Design and Analysis of Composite Rotor Blades For ActivePassive Vibration Reduction

Design and Analysis of Composite Rotor Blades for
Active/Passive Vibration Reduction
by
Devesh Kumar
A dissertation submitted in partial fulfillment

of the requirements for the degree of
Doctor of Philosophy
(Aerospace Engineering)
in The University of Michigan
2013
Doctoral Committee:
Professor Carlos E. S. Cesnik, Chair

Professor Peretz P. Friedmann
Associate Professor Joaquim R. R. A. Martins
Professor Kon-Well Wang
 Devesh Kumar, 2013
All Rights Reserved
To my parents, sister and brother
ii
ACKNOWLEDGEMENTS
I would like to extend my sincere thanks to my family, teachers, friends and
colleagues who helped me immensely during my days in the graduate school.
First of all, I would like to thank my family for providing unyielding support and
encouragement over the years. They not only provided me with the best opportunities but
also created a protected environment around to help me to succeed. The values instilled
in me by my family have been a constant guiding force for me. I hope the completion of
my thesis gives them some happiness and relief which they truly deserve.
Next, I would like to thank Prof. Cesnik for being my advisor and mentor during the
course of my long and uneventful (fortunately!) doctoral research. I appreciate the
encouragement and opportunities that he has provided me at different times during the
graduate school to help me learn better, to grow as a researcher and to build my
confidence. I am also very grateful for the patience and effort that he has put in
improving my writing and presentation skills to reach the high standards that he has set.
His enthusiasm for research and the ability to pay attention to finer details has been very
inspiring. Besides these, his passion for sports, flying, food and johnnies has always been
very contagious.
I deeply appreciate the time and effort and useful feedback from my thesis committee
members, Prof. Friedmann, Prof. Martins and Prof. Wang. I would specially like to
acknowledge encouragement and guidance provided by Prof. Friedmann during my
iii
research and graduate courses. I would also like to thank my collaborators at Advatech
Pacific Inc, Dr. Peter Röhl and Mark Sutton, for their help in the implementation of
ModelCenter solution. I am very grateful to Dr. Bryan Glaz for his help in the
implementation of the basic surrogate-based optimization technique. I also benefitted
greatly from technical discussions with Dr. Li Liu, Dr. Rafael Palacios, and Dr. Ashwani
Padthe.
My research group, A2SRL, has always been very supportive and I have had the
opportunity to learn something new from each of my group members. I wish to thank Dr.
Weihua Su, Dr. Jiwon Mok, Dr. Satish Chimakurthi, Dr. Ken Salas, Dr. Torsten Skujins,
Dr. Nate Falkiewicz, Kalyan Nadella, Smith Thepvongs, Ben Hallisy, Matt Dillsaver,
Matt Obenchain, Jessica Jones, and members of X-HALE team (Blake Davis, Elizabeth
Prentice, etc.) for making the research in a windowless office and basement lab a very
enjoyable experience. I would specially like to thank Dr. Jiwon Mok and Smith
Thepvongs for help with their numerical codes, and Dr. Ken Salas for help in the
experimental laboratory.
The experimental study presented in my thesis would not have been possible without
the help of the technical staff in the Department of Aerospace Engineering. I am very
grateful to Terry “it will never work” Larrow for being patient with me and for
fabricating all my crazy and tiny parts multiple times. I would like to thank Eric Kirk,
Chris Chartier, and Thomas Griffin for being readily available whenever required and for
providing useful suggestions. I would also like to thank Amit Salvi for his help with
tensile testing of fabricated blade sections.
iv
My long stay in Ann Arbor was made wonderful by friends around me. I would like to
thank my friend Dr. Pavana Prabhakar for her joyous company, for being part of
innumerable happy memories and for smoothing out the difficult part of my PhD. I would
like to thank my friend Dr. Gayathri Seenumani for trusting in me and for giving me lot
of confidence when it was needed the most. I also wish to thank Kalyan Nadella, Dr.
Rajeev Verma and Pritam Sukumar for their friendship and belief in me. I will also
cherish the time spent with my friends: Dr. Christian Heinrich, Dr. Saumil Ambani,
Nicholas Lamorte, Dr. Daniel Zaide, Dr. Qing Zhu, Nhung Nguyen, Karthik
Ananthraman, Dipankar Dua, Dr. Amit Kaushik and my lunch group.
And finally, I would like to acknowledge the financial support provided by Georgia
Tech/University of Michigan/Washington University Vertical Lift Rotorcraft Center of
Excellence (VLRCOE), funded by the U.S. Army, with Dr. Michael Rutkowski as the
technical monitor. This work was also supported by U.S. Army Research, Development
and Engineering Command (AMRDEC), Aviation Applied Technology Directorate
(AATD) through SBIR Phase II contract W911W6-10-C-0030.
v
TABLE OF CONTENTS
DEDICATION................................................................................................................... ii
ACKNOWLEDGEMENTS ............................................................................................ iii
LIST OF FIGURES .......................................................................................................... x
LIST OF TABLES ....................................................................................................... xviii
LIST OF APPENDICES .............................................................................................. xxii
ABSTRACT .................................................................................................................. xxiv
CHAPTER 1. INTRODUCTION................................................................................. 1
1.1 Introduction to Vibration in Helicopters ................................................................ 1
1.2 Methods for Vibration Reduction ........................................................................... 5

1.2.1 Passive Approaches ............................................................................................ 6
1.2.2 Active Approaches .............................................................................................. 7
1.2.3 Active Vibration Control in Fuselage ................................................................. 8
1.2.4 Vibration Control with Blade Pitch Actuation.................................................... 9
1.2.5 Vibration Control with on-Blade Actuators ...................................................... 10
1.2.5.1 Discrete Actuators ..................................................................................... 12
1.2.5.2 Continuous/Embedded Actuators ............................................................. 13
1.3 Literature Review Relevant to the Thesis ............................................................. 14

1.3.1 Active Twist Rotors .......................................................................................... 15
1.3.1.1 Modeling of Active Twist ......................................................................... 15
1.3.1.2 Design and Parametric Studies ................................................................. 17
1.3.1.3 Active Twist Optimization ........................................................................ 18
1.3.1.4 Hybrid Active/Passive Optimization ........................................................ 19
1.3.2 Passive Optimization ........................................................................................ 20
1.3.3 Active Flaps ...................................................................................................... 21
1.3.3.1 Cross-sectional Design for Composite Blade with Active Flaps .............. 22
1.3.3.2 Dual/Multiple Trailing Edge Active Flaps ............................................... 22
1.3.3.3 Piezoelectric Actuator for Active Flaps .................................................... 23
1.3.1 Optimization for Composite Structures ............................................................ 25
1.3.2 Camber Actuation ............................................................................................. 26
1.4 Objectives of the Dissertation ................................................................................ 27
vi
1.5 Outline of the Thesis ............................................................................................... 30
CHAPTER 2. A MULTIDISCIPLINARY DESIGN ENVIRONMENT FOR

PASSIVE COMPOSITE ROTOR BLADES ............................................................... 34
2.1 Introduction ............................................................................................................. 34
2.2 Design Approach ..................................................................................................... 37

2.2.1 Blade Modeling Tool IXGEN ........................................................................... 39
2.2.2 UM/VABS ........................................................................................................ 41
2.2.3 RCAS ................................................................................................................ 42
2.2.4 ModelCenter Integration ................................................................................... 43
2.3 Application Example: Vibratory Hub Load Minimization ................................ 44

2.3.1 Description of the Baseline Rotor Blade .......................................................... 45
2.3.2 Definition of the Rotor Blade Optimization Problem ....................................... 49
2.3.3 Preliminary Parametric Study ........................................................................... 51
2.4 Optimization Studies for Vibration Reduction .................................................... 52

2.4.1 Minimizing Vertical Component of the 4/rev Vibratory Hub Load (min FZ4) . 53
2.4.2 Minimizing Combined Vibratory Hub Load (min FH4) .................................... 61
2.5 Concluding Remarks .............................................................................................. 64
CHAPTER 3. OPTIMIZATION FRAMEWORK FOR THE DYNAMIC

ANALYSIS AND DESIGN OF ACTIVE TWIST ROTORS ..................................... 66
3.1 Optimization Framework ....................................................................................... 67

3.1.1 High-Fidelity Analysis Framework .................................................................. 68
3.1.2 Surrogate-based Optimization .......................................................................... 69
3.1.3 EGO Algorithm ................................................................................................. 70
3.2 Numerical Studies ................................................................................................... 71
3.3 Optimization Results .............................................................................................. 75

3.3.1 Construction and Verification of the Surrogate Models ................................... 76
3.3.2 Optimization Cases ........................................................................................... 77
3.3.2.1 Static Twist Optimization (Max θstat)........................................................ 77
3.3.2.2 Dynamic Twist Optimization (Max θdyn,4/rev)............................................ 78
3.3.3 Analysis of Optimized Results.......................................................................... 81
3.3.3.1 Relation between blade twist and induced FZ4.......................................... 81
3.3.3.2 Effect of Frequency of Actuation ............................................................. 83
3.3.3.3 Effect on Vibratory Loads in Forward Flight Conditions for Zero Twist
Actuation …………………………………………………………………………...84
3.3.3.4 Circle Plot for the Optimized Results ....................................................... 85
3.3.3.5 Effect of Aeromechanic Analysis on Optimized Results ......................... 87
3.3.4 Trend Analysis .................................................................................................. 92
vii
3.4 Optimization at 3/rev Actuation Frequency (Max θ3/rev) .................................... 94
3.5 Optimization at 5/rev Actuation Frequency (Max θ5/rev) .................................... 95
3.6 Optimization at a Range of Actuation Frequencies (Max θ345/rev) ...................... 97
3.7 Effect of Advance Ratio ........................................................................................ 101
3.8 Concluding Remarks ............................................................................................ 103
CHAPTER 4. MIXED-VARIABLE OPTIMIZATION FOR DESIGN OF

ACTIVE TWIST ROTOR BLADES .......................................................................... 106
4.1 Architecture for New Mixed-variable Optimization Framework .................... 107
4.2 Optimization with Normalized Ply Thickness .................................................... 114

4.2.1 Optimization with Continuous Design Variables ........................................... 117
4.2.2 Optimization with Mixed Design Variables ................................................... 122
4.2.2.1 Maximizing θstat ...................................................................................... 123
4.2.2.2 Maximizing θ3/rev..................................................................................... 124
4.2.2.3 Maximizing θ4/rev..................................................................................... 126
4.2.2.4 Maximizing θ5/rev..................................................................................... 127
4.2.2.5 Maximizing θ345/rev .................................................................................. 129
4.3 Optimization Study with 8 Design Variables ..................................................... 130

4.3.1 Optimization Results with Continuous Design Variables .............................. 132
4.3.2 Optimization Results with Mixed Design Variables ...................................... 135
4.4 Optimization Studies with Ply Angles and Ply Thicknesses ............................. 137
4.4.1 Optimization Results with Continuous Design Variables .............................. 138
4.4.2 Optimization with Mixed Design Variables ................................................... 140
4.5 Post Processing of Optimization Results............................................................. 142

4.5.1 Effect on Baseline Vibration ........................................................................... 143
4.5.2 Effect of Advance Ratio.................................................................................. 144
4.5.3 Circle Plot for Optimized Cases ..................................................................... 146
CHAPTER 5. NEW STRATEGY FOR DESIGN OF COMPOSITE ROTOR

BLADE WITH ACTIVE FLAPS ................................................................................ 152
5.1 Aeroelastic Analysis with Active Flaps ............................................................... 152
5.2 Baseline Rotor Blade............................................................................................. 153
5.3 Preliminary Analysis ............................................................................................ 156
viii
5.4 Optimization Problem Definition ........................................................................ 160
5.5 Optimization Results ............................................................................................ 163
CHAPTER 6. PERFORMANCE ENHANCEMENT AND VIBRATION

REDUCTION IN DYNAMIC STALL CONDITION USING ACTIVE CAMBER
DEFORMATION .......................................................................................................... 174
6.1 Optimization Framework ..................................................................................... 175

6.1.1 Surrogate-based Optimization (SBO) ............................................................. 176
6.1.2 Gradient-based Optimization .......................................................................... 177
6.1.3 Aeroelastic Framework (UM/NLABS-A) ...................................................... 177
6.2 Camber Deformation Shape ................................................................................ 182
6.3 Preliminary Numerical Results ........................................................................... 184

6.3.1 Effect of Camber Actuation on MZ0 ............................................................... 186
6.3.2 Circle Plot for Different Actuation Frequencies ............................................. 187
6.3.3 Effect of Amplitude of Actuation ................................................................... 189
6.4 Optimization Problem Definition ........................................................................ 190

6.4.1 Optimization Results....................................................................................... 191
6.4.2 Advantage of Two-step Optimization Process ............................................... 193
6.4.3 Effect of Optimization on other Hub Load Components ................................ 196
6.4.4 Analysis of Optimized Cases .......................................................................... 198
6.4.5 Pareto Optimization ........................................................................................ 201
CHAPTER 7. CONCLUSIONS AND FUTURE WORK ...................................... 204
7.1 Summary ................................................................................................................ 204
7.2 Main Results .......................................................................................................... 206
7.3 Key Contribution .................................................................................................. 210
7.4 Future Work .......................................................................................................... 211
APPENDICES ............................................................................................................... 214
REFERENCES .............................................................................................................. 359
ix
LIST OF FIGURES
Figure 1-1: Trends for Helicopter Vibration Levels [3] ..................................................... 2

Figure 1-2: Vibration Amplitude Spectrum for BO 105 in Level Cruise Flight [4] ........... 3
Figure 1-3: Unsteady Aerodynamics on the Main Rotor Disk ........................................... 4
Figure 1-4: Vertical Cabin Vibration in BO 105 as a function of Airspeed [6] ................. 5
Figure 1-5: Methods used for Vibration Reduction in Rotor Blades .................................. 6
Figure 1-6: Area of Focus for the Thesis .......................................................................... 28
Figure 2-1: Different Stages of Product Design [100] ...................................................... 36
Figure 2-2: 2D - 1D Design Approach.[148] .................................................................... 38
Figure 2-3: Multidisciplinary Optimization Process for Design of Passive Blade. .......... 39
Figure 2-4: Representative Cross Sections Developed using IXGEN [148] .................... 40
Figure 2-5: Basic Process of UM/VABS [71] .................................................................. 42
Figure 2-6: ModelCenter-Based Blade Optimization Process. ......................................... 43
Figure 2-7: Top View of the UH-60 Rotor Blade [157] ................................................... 45
Figure 2-8: Cross Section of the UH-60 Rotor Blade ....................................................... 46
Figure 2-9: Cross Section Layup ...................................................................................... 47
Figure 2-10: Contribution of each Ply Thickness to Overall Variation in the Response . 53
Figure 2-11: Variation of Objective Function (FZ4) with Generation Number................. 56
Figure 2-12: Variation of Ply Thickness and Ply Angle with Generation Number .......... 58
Figure 2-13: Variation of Ballast Mass and Auxiliary Spar Web with Generation Number
........................................................................................................................................... 59
Figure 2-14: Variation of Constraints with Generation Number ...................................... 60
Figure 2-15: Percentage Change in Amplitude of 4/rev Vibratory Hub Load Components
for the Optimized Case (min FZ4) with respect to the Baseline Case ............................... 60
Figure 2-16: Variation of FH4 with Generation Number ................................................... 62
Figure 2-17: Percentage Change in Vibratory 4/rev Hub Loads for the Optimized Cases64
Figure 3-1: Optimization Framework for Designing Active Twist Rotor Blades ............ 67
Figure 3-2: Analysis Framework for Active Twist Rotors ............................................... 68
Figure 3-3: Planform View of the ATR Rotor Blade (Dimensions in Meters) [62] ......... 72
Figure 3-4: Cross-Sectional Shape of the Rotor Blade (NACA 0012 Airfoil) ................. 73
Figure 3-5: Baseline and the Optimized Cross Sections ................................................... 80
x
Figure 3-6: Objective Function Results in Order of Increasing Static Twist ................... 82
Figure 3-7: Objective Function Results in Order of Increasing Dynamic Twist .............. 82
Figure 3-8: Effect of Frequency on Dynamic Twist ......................................................... 83
Figure 3-9: Effect of Frequency on Amplitude of Vertical Tip Displacement ................. 84
Figure 3-10: Percentage Increase in 4/rev Vibratory Loads at the Rotor Hub with no
Actuation at μ = 0.24 ........................................................................................................ 85
Figure 3-11: Vibratory Hub Vertical Shear Force (μ = 0.24) ........................................... 86
Figure 3-12: Vibratory Hub Lateral Cyclic Moment (μ = 0.24) ....................................... 86
Figure 3-13: Variation of Mean Value and Amplitude of Tip Twist for twist actuation at
μ=0.0 ................................................................................................................................. 88
Figure 3-14: Variation of Amplitude with Torsional Stiffness and 1st Torsion Frequency
(sorted with respect to amplitude obtained from low density analysis)............................ 93
Figure 3-15: Variation with Iteration Number for Other Parameters (sorted with respect to
amplitude obtained from low density analysis) ................................................................ 93
Figure 3-16: Variation of Dynamic Twist Amplitude with Torsion Frequency ............. 101
Figure 3-17: Effect of Advance ratio on Dynamic Twist Amplitude ............................. 103
Figure 4-1: Augmented Optimization Framework for Continuous/Discrete Design
Variables ......................................................................................................................... 108
Figure 4-2: Variation of objective function with Iteration number for optimization with
Continuous Design Variables ......................................................................................... 120
Figure 4-3: Variation of Constraints for Max θ4/rev Optimization .................................. 121
Figure 4-4: Variation of Design Variables for Max θ4/rev Optimization ......................... 121
Figure 4-5: Modified Baseline Case (Baseline 2) ........................................................... 131
Figure 4-6: Cross Section for the Optimized Cases obtained with Mixed Design Variables
......................................................................................................................................... 142
Figure 4-7: Percentage Increase in Vibratory Loads ...................................................... 144
Figure 4-8: Effect of advance ratio at 3/rev actuation frequency ................................... 145
Figure 4-9: Effect of Advance Ratio at 4/rev Actuation Frequency ............................... 145
Figure 4-10: Effect of Advance Ratio at 5/rev Actuation Frequency ............................. 146
Figure 4-11: Circle Plot for 3/rev Actuation Frequency ................................................. 147
Figure 5-1: University of Michigan Spin Test Stand ...................................................... 153
Figure 5-2: Cross-Sectional Layup for the Baseline Rotor Blade .................................. 154
Figure 5-3: Active Flap Mechanism ............................................................................... 155
xi
Figure 5-4: Variation of Tip Twist with Actuation Frequency ....................................... 157
Figure 5-5: Variation of Amplitude of Tip Twist with Advance Ratio .......................... 158
Figure 5-6: Variation of Trim Variables with Advance Ratio ........................................ 158
Figure 5-7: Circle Plot for 4/rev Actuation Frequency ................................................... 160
Figure 5-8: Spanwise Location for Active Flaps ............................................................ 161
Figure 5-9: Circle Plot at rf = 0.78R ............................................................................... 166
Figure 5-10: Circle Plot at 0.7R ...................................................................................... 168
Figure 5-11: Circle Plot for Optimized Result at rf = 0.85R........................................... 170
Figure 5-12: Optimized Results for different Spanwise Locations ................................ 170
Figure 5-13: Optimized Cross Sections .......................................................................... 171
Figure 6-1: Two-step Optimization Process ................................................................... 176
Figure 6-2: UM/NLABS-A Framework [145] ................................................................ 178
Figure 6-3: Schematic of Unified Airloads Model ......................................................... 179
Figure 6-4: Camber Deformation Shape Function.......................................................... 183
Figure 6-5: Airfoil Cross Section with 5% Camber Deformation .................................. 183
Figure 6-6: Aerodynamic Properties of Cambered and Baseline Airfoil Section .......... 185
Figure 6-7: Variation of Camber Deformation along the Blade Span ............................ 186
Figure 6-8: Effect of Camber Actuation on MZ0 at µ = 0.33 .......................................... 187
Figure 6-9: Circle Plot for FZ4......................................................................................... 188
Figure 6-10: Circle Plot for MX4 ..................................................................................... 188
Figure 6-11: Effect of Amplitude on Circle Plot for FZ4 ................................................ 189
Figure 6-12: Effect of Amplitude of Actuation on MZ0 .................................................. 190
Figure 6-13: Amplitude of Actuation for Optimized Cases............................................ 192
Figure 6-14: Phase of Actuation for Optimized Cases ................................................... 193
Figure 6-15: Variation of Objective Function with Iteration Number for SBO ............. 194
Figure 6-16: Variation of Design Variables with Iteration Number for SBO for Min MZ0
case for analysis with Ψcubic ............................................................................................ 194
Figure 6-17: Percentage Reduction in FZ4 using Two-Step Optimization Process ......... 195
Figure 6-18: Variation of Angle of Attack for the Baseline Case (Units: Deg) ............. 198
Figure 6-19: Difference in Angle of Attack for the Optimized Cases (Unit: Deg) ........ 199
Figure 6-20: Variation of Angle of Attack at r = 0.74R ................................................. 200
Figure 6-21: Variation of Camber Deformation for the Optimized Cases (Unit: %c) ... 200
Figure 6-22: Camber Deformation at the Blade Tip ....................................................... 201
xii
Figure 6-23: Pareto Front for Vibration Reduction and Performance Enhancement ..... 202
Figure A-1: Correlation Functions .................................................................................. 217
Figure B-1: Unified Airloads Model............................................................................... 223
Figure B-2: General Airfoil Coordinate System ............................................................. 224
Figure B-3: Static Stall Residual..................................................................................... 231
Figure B-4: Direction of Aerodynamic Forces ............................................................... 232
Figure C-1: Schematic of Quasi-Static Test ................................................................... 237
Figure C-2: Experimental Setup for Quasi-Static Tests ................................................. 237
Figure C-3: Cage Region for holding the Actuator ........................................................ 238
Figure C-4: Characterization of X-frame Actuators ....................................................... 240
Figure C-5: Location of Flaps and Actuators ................................................................. 241
Figure C-6: VR7 airfoil with a Plain Flap ...................................................................... 243
Figure C-7: Grids for the Airfoil with Flap .................................................................... 243
Figure C-8: Pressure Contour for α=4° and M=0.538 .................................................... 243
Figure C-9: Hinge Moment Curve Slope (CHδ) .............................................................. 244
Figure C-10: Variation of Lift Coefficient due to Flap Deflection ................................. 245
Figure C-11: Variation of CLδ with Hinge Location ....................................................... 245
Figure C-12: Schematic of the Dual Flap Section of Rotor Blade ................................. 246
Figure C-13: Detailed View of the Flap Supports .......................................................... 247
Figure C-14: Ideal operating point obtained using impedance matching ....................... 253
Figure C-15: Effect of Moment Arm Modification on Operating Point ......................... 254
Figure C-16: CAD Model of the Flap Parts .................................................................... 255
Figure C-17: Actual Fabricated Parts for Flap Hinge Mechanism ................................. 256
Figure C-18: Setup for Bench Test of Active Flap ......................................................... 256
Figure C-19: Hysteresis in Flap Actuation ..................................................................... 257
Figure D-1: Blade plan-form View [58] ......................................................................... 260
Figure D-2: Blade Cross Section .................................................................................... 261
Figure D-3: Blade Twist Distribution ............................................................................. 261
Figure D-4: Cross Section of the Rotor Blade ................................................................ 262
Figure D-5: CH47D Rotor Blade with Dual Active Flaps .............................................. 262
Figure D-6: Finite Element Mesh for Root Section ........................................................ 264
Figure D-7: Finite Element Mesh for Main Blade Sections ........................................... 265
Figure D-8: Location of Hinges in Spin Test Stand ....................................................... 265
xiii
Figure D-9: Steps used in Blade Design ......................................................................... 266
Figure D-10: CAD Model of the Metal insert used for Pull Tests.................................. 275
Figure D-11: Set up for pull test of blade root section ................................................... 276
Figure D-12: Location of Strain Gages used for the Pull Test of Root Section ............. 276
Figure D-13: Damaged Section after the Pull Test for Root Section ............................. 277
Figure D-14: Blade Loading Profile used for the Pull Test of Root Section .................. 278
Figure D-15: Strain Recorded by different Strain Gages during the Blade Loading...... 278
Figure D-16: Test Section Fabricated for Pull Test of Cutout Section ........................... 279
Figure D-17: Setup used for the Pull Test for the Cutout Region .................................. 279
Figure D-18: Location of Strain Gages used for the Pull Test of Cutout region ............ 280
Figure D-19: Loading Cycle used for Pull Test of Cutout Region ................................. 281
Figure D-20: Strain Gage Output .................................................................................... 281
Figure D-21: Instrumentation used in the Active Blade ................................................. 283
Figure E-1: Exploded View of the Airfoil Cross Section ............................................... 285
Figure E-2: Assembled View of the Airfoil Cross Section............................................. 286
Figure E-3: Shape of the Foam Core for Spar and Fairing Section ................................ 287
Figure E-4: Joined pieces of Foam Core......................................................................... 288
Figure E-5: Cutout made in the Spar Region for Actuators............................................ 288
Figure E-6: Instrumented Spar Section ........................................................................... 289
Figure E-7: Trough made in the Blade Spar for Instrumentation Wires......................... 290
Figure E-8: Wiring Diagram for Full Bridge Flap wise Bending Strain Gage ............... 290
Figure E-9: Wiring diagram for Full Bridge Torsion Strain Gage ................................. 290
Figure E-10: High Voltage Wires in the Cutout Region................................................. 291
Figure E-11: Accelerometers mounted on the Blade Tip in the Spar Region ................ 291
Figure E-12: Different Configurations used for Calibration of Accelerometer .............. 292
Figure E-13: Spar Mandrel Parts .................................................................................... 293
Figure E-14: CAD Model Developed for the Parts of Spar Mandrel ............................. 293
Figure E-15: Cutout made in Spar ply 1, 3, 4, 5 and 6. .................................................. 297
Figure E-16: Cut Prepreg Plies prior to Layup ............................................................... 297
Figure E-17: Adhesive Film wrapped around Foam Core .............................................. 298
Figure E-18: Additional Plies in the Cutout Region ....................................................... 298
Figure E-19: Cutout Region with Ribs ........................................................................... 298
Figure E-20: Leading Edge Ballast Mass ....................................................................... 300
xiv
Figure E-21: Wrapping Root plies around the Root Pin ................................................. 300
Figure E-22: All Root Plies wrapped around the Root Pin ............................................. 300
Figure E-23: Leading Edge Weights and Spar web plies ............................................... 301
Figure E-24: Cutout Region with Spar Mandrel and Additional Leading Edge Weights
......................................................................................................................................... 301
Figure E-25: First main Spar ply Wrapped Around Spar in the Cutout Region ............. 302
Figure E-26: First Main Spar Ply near the Root Region ................................................. 302
Figure E-27: 4th Spar Ply in the Cutout Region ............................................................. 302
Figure E-28: Unidirectional Plies moved around the Cutout Region ............................. 303
Figure E-29: Cutout Region with All Spar Plies ............................................................ 303
Figure E-30: cutout region with spacer and peel plies .................................................... 304
Figure E-31: Bottom part of the cutout region................................................................ 304
Figure E-32: Root region before spar cure ..................................................................... 305
Figure E-33: Instrumentation wires exiting the mold ..................................................... 305
Figure E-34: Blade Molds Closed with Heavy Duty Steel Clamps ................................ 306
Figure E-35: Active Blade Spar after Cure ..................................................................... 306
Figure E-36: Cutout Region after Cure ........................................................................... 307
Figure E-37: Spar Region after removing Spar Mandrel ................................................ 307
Figure E-38: Holes machined in the bottom mold for alignment of flaps during the cure
......................................................................................................................................... 308
Figure E-39: CAD model of the flap mandrel used during fairing cure ......................... 308
Figure E-40: Location of Flap Supports mounted in the Flap Region during Fairing Cure
......................................................................................................................................... 309
Figure E-41: Root Region with Fairing Foam Core ....................................................... 310
Figure E-42: Instrumented Flap Region for Fairing Cure with Flap Mandrel ................ 311
Figure E-43: Flap Region with Additional Plies for holding Flap Supports .................. 311
Figure E-44: Flap Region with Fairing Plies .................................................................. 312
Figure E-45: Bottom Part of the Fairing Region ............................................................ 312
Figure E-46: Fairing region before Final Cure ............................................................... 313
Figure E-47: Root part of the Blade before Fairing Cure ............................................... 313
Figure E-48: Flap part of the Blade after Cure ............................................................... 314
Figure E-49: Cured Active Blade ................................................................................... 314
Figure E-50: Cured Passive Blade .................................................................................. 315
xv
Figure E-51: Ballast mass added in passive blade instead of actuator ........................... 316
Figure E-52: Cross-sectional Shape of the Flap ............................................................. 316
Figure E-53: CAD Model of the Active Flap ................................................................. 317
Figure E-54: Flap section before cure ............................................................................. 318
Figure E-55: Flap inside the Mold before Cure .............................................................. 318
Figure E-56: Flap inside the Autoclave for Cure ............................................................ 318
Figure E-57: Cured Flap Sections ................................................................................... 319
Figure E-58: CAD model for the Metal Insert ................................................................ 320
Figure E-59: Metal Insert Attached to the Foam Core ................................................... 320
Figure E-60: Fabricated Section for Pull Test ................................................................ 321
Figure F-1: Spin-test Stand ............................................................................................. 323
Figure F-2: Axes Convention for the Load Cell ............................................................. 324
Figure F-3: Setup for Data Acqisition and Power Supply .............................................. 327
Figure F-4: Aluminum Fixture for Static Balancing....................................................... 328
Figure F-5: Static Balancing of Blades ........................................................................... 328
Figure F-6: Schematic of Blade Tracking ....................................................................... 330
Figure F-7: Time Averaging of Data .............................................................................. 332
Figure F-8: Effect of Averaging on FFT......................................................................... 333
Figure F-9: FFT of the Flap Deflection .......................................................................... 336
Figure F-10: Variation of Mean Loads for Base1 ........................................................... 336
Figure F-11: Variation of Mean Loads for Different Cases ........................................... 337
Figure F-12: FFT of Fz for the Baseline Cases ............................................................... 338
Figure F-13: FFT of Mz for the Baseline Cases .............................................................. 339
Figure F-14: Difference in FFT for Fz Component......................................................... 340
Figure F-15: Difference in FFT for Mz component ........................................................ 340
Figure F-16: Comparison for Mean Loads ..................................................................... 341
Figure F-17: Comparison for Fz component for Case 1.................................................. 343
Figure F-18: Comparison for Fz component for Case 5.................................................. 343
Figure F-19: Comparison for Mz component for Case 1 ................................................ 344
Figure F-20: Comparison for Mz component for Case 5 ................................................ 344
Figure F-21: Flap Deflection to Input Voltage Transfer Function for Flap 1 ................. 346
Figure F-22: Flap Deflection to Input Voltage Transfer Function for Flap 2 ................. 346
Figure F-23: Hub Thrust (Fz) to Input Voltage Transfer Function ................................. 347
xvi
Figure F-24: Hub Torque (Mz) to Input Voltage Transfer Function............................... 347
Figure F-25: Variation of CT for Baseline Conditions [177] .......................................... 350
xvii
LIST OF TABLES
Table 1-1: Active Flaps Implemented on Full Scale Blades............................................. 23

Table 1-2: Active Controlled Flaps Tested on a Model Scale Rotor in Whirl Tower or
Wind Tunnel ..................................................................................................................... 24
Table 1-3: Conceptual Designs for Active Flaps .............................................................. 24
Table 2-1: Characteristics of UH-60 Rotor ....................................................................... 45
Table 2-2: Material Properties .......................................................................................... 47
Table 2-3: Thickness of Plies used in the Layup .............................................................. 48
Table 2-4: Structural Frequencies of the Blade at 100% RPM ......................................... 49
Table 2-5: Mean Value and Amplitude of 4/rev Vibratory Loads at the Rotor Hub for µ =
0.24.................................................................................................................................... 49
Table 2-6: Variation in Blade Parameters Observed During the Parametric Study with Ply
Thickness .......................................................................................................................... 51
Table 2-7: Design Variables used in the Optimization Study for Min FZ4 ....................... 54
Table 2-8: Constraints used in the Optimization Problem for Min FZ4 ............................ 55
Table 2-9: Parameters used for NSGA II Algorithm ........................................................ 56
Table 2-10: Design Variables and Constraints for the Optimized Case ........................... 57
Table 2-11: Ratio of Cross-Sectional Properties for the Optimized Case ........................ 58
Table 2-12: Blade Structural Frequencies for the Optimized Case .................................. 58
Table 2-13: Design Variables and Constraints for Min FH4 ............................................. 63
Table 2-14: Cross-sectional Properties for the Optimized Case (min FH4)....................... 63
Table 2-15: Structural Frequencies for the Optimized Case ............................................. 64
Table 3-1: Baseline ATR Cross Section Ply Angles ........................................................ 73
Table 3-2: Material Properties .......................................................................................... 74
Table 3-3: Characteristics of the Baseline ATR ............................................................... 74
Table 3-4: Hub Loads for the Baseline ATR Case (μ = 0.24) .......................................... 75
Table 3-5: Design Variable Used for Optimization Study................................................ 76
Table 3-6: Constraints used in Optimization .................................................................... 76
Table 3-7: Error Obtained in the Prediction of Design Variable and Constraints for
Surrogate models .............................................................................................................. 77
Table 3-8: Static Twist Optimization Result .................................................................... 78
xviii
Table 3-9: Dynamic Optimization Results ....................................................................... 80
Table 3-10: Optimized Results ......................................................................................... 89
Table 3-11: Optimization Constraints and Other Parameters ........................................... 90
Table 3-12: Design Variables for the Optimized Cases.................................................... 91
Table 3-13: Optimized Results for 3/rev Actuation Frequency ........................................ 94
Table 3-14: Optimization Results for 5/rev Actuation Frequency.................................... 96
Table 3-15: Results obtained from all the Optimization Cases ........................................ 99
Table 3-16: Optimization Results for 3, 4 and 5/rev Actuation Frequencies ................. 100
Table 4-1: Modified Bounds for Constrained Mixed-variable Genetic Optimization.... 112
Table 4-2: Modified Set of Design Variables for Sequential Gradient-Based Optimization
......................................................................................................................................... 113
Table 4-3: Parameters used in GA Optimization ............................................................ 114
Table 4-4: Optimization Parameters for GBO ................................................................ 114
Table 4-5: Design Variables and their Bounds ............................................................... 115
Table 4-6: Constraints for Optimization Problem .......................................................... 115
Table 4-7: Final Result obtained from Optimization Studies ......................................... 116
Table 4-8: Percentage Difference between the Objective Function for Continuous
Variable Optimization and Mixed-variable Optimization .............................................. 117
Table 4-9: Constraints and Design Variables for Optimization with Continuous Design
Variables ......................................................................................................................... 118
Table 4-10: Performance of Optimized Cases at other Actuation Frequencies .............. 122
Table 4-11: Optimization Results for Maximizing θstat .................................................. 124
Table 4-12: Optimization Results for Maximizing θ3/rev ................................................ 125
Table 4-13: Optimization Results for Maximizing θ4/rev ................................................ 126
Table 4-14 Optimization Results for Maximizing θ5/rev.................................................. 128
Table 4-15: Optimization Results for Maximizing θ345/rev .............................................. 129
Table 4-16: Design Variable and their Bounds............................................................... 131
Table 4-17: Results obtained for Optimization with 8 Design Variables ....................... 132
Table 4-18: Results Obtained with Continuous Design Variables ................................. 133
Table 4-19: Optimization Results with Mixed Design Variables ................................... 136
Table 4-20: Design Variables for Optimization with Ply Thicknesses and Ply Angles . 138
Table 4-21: Results for Optimization with Continuous Design Variables ..................... 139
Table 4-22: Results Obtained for Optimization with Mixed Design Variables.............. 141
xix
Table 5-1: Rotor Properties ............................................................................................. 154
Table 5-2: Flap and Actuator Dimensions ...................................................................... 155
Table 5-3: Structural Frequencies of the Rotor Blade .................................................... 156
Table 5-4: Design Variable used for Optimization Study .............................................. 162
Table 5-5: Constraints used in the Optimization ............................................................ 162
Table 5-6: Results for Flap centered at rf = 0.78R .......................................................... 164
Table 5-7: Results at rf = 0.7R ........................................................................................ 167
Table 5-8: Results for Flap at 0.85R ............................................................................... 169
Table 6-1: Baseline Blade Properties .............................................................................. 185
Table 6-2: Range for Design Variable used in the Optimization Study ......................... 191
Table 6-3: Optimization Results ..................................................................................... 192
Table 6-4: Optimum Design Variables for min FZ4 Case ............................................... 196
Table 6-5: Variation in Hub Loads for the Optimized Cases ......................................... 196
Table C-1: Comparison between X-frame Actuator and APA Actuators....................... 236
Table C-2: Variation of Load stiffness with the Length of Wire .................................... 239
Table C-3: Actuator Stiffness obtained from Quasi-Static Tests (Units: lbf/in) ............ 240
Table C-4: Difference between Ideal and Actual Operating Condition ......................... 254
Table C-5: Experimental Results for Bench Tests .......................................................... 257
Table D-1: Blade Properties............................................................................................ 260
Table D-2: Spanwise Regions for Cross-sectional Analysis .......................................... 263
Table D-3: Initial and Final Blade Design ...................................................................... 269
Table D-4: Cross-sectional properties for Section 1F and Section 2F............................ 270
Table D-5: Dynamic Blade Frequencies ......................................................................... 270
Table D-6: Maximum Strains for the Baseline Case (Units: με) .................................... 271
Table D-7: Maximum Blade Strain for the Final Design (Units: με) ............................. 272
Table D-8: Percentage Variation in Cross-Sectional Strains .......................................... 272
Table D-9: Alternating Strains for the Baseline Case (Units: με) .................................. 273
Table D-10: Alternating Strains for the Final Design (Units: με) .................................. 273
Table D-11: Percentage Difference for the Alternating Strains...................................... 274
Table D-12: Details of the Sensors used on the Active Blade ........................................ 282
Table E-1: Instrumentation used in the Spar Section...................................................... 289
Table E-2: Measured Voltage from Accelerometers ...................................................... 292
Table E-3: Dimension of the Plies cut prior to Blade Fabrication .................................. 294
xx
Table E-4: Ballast Mass used in each Cross Section ...................................................... 299
Table F-1: Spin-test Stand Characteristics ...................................................................... 323
Table F-2: Maximum loads ............................................................................................. 324
Table F-3: Preliminary Accuracy and Resolution of Various Load Components .......... 325
Table F-4: List of working Sensors ................................................................................ 325
Table F-5: Test Matrix for Analysis at Different Frequency .......................................... 331
Table F-6: Mean value of Flap Deflection ...................................................................... 335
xxi
LIST OF APPENDICES
APPENDIX A. SURROGATE BASED OPTIMIZATION ..................................... 215
A.1 Construction of Surrogate ................................................................................ 215
A.2 EGO Algorithm ................................................................................................. 220
APPENDIX B. UM/NLABS-A AERODYNAMICS MODEL ................................ 223
B.1 Peters Flexible Airfoil Theory.......................................................................... 223
B.2 Dynamic Inflow Model ..................................................................................... 227
B.3 Dynamic Stall Model......................................................................................... 228
APPENDIX C. DESIGN OF ACTIVE FLAP........................................................... 233
C.1 Quasi-static Tests .............................................................................................. 237
C.2 Dual-flap Design ................................................................................................ 240

C.2.1 Aerodynamic Hinge Moment ......................................................................... 242
C.2.2 Sizing of the Parts for Flap Hinge Mechanism ............................................... 245
C.2.3 Calculation of Compliance in the System ....................................................... 249
C.2.4 Stiffness of the Load Path ............................................................................... 251
C.2.5 Impedance Matching ....................................................................................... 252
C.3 Bench Test for Active Flaps ............................................................................. 254
C.4 Conclusion ......................................................................................................... 258
APPENDIX D. ACTIVE BLADE DESIGN .............................................................. 259
D.1 Geometry of the Baseline Blade ....................................................................... 259
D.2 Blade-Layup Design Process ............................................................................ 262
D.3 Failure Analysis ................................................................................................. 270
D.4 Strength Test ..................................................................................................... 274
D.5 Blade Instrumentation ...................................................................................... 282
D.6 Passive Blade design ......................................................................................... 283
xxii
APPENDIX E. BLADE MANUFACTURING ......................................................... 285
E.1 Fabrication of Foam Core .................................................................................... 286
E.2 Instrumentation..................................................................................................... 288
E.3 Spar mandrel Design ............................................................................................ 292
E.4 Spar Manufacture ................................................................................................. 294
E.5 Fairing Manufacture ............................................................................................ 307
E.6 Fabrication of Active Flap.................................................................................... 316
E.7 Blade Section for Pull Test ................................................................................... 319
APPENDIX F. RESULTS FROM THE DUAL FLAP EXPERIMENTS.............. 322
F.1 Introduction ........................................................................................................... 322
F.2 Testing Process ...................................................................................................... 330
F.3 Sample Results ...................................................................................................... 334
F.4 Comparison with RCAS ....................................................................................... 341
F.5 Conclusion ............................................................................................................. 347
xxiii
ABSTRACT
The problem of vibration has limited the use of helicopters in both civil and military
applications. In this research, further analysis has been performed for the various on-
blade approaches available for vibration reduction using a unique optimization
framework.
For passive optimization, an aeroelastic environment with several well-established
analysis codes from different sources was developed that can be used to analyze and
design composite rotor blades for minimum vibration or maximum performance. This
design environment enables conceptual/early preliminary multidisciplinary rotor blade
design with realistic structural properties for modern composite rotor blades.
For the design of a rotor blade with active twist, a new design strategy was introduced
where the amplitude of dynamic twist is maximized. The optimization framework
included the aeroelastic design environment described earlier along with surrogate based
optimization technique. The surrogate based optimization is performed in combination
with Efficient Global Optimization algorithm. Results showed that the amplitude of
dynamic twist is a true indicator of control authority of active twist rotor for vibration
reduction. Furthermore, the optimization framework was extended to include discrete
design variables in the optimization and the solution for mixed-variable design problem
was obtained using three different techniques.
After modifying the aeroelastic analysis to account for the presence of active flaps, a
Mach-scaled composite rotor blade was designed using the same mixed design variable
xxiv
optimization framework to enhance the vibration reduction capabilities of the active flap.
In this case also, the amplitude of dynamic twist was used as the objective function and
the analysis was carried out at three different spanwise flap locations. This thesis also
includes work related to the design and fabrication of a composite rotor blade with dual
flaps which can be tested in a Mach-scaled spin test stand.
Finally, the use of camber actuation with quadratic and cubic camber deformation
shapes for vibration reduction and performance enhancement in dynamic stall region was
studied. The aeroelastic analysis was augmented with a modified version of the ONERA
dynamic stall model that accounts for morphing airfoil section.
xxv
Chapter 1. Introduction
A helicopter can take off and land, fly forward or backward, climb and descend and
move in almost any direction. These combinations of maneuvers, which are not possible
with a fixed-wing aircraft, have made the helicopter an ideal vehicle for a number of
challenging tasks in both civil and military operations. Modern civilian roles of the
helicopter include sea and land rescue mission, police surveillance, oil rig servicing,
homeland security etc. However, the issues of high vibration and noise have limited the
helicopter`s community acceptance and reduced its mission effectiveness.
The current chapter provides an introduction to the problem of vibration reduction in
helicopters. It highlights the main sources of vibration in helicopter and various
techniques currently being used to obtain vibration reduction. Different methods
available for vibration reduction are classified depending upon their nature
(active/passive) and implementation (components of the helicopter that it influences).
Based on the literature review, key areas where further improvement can be made in the
vibration reduction techniques are identified and that forms the motivation for the
research presented in this thesis. Finally, an outline of the thesis is presented.
1.1 Introduction to Vibration in Helicopters
Vibration in a helicopter leads to passenger/pilot discomfort and fatigue and it affects
the reliability and fatigue life of the airframe and its components. Other effects of
vibration are reduced weapon effectiveness, difficulty in reading instruments, etc. A
1
study done by Sikorsky Aircraft in 1973 [1] showed that reduction in vibration can
significantly improve the reliability and reduce the costs associated with maintenance and
life cycle. The current overall level of vibration in helicopters (approximately 0.05g to
0.1g) [2] remains significantly higher than those for a fixed-wing aircraft (0.01g). Even
though a significant reduction in helicopter vibration has been obtained over the last few
decades due to improved designs, the modern helicopters still do not meet the ultimate
goal of a jet smooth ride, as highlighted in Figure 1-1.
The main sources of vibration in a helicopter are the main rotor, tail rotor, engine,
gearbox and fuselage. However, the most dominant source of vibration is the main rotor
(more than 90% for a UH-60 Helicopter). The vibratory loads are produced by the main
rotor in the rotating frame. However, when the loads are transferred to the hub or the
fuselage in the fixed system, only the loads corresponding to the Nb/rev frequency are
observed, where Nb is the number of rotor blades. Thus, the rotor acts as a filter for the
vibratory loads. Figure 1-2 shows a typical vibration amplitude spectrum of a BO 105
helicopter in cruise condition.
Figure 1-1: Trends for Helicopter Vibration Levels [3]
2
Figure 1-2: Vibration Amplitude Spectrum for BO 105 in Level Cruise Flight [4]
The main rotor blade in a helicopter experiences highly unsteady aerodynamic loads
[5] as shown in the Figure 1-3, in addition to the time varying pitch angles. For a rotor in
forward flight, the advancing side of the blade experiences different aerodynamic
conditions as compared to the retreating side. On the advancing side, highly unsteady
aerodynamic loads are produced due to the blade-vortex interactions. This occurs when
the rotating blades encounter tip vortices shed by the preceding blades. The effects of
blade-vortex interaction are more pronounced at low forward flight speeds (μ = 0.15). At
higher advance ratios (μ = 0.35), very high mach numbers are observed at the blade tip on
the advancing side, and the flow reaches transonic conditions (supercritical flow). On the
retreating side, the dynamic stall condition is observed which is characterized by flow
separation. Also, near the root section on the retreating side, reverse flow occurs in the
region where the rotational speed at a radial location is smaller than the forward flight
speed. In this region, the flow over the airfoil section is from the trailing edge to the
leading edge. In addition to these, the finite length of the rotor blade results in tip
vortices, as in a fixed wing aircraft. Thus, the combination of unsteady aerodynamics and
large structural deformations in flap, torsion and lag due to blade flexibility and
slenderness leads to the generation of large oscillatory loads by the main rotor blades.
3
Figure 1-3: Unsteady Aerodynamics on the Main Rotor Disk
The variation in the vibratory loads at the cabin in a BO 105 helicopter with forward
flight speed is shown in Figure 1-4. As discussed above, large vibration is observed at
low flight speeds due to the blade-vortex interaction and then at high forward flight
speeds due to the dynamic stall and high speed flow effects. The vibrations due to the
blade-vortex interactions are increased by the maneuvers that retain the wake near the
plane of the disk, such as decelerating or descending flight (flare condition as shown in
Figure 1-4).
4
Figure 1-4: Vertical Cabin Vibration in BO 105 as a function of Airspeed [6]
The noise generated by helicopters has constrained the rotorcraft operation near cities
and have resulted in restrictions on the frequency of operation, time of day of specific
operation and types of rotorcraft that can be used. In a helicopter, main sources of noise
are the main rotor, the tail rotor and the engine. Among these, the most important one is
the main rotor. The low frequency noise that the main rotor generates is made up of basic
loading noise and broadband turbulence noise, each a function of lift and rotor speed. In
addition to these, BVI noise and High Speed Impulsive (HSI) noise become dominant in
descents and forward flight airspeeds, respectively. Further details related to the physical
mechanism of noise generation and acoustic modeling can be found in [7-9]. In the
research presented in this thesis, the main focus will be on vibration reduction.
1.2 Methods for Vibration Reduction
From the earliest days of rotorcraft development, the problem of airframe vibration
has been a serious concern. In a very early study [10], researchers identified three
different approaches to obtain vibration reduction, namely: a) by minimizing the source
of vibration, b) by reducing the response of the structure to the vibration, and c) by
5
isolating the structure from the source of vibration. Based on these approaches, the
problem of vibration reduction can be solved by both passive and active techniques.
Different methods for vibration reduction that have been discussed in the literature are
graphically summarized in Figure 1-5.
Vibration Reduction in Rotor
Passive Approach Active Approach
Dampers and Aeroelastic Vibration Control Vibration Control with Vibration Control with
Isolators Tailoring in Fuselage Blade Pitch Actuation on-Blade actuators
ACSR
HHC (Using IBC Using

Swashplate) 1. Pitch Links
2. Multi-Swashplates
Continuous/Embe-
Discrete Actuators
ded Actuation
Active Flaps Active twist

Leading Edge Slats Active Camber
Microflaps/Tabs Active Trailing Edge
Figure 1-5: Methods used for Vibration Reduction in Rotor Blades
1.2.1 Passive Approaches
The earliest approach for vibration reduction involved the usage of passive and semi-
passive devices like pendulum absorbers and isolators to reduce vibration. It is usually a
single DOF system with a small mass and a spring. Good reviews of passive vibration
reduction methods are given by Reichert [3] and Loewy [11]. Although the aerodynamic
vibratory loads persist, the transmission of these loads to the rest of the helicopter is
reduced. The passive devices have been used in many operational helicopters by tuning
their characteristics to filter out specific frequencies [12]. A brief description of the
6
passive systems like SARIB (Vibration reduction system using integral bar absorption)
and ARIS (Anti-Resonance Isolation Systems) implemented in the pylon assembly of
various Eurocopter helicopters to filter the dynamic loads transmitted to the airframe is
provided in [13]. However, these devices introduce weight penalties and are designed to
be effective over a narrow range of operating conditions only. Some of the studies have
considered the use of passive devices on the rotor blade themselves like the bifilar
pendulum implemented on the S-76 helicopters [14].
The second passive approach involves the tailoring of the structural and the
aerodynamic properties of a rotor blade using optimization techniques [15-17]. The
increased use of composite material in the blade allows easier fabrication of advanced
geometry blades and provides the potential of aeroelastic tuning. In this approach, the
vibration reduction problem is formulated as a mathematical optimization function with
appropriate objective function and constraints. For a composite rotor blade, the ply
angles, the ply thicknesses or any other cross-sectional parameters can be used as design
variables. For the aerodynamic shape, sweep, anhedral, droop, etc., at the blade tip are
considered as the design variables. In many cases, both structural and aerodynamic shape
optimization design variables are used simultaneously. Further discussion on
multidisciplinary optimization frameworks is provided in Section 1.3.2.
1.2.2 Active Approaches
An active control approach for the rotor has the potential to be a more effective
solution for vibration reduction since it can directly influence the source of vibratory
loads which are the main rotor blades [18]. This is fundamentally different from the
7
passive devices which attempt to reduce vibratory loads in the hub or the fuselage, which
are far away from the source. Also, the optimization technique has limited capability for
vibration reduction (20-40%) and in most of the cases, the final design is optimum at one
flight condition only. Most recent review of the active control methods for vibration
reduction is given by Kessler [19, 20]. A summary of all the methods shown in Figure
1-5 is given next.
1.2.3 Active Vibration Control in Fuselage
The ACSR (Active Control of Structural Response) scheme is based on the fact that in
a linear system it is possible to superimpose two independent responses such that the total
response is minimized [21]. In practice, the principle consists of connecting a number of
(hydraulic) actuators among strategic points on the fuselage and applying control forces
to the structure so to destruct the vibration signal [22, 23]. It has the advantage that it is
easy to maintain and the potential for vibration reduction is high since the actuator can
always produce the right amount of load at the right amplitude and phase to counteract
the primary vibration. In most of these cases, the vibration reduction is localized e.g.,
pilot seat, instrumentation panel, passenger cabin, etc. The ACSR technology has been
implemented for vibration reduction in the EH101 helicopter produced in Europe. An
AVCS (active anti-vibration control system) based on electromagnetic actuation was
implemented on the EC225 helicopter, while a piezo-based AVCS was successfully
tested on the EC135 [13]. The drawbacks of this system are that it requires a detailed
model for the rotor-fuselage dynamics in order to determine the optimum placement of
the actuator for maximum vibration reduction and it does not address the noise and
performance issues of the rotor.
8
In the early studies of vibration reduction, it was observed that the built-in twist of the
blade has a strong influence on blade vibratory loads. For vibration reduction, a
decreased negative twist is desired on the advancing side and a simultaneous increased
negative twist is required on the retreating side [24, 25]. This effect can either be
obtained by changing the pitch of the entire rotor blade, as described in Section 1.2.4, or
by using the twisting moment generated by the on-blade actuators to twist the rotor blade,
as described in Section 1.2.5.
1.2.4 Vibration Control with Blade Pitch Actuation
In the case of HHC (Higher Harmonic Control) and conventional IBC (Individual
Blade Control) methods, the aeroelastic behavior of a rotor blade is influenced by using
the actuators either mounted on the swashplate or by the use of pitch links to induce rigid
body actuation of the blade in pitch, respectively. The pitching motion consists of a high
frequency actuation signal on top of the primary collective and cyclic commands. This
method has potential for influencing the vibratory loads as they reduce the loads at the
source in rotating system.
In the HHC technique, the blade pitch actuation is introduced in the non-rotating
swashplate by superimposing the appropriate time dependent pitch commands [26]. The
HHC approach is the most mature active control approach for vibration reduction. Here,
all the blades experience the same input. The vibration levels in the fuselage or at the hub
are modified at their source before they propagate into the airframe. Numerical
simulations demonstrating the effectiveness of HHC technique for vibration reduction are
presented in [27, 28], while various experimental tests on model scale and full scale
rotors in wind tunnel are discussed in [29-32]. Some of the limitations of the HHC
9
approach are: 1) the considerable cost of implementing the HHC on a production
helicopter, 2) the power required for actuating the blades at the root and 3) limitations on
the objective that can be achieved with an HHC implemented through a conventional
swashplate as the actuation frequencies available are limited.
In the conventional IBC approach, each blade is actuated independently in the rotating
frame [33, 34] at any desired frequency, thus overcoming some of the limitations of the
HHC technique. The IBC approach involves independent feedback control of each blade
in the rotating frame. As compared to HHC, which provides a maximum of three DOF,
the IBC approach provides more freedom for vibration control. IBC can be obtained by
using active pitch link for each of the blades or by using multiple swashplates [20].
Experimental tests demonstrating the feasibility of the IBC for vibration reduction,
performance enhancement and noise reduction have been done in both the U.S. [35, 36]
and Europe [37]. Implementing the IBC approach brings significant challenges since
supplying hydraulic power to the rotating system is only possible by means of hydraulic
slip rings which are heavy and complex.
1.2.5 Vibration Control with on-Blade Actuators
In another approach, the actuation was moved onto the rotor blade. Unlike the HHC
and the conventional IBC, failure of the on-blade actuation system would not
catastrophically affect the flight safety. The actuation with the on-blade actuators requires
significantly less power as compared to HHC and conventional IBC [18] and it is
relatively less complex to implement them in the rotor blades. Vibration control with the
on-blade actuators can be considered as a subset of the IBC approach since the controller
10
has the freedom to control the actuator on each blade individually. (Hence, the IBC
approach described in the previous section is referred as “Conventional IBC.”)
The possibility of using an on-blade actuator to reduce vibration and noise was also
supported by the advent of smart materials. These are light weight, compact and have
small power requirements. A summary for the applications of smart material based
actuations for aeroelastic and vibration control is provided in the following references.
Giurgiutiu and co-workers [38-41] demonstrated the use of induced strain actuation
principles and capabilities for a smart rotor blade application like inducing twist, active
blade tip and active flaps. The application of these technologies for a fixed wing aircraft
was also discussed for active flutter control, buffet suppression, gust load alleviation and
sonic fatigue reduction. Straub [42, 43] stressed on the use of smart materials for “on the
blade” actuation to overcome the size, weight and complexity issues associated with the
hydraulic and electrical on-rotor actuation. Preliminary results showed the servoflap
control to be more effective as compared to the embedded actuators concepts like pitch
twist and camber control. Chopra [44] highlighted that the use of smart materials
methodology is equally applicable to other helicopter problems like aeromechanical
stability augmentation, handling qualities enhancement, stall alleviation, reduction of
interior/exterior acoustic signatures, minimization of the blade dynamic stresses and rotor
head health monitoring. Friedmann [45] obtained the scaling laws for the rotary-wing
aeroelastic and aeroservoelastic problem to be used for the scale model tests intended to
demonstrate the active control of vibration using an adaptive materials based actuation.
The on-blade actuation system mechanism can be further classified into discrete
actuation system and embedded actuation system.
11
1.2.5.1 Discrete Actuators
This approach includes an actuator installed inside the rotor blade and it is connected
to a movable device usually mounted on the trailing edge of the blade. The actuator
mainly consists of active materials capable of operating at large actuation frequencies and
an amplification mechanism to amplify the displacement produced by the active material.
Examples of discrete actuators are active flaps, active microflaps/tabs, leading edge slats,
etc. Discrete actuators modify the sectional aerodynamic properties of the region where
they are installed. The aerodynamic parameters that may be influenced by a discrete
actuator are cLα, cL,max, cMα, or cL/cD for the cross section. The control of the aerodynamic
loads acting on the blade is obtained either through changes in lift (lift effect) or by
elastically twisting the blade using the pitching moment generated by the movable
surface (servo-effect).
Actively controlled flaps (ACF) are usually installed between 0.6R to 0.9R along the
span of a rotor blade. The ACF can be implemented in a single, dual or multiple flap
configurations. ACF influences the blade vibrations by the combination of the servoeffect
and the direct lift effect [46]. Thus, like the HHC and conventional IBC, it reduces
vibration at the source which is the main rotor, but the power it consumes is an order of
magnitude less than IBC [18]. Substantial amount of work has been done to model the
effect of active flaps and use them for vibration reduction, performance enhancement and
noise reduction. Most of this work has been summarized in [47-49]. In these studies, it
was concluded that a flap deflection of ±4 deg at full RPM is sufficient to obtain
substantial benefit without incurring penalties on the stability or performance of the
helicopter.
12
As an active control device, microflaps have the potential for a high bandwidth control
with low actuation power requirement and minimal loss in the stiffness due to their small
size and low inertia. Microflaps are small, usually less than 5% of chord in height and are
mounted normal to the pressure surface and produce an increase in sectional c L,max by
approximately 25%. Since microflaps work within the boundary of the airfoil, they
produce a very small profile drag and are expected to have even smaller performance
penalty as compared to the active flaps. Microflaps were first proposed for fixed wing
aircraft [50, 51] to solve the problem of flutter. Many numerical studies with microflaps
for vibration reduction in helicopters [52-55] have been conducted in the last few years;
however, the experimental tests with microflaps mounted on a rotor blade in rotating
condition have yet to be performed. This is due to the difficulty in identifying an actuator
suitable for actuating [56] microflaps and the size constraints on the airfoil thickness.
Besides active flaps and microflaps, another on-blade discrete actuator called Active
Tab was developed by the research cooperation between JAXA and Kawada Industries
Ltd [57]. The low-speed wind tunnel test was conducted to prove the capability of BVI
noise reduction for a helicopter using the active tab control.
1.2.5.2 Continuous/Embedded Actuators
In this approach, the active material is usually embedded in the cross section or
bonded on the surface of the rotor blade. In most of the cases, the deformation is obtained
by the use of piezoceramic layers inducing shear strains. As compared to discrete
actuators, it does not have any external moving parts like hinges or bearing and, hence,
13
they have less profile drag. A possible disadvantage is that the maintenance of the
actuator is difficult since they are integrated inside the blade.
Active twist in the rotor blade is obtained by the active torsional moment generated by
embedded active piezoelectric fibers at ±45 deg orientation. The advantages of active
twist lies in the simplicity of the mechanism and no increase in the profile drag. Review
for the recent advances made in active twist technology is provided in [38, 58-62].
Numerical results have shown that a tip twist of ±2 deg obtained due to active twist
actuation is sufficient for vibration and noise reductions.
Active camber approach is relatively new as compared to the active twist and active
flap methods proposed in the earlier sections. Camber deformation is considered to be a
more efficient way of modifying blade sectional loads to influence vibratory loads at the
hub in fixed system. A good review of the active camber methodology is provided in [63]
and it discusses the use of camber deformation for vibration reduction. Active trailing
edge [64] is a new concept being developed as part of the Friendcopter program in
Europe and it consists of a trimorph bender integrated into the blade cross section. Like
active flaps, it moves the blade trailing edge upwards and downwards to generate servo
effects, but unlike the active flaps, there are neither moving parts nor discrete hinges.
1.3 Literature Review Relevant to the Thesis
This thesis focuses on the methods of vibration reduction where the vibration is
reduced at the source, that is, the main rotor. These include:
a) Active Twist Rotors
b) Rotor Blade with Active Flaps
14
c) Rotor Blade with Morphing Airfoil Section
d) Aeroelastically tailored Blades
1.3.1 Active Twist Rotors
Active twist is obtained in the rotor blades by including active MFC (macro fiber
composite) or AFC (active fiber composite) plies in the cross-sectional layup of the rotor
blade. Modeling and design of a rotor blade with active fiber plies present many
challenges since the actuator itself is part of the rotor blade and it is a load-bearing
member.
1.3.1.1 Modeling of Active Twist
The structural modeling of a cross section with embedded piezoelectric fibers must
take into account not only the contribution to mass and stiffness of the integral actuators,
but also the induced strain effects. Some of the works related to capturing these effects is
presented here.
Cesnik and Shin (2001) [65, 66] developed an asymptotic analysis that takes into
account the electromechanical three-dimensional nature of the problem and reduces it
into a linear analysis over the cross section and a nonlinear analysis of the resulting beam
reference line. The analysis results showed very good correlation with experimental data
obtained at MIT for active model blades. In [67], they analyzed the active cross section of
a rotor blade with multiple cell and showed that an increase in torsional stiffness does not
necessarily reduce the twist actuation. This approach was used in the design of the
NASA/ARMY/MIT Active Twist rotor [68]. The closed loop vibration control tests were
15
carried out with this rotor blade in NASA Langley’s Transonic Dynamic Tunnel in
forward flight conditions [69]. The experimental test results showed 40db reduction in
vertical shear vibratory loads and some reduction in other hub force and moment
components. This analytical work was further extended by Palacios and Cesnik to include
a modal solution procedure that allows arbitrary definition of the one dimensional elastic,
thermal and electric variables [70]. This methodology has been implemented in a
software code called UM/VABS [71] providing cross-sectional parameters for the active
beam model of the blades. For the nonlinear beam analysis of active rotor blades, an in-
house analysis code called UM/NLABS (University of Michigan, Nonlinear Active Beam
Solver) was developed [72]. It includes the mixed form of beam dynamic equations and
is expanded to account for the deformation of cross section through a set of finite section
modes. This resulting beam formulation explicitly captures both large elastic beam
deformation of the beam reference line and small local deformations at the cross section
for active/passive beams.
In the European Friendcopter Project, an FEA based procedure was used to determine
the stiffness and piezoelectric properties of an active twist rotor blade [73]. The cross-
sectional properties obtained were used in a multibody analysis of the active twist rotor to
obtain vibration suppression by open and closed loop controls [74]. Glukhikh et al. [75]
modeled the piezoelectric effect by the means of temperature analogy and thus converting
the electro-elastic problem to a thermo-elastic problem, which was solved using FEA in
ANSYS. Hoffman et al. [76] presented two simulation models for active twist by
prescribing active twist and twist moment based on modal shape function and validated
results from whirl tower test data. Brockmann and Lammering [77] derived a three
16
dimensional beam finite element model with regard to all the gyroscopic terms and the
actuation capabilities in a shear-flexible formulation with out-of-plane torsional warping.
Results obtained were compared with analytical solution for the static case and results
from the finite element shell model. The model showed good agreement with the finite
element shell model except in the cases where deformation in the cross-sectional plane
was observed since the beam formulation assumed undeformable shape.
1.3.1.2 Design and Parametric Studies
Cesnik et al. [78] performed numerical parametric studies with UM/VABS for wing
sections with double and triple cells to determine a cost effective way to add active
material to the cross section. Sekula and Wilbur conducted a series of parametric design
studies with structural and aerodynamic parameters to understand the twist actuation in
rotor blades [79-81]. In the parametric study with structural variables [80], the effect on
blade active twist, rotor power required, blade loads and vibratory hub loads were studied
due to the variation in blade torsional, flap-wise and lead-lag stiffnesses, sectional mass
and torsional inertia, and center of gravity and elastic axis locations. The analysis was
done using CAMARAD II [82] and the effect of actuators was represented by two torsion
moments producing equal but opposing loads at the blade ends. In a similar study, the
effect of aerodynamic parameters like linear blade twist, blade tip sweep, droop and taper
on active twist performance was studied [79, 81]. Based on the analysis, a candidate
design of AATR (Advance Active Twist Rotor) with −10 deg linear twist, 30 deg sweep,
10 deg droop and 2.5:1 taper ratio was proposed. In these studies, the external active
twisting moment applied was assumed to be independent of the variation in blade
17
structural and aerodynamic properties. Thornburgh et al. [83] performed parametric
studies on model-scale blades in order to determine the variables critical for active twist
response and to determine the effect of twist rate on cross-sectional constraints like mass
per unit length, chordwise location of shear center and CG and natural frequencies of
blade and material stresses. They also looked at the effect of scaling changes on optimal
structural design.
1.3.1.3 Active Twist Optimization
Active twist obtained from the active blade is dependent on the cross-sectional
properties of the rotor blade. Due to the large number of variables involved, the principle
of mathematical optimization provides a reliable way to explore the design space.
Cesnik and co-workers [84-86] developed an optimization framework to design an
active blade that maximizes the static twist actuation while satisfying constraints on
various blade requirements. The framework included UM/VABS for active cross section
analysis, DYMORE for one dimensional geometrically exact beam analysis, a MATLAB
based cross-sectional parametric mesh generator and MATLAB’s gradient based
optimizer. Results showed that the ATR (Active Twist Rotor) blade [68] could exhibit
30% higher twist actuation than the tested one. The same framework was used to design
the ATR-A blade for tests in NASA’s Transonic Dynamics Tunnel. The ATR-A was
based on a scaled model of AH-60D blade [84] that has a more complex geometry and
the final design obtained included manufacturing constraints. In the Friendcopter
program [87], the objective was to maximize the twist per unit span of a uniform beam
section under given constraints on airfoil shape, chordwise location of CG (center of
18
gravity) and SC (shear center), torsional frequency and beam stiffness. The design
variables used were chordwise location, length and thicknesses of piezoelectric layers,
ballast mass and four parameters that define the front C spar. Similar framework was
used with response surface technique for optimization in [88, 89]. Here approximations
of the original functions for constraints and objective function were obtained using a low
order polynomial.
1.3.1.4 Hybrid Active/Passive Optimization
In a hybrid active/passive optimization process, an optimal control law is combined
with the nonlinear optimization programming. The optimal controller tries to minimize an
objective function usually consisting of vibratory hub loads and flap control inputs while
the passive structural optimization aims to enhance the effectiveness of controller. Due to
the fact that the performance of active control strategies like active flaps or active twist
rotors is heavily dependent on the blade dynamics properties, it is useful to perform an
active/passive hybrid optimization to take advantage of the structural optimization to
enhance the effects of active control. Hybrid optimization technique has been used for
active flaps where a trailing edge flap controller design is combined with blade structural
optimization [90, 91]. Results obtained in [91] showed that an active-passive hybrid
method can outperform an optimal passive blade or an active flap retrofitted to a baseline
blade by achieving more vibration reduction with less control effort. This occurs due to
the tuning of blade flapwise bending frequencies and first torsional frequency close to
actuation frequency of the trailing edge flap. A multi-objective function optimization
approach was used in [90] to obtain simultaneous vibration and power reduction. The
19
results obtained highlighted strong tradeoff between performance enhancement and
vibration reduction, and both combined and sequential active/passive approaches led to
useful designs.
1.3.2 Passive Optimization
For the purpose of passive optimization, it is very important to perform the analysis of
rotor blades using a high fidelity aeroelastic framework. The current state of art with
respect to the analysis of rotor blades is discussed in [92] and [93] and the future needs
are described in [94]. The most recent reviews of different optimization methods used for
helicopter vibration reduction are provided in [15, 95]. Here, the literature review related
to recent work done in multidisciplinary optimization is presented.
In [96], Glaz used the efficient global optimization (EGO) algorithm with surrogate
models for rotor blade design optimization and a modified version of EGO based on
weighted expected improvement function (WEIF) for multi-objective function
optimization. The multi-objective optimization problems considered in the study were: 1)
vibration reduction through entire flight envelope, 2) noise and vibration reduction at low
advance ratio, and 3) vibration reduction and performance enhancement at high advance
ratios. The aeroelastic analysis in these studies was performed using high fidelity yet
computationally efficient aeroelastic code called AVINOR [97] developed at UCLA and
at University of Michigan. Jieun Ku [98] developed a rotor blade multilevel optimization
framework by including computationally efficient yet realistic and sufficiently accurate
tools like VABS and DYMORE. At the global level, optimization seeks structural
configuration that satisfies global constraints and focuses on rotorcraft dynamics. The
20
goal of the local level optimization is to find specific cross-sectional layout that satisfies
constraints obtained from the global level. Li et al. [99] developed a design tool which
incorporated manufacturing constraints, fatigue analysis and manufacturing uncertainty.
The design tool included VABS for cross section analysis and a parametric geometry
generator. A hybrid optimization procedure was developed that could handle a mixed
variable (discrete and continuous) problem. Khalid [100] incorporated the effects of
vehicle engineering, stability and control, aerodynamics, propulsion, transmission, weight
and balance, noise and cost calculations in the rotorcraft design environment using low
fidelity techniques. Collins [101] developed an automated high fidelity CFD based
simulation framework capable of predicting acoustic noise. A novel method using
combination of low fidelity and high fidelity results with statistical analysis was
developed and was used for optimization studies. Mustafa [95] used ModelCenter to
combine various software tools like: CATIA as the CAD tool, ANSYS as the FEA tool,
VABS for obtaining cross-sectional structural properties and DYMORE for frequency
and dynamic analysis of the rotor and MATLAB codes for generating input files and
reading output files.
1.3.3 Active Flaps
In the case of active flaps, the analysis performed in this thesis focuses on the design
of cross section for a composite rotor blade in order to maximize the control authority for
vibration reduction. In the next step, design and testing of a flap-actuation mechanism
using the X-frame actuator is presented which can be used for actuating flaps on a Mach-
scaled rotor blade.
21
1.3.3.1 Cross-sectional Design for Composite Blade with Active Flaps
There have been very few studies that have focused on the design of a composite rotor
blade with active flaps. For the design of a rotor blade with active flap, hybrid active-
passive techniques [90, 91] have been proposed. In [91], the optimal cross-sectional
stiffness values were determined using an integrated active-passive approach to reduce
vibration while minimizing control effort. In [90], simultaneous vibration reduction and
performance enhancement were obtained using active/passive optimization for a
simplified blade cross section. Ganguli and co-workers [102, 103] have performed
optimization using response surface and neural networks metamodels to determine the
optimal flap locations and the blade stiffness (torsional) for a rotor blade with multiple
trailing edge flaps to achieve minimum hub vibration level. In most of these studies,
either the cross-sectional stiffnesses are used as design variables or a simplified cross
section is used. Thus, there is a need for an optimization framework which can analyze a
“realistic” composite rotor blade with all the cross-sectional details such that the final
design obtained at the end of optimization is suitable for fabrication.
1.3.3.2 Dual/Multiple Trailing Edge Active Flaps
Patt et al. [104] used a single 12% long flap and dual 6% flaps for vibration reduction
and showed that the dual flap configuration is more effective. The increase in noise level
as an adverse effect of vibration reduction is smaller in the case of dual flaps. It was also
shown that dual flaps work better than a single flap for BVI noise reduction [105] and for
simultaneous vibration reduction and performance enhancement in dynamic stall
conditions [106]. Kim [107] used dual flaps to reduce both vibratory hub loads and
22
bending moments without significant change of control settings. Visvamurthy and
Ganguli [108] and Dalli [109] used multiple trailing edge flaps with differential
weighting to modify the contribution of second flapwise bending mode at a much lower
control power as compared to single and dual flaps. Some of the recent studies have
explored the experimental analysis of dual active flaps for BVI noise neduction [110] and
vibration reduction [111]. Thus, earlier work has demonstrated the advantages of multiple
flaps; however there have been no experimental studies to demonstrate their
effectiveness.
1.3.3.3 Piezoelectric Actuator for Active Flaps
Over the last two decades, a variety of actuators have been developed for rotor
blades with active flaps. A summary of these actuators is provided in [38] and [112].
More recently, trailing edge flap actuation system using Pneumatic Artificial Muscles
(PAM) [113] were tested which can produce required levels of blocked force and free
strain without the need for an amplification mechanism. Different types of actuators for
oscillating flaps that have been used in the past are given in Table 1-1, Table 1-2 and
Table 1-3.
Table 1-1: Active Flaps Implemented on Full Scale Blades

Num Baseline Program Actuator Details Blade Dim Flap Dimension
1 BK 117 ONERA/Eurocopter/ CEDRAT 5.5m radius 0.109 R and 0.156 c
Helicopter DLR [46] Actuator 0.325m chord centered at
0.75, 0.8, 0.85
2 Modified MD900 Boeing Double X-frame 5.15 m radius 0.18R, 0.35c,

[114, 115] Actuator 0.25m chord centered at 0.82R
3 Blade with AK120g JAXA [110] Two piezo stacks 5.8 m radius 0.1c, 0.1R
and AK100g airfoil with an amplifying 0.4 m chord centered at 0.75R
sections mechanism
4 Modified S-434 Sikorsky, UTRC, Electromechanical 4.45m radius 0.24c, 0.12R
rotor blades AATD [116] Actuator 0.2 m chord centered at 0.72R
23
Table 1-2: Active Controlled Flaps Tested on a Model Scale Rotor in Whirl Tower
or Wind Tunnel
Num Blade Program Actuator Details Blade Dimension Flap Dimension

1 2.1 m radius ONERA/Eurocopter/ CEDRAT Actuator 2.1 m R, 0.14m c 0.1R and 0.15c,
Adv Tech DLR/Eurocopter centered at
Rotor [117] Deutschland [118] 0.75R, 0.8R, 0.85R
2 1/6th scaled MIT [119] X-frame Actuator 1.54m span 0.12R and 0.2c
CH 47D , 0.137m chord centered at 0.78R
rotor blade
3 Blade with Univ of Maryland Piezo bender 0.914m, 0.0762 c 0.05R and 0.2c
NACA 0012 [120, 121] with mech level
airfoil
4 Blade with Univ of Maryland Cam Follower 1.848m, 0.1334m 0.25c , 0.18R
NACA 0015 Boeing [122, 123] Assembly centered at 0.88R
airfoil
5 Blade with AFDD [124] PZT bimorph bender 1.143m,0.0864m 0.1c, 0.12R
uniform beam with elevon centered at 0.75R
NACA 0012 lever arm mechanism
Table 1-3: Conceptual Designs for Active Flaps
Num Program Actuator Description Results

1 Penn State Induced shear 2.8 deg at 0 RPM
University [125] Piezoelectric actuator 1.4 deg at 400RPM
2 Univ of Michigan [126] Piezoceramic C Blocks 20 deg flap deflection in

wind tunnel tests
3 MIT [127] Piezoelectric bender flap deflection of 11.5

with flexure mechanism in no load at 100Hz
4 Boeing [128] Biaxial piezostack column Tested in 814g steady

and 29g vibratory
5 Univ of Maryland Double-lever (L-L) Amplification of 20

[129] Actuator upto 8/rev frequency
24
1.3.1 Optimization for Composite Structures
Different optimization techniques have been proposed for determining the minimum
number of layers in a composite laminate and the best fiber orientation and thickness for
each layer. In review papers [130, 131], the main optimization methods are described and
their characteristic features are contrasted for constant stiffness design and variable
stiffness design. In [132], composite laminate optimization with discrete variables is
discussed and issues associated with design of composite laminates are highlighted. In,
[133, 134] a novel laminate parameterization technique based on discrete material
optimization is used which is well suited for gradient based design optimization to handle
problems where ply angles and ply thicknesses are treated as discrete. Most of these
studies focus on the design of a simplified composite laminate.
For optimizing complex composite structures where time consuming finite element
analysis is required, surrogate modeling and response surface methods are proposed that
efficiently explore the design space and limit the number of FEA runs. Surrogate based
optimization technique have been used earlier for the design of composite rotor blade [97,
135]. However in these studies, only continuous design variables were considered. Guido
et al. [136] presented a mixed continuous-discrete variable optimization for design of
composite panel using surrogate modeling. Here, first a solution with continuous design
variables is determined and the solution with mixed design variable is obtained by
branching into sub-problems.
25
1.3.2 Camber Actuation
Airfoil camber deformation can potentially be achieved by embedded smart actuators
such as piezoelectric materials in the wing structure [137], or through compliant
substructures such as airfoils with deformable leading edge [138, 139]. Continuously
deformable airfoils have already been considered for performance and handling-quality
improvement of fixed-wing aircraft. Kota and co-workers [138, 140], demonstrated the
use of compliant mechanism for design of morphing aircraft structures. They suggested
the use of passive compliant structures with a generic force actuator to produce static
shape control of an airfoil camber. Gandhi and Anusonti-Inthra [141] looked at desirable
attributes of a flexible skin on a morphing wing. Parametric study was conducted to
determine the required in-plane and out-of-plane stiffness by considering the requirement
for actuation force, local and global deformation under aerodynamic loading and local
buckling of skin. Santer and Pellegrino [139] introduced network analysis technique to
determine an optimized compliant structure that deforms in conjunction with the wing
skin in response to a single displacement actuation. Rediniotis et al. [142] demonstrated
the use of shape memory alloys as artificial muscles to actuate a biomimetic hydrofoil.
Kudva and co-workers [137, 143], as a part of the Defense Advanced Research Projects
Agency (DARPA) sponsored Smart Wing program, developed deformable airfoil
surfaces using ultrasonic piezoelectric motors and eccentuator to ensure effective
transmission of motor torque to deflect control surface and demonstrated that the airfoil
could achieve trailing-edge deflections of up to 20 deg at deflection rate of over 80 deg/s.
The use of conformable airfoils in rotorcraft blades has been limited. Anusonti-Inthra
et al. [56] conducted research on conformable rotor airfoils using an optimized ground
26
structure of piezoelectric elements. The predicted trailing-edge deflections were 4 deg,
but the structure required a large number of piezoelectric elements. Later, Gandhi et al.
[144] proposed a conformable rotor airfoil design consisting of a passive compliant
structure coupled with a limited number of piezoelectric actuators to reduce the
complexity of the design. In this research, a detailed numerical analysis is performed with
different camber deformation shape function to explore the possibility of vibration
reduction and performance enhancement in the dynamic stall condition using camber
actuation.
In order to analyze morphing-type rotors, an analysis code called UM/NLABS-A
(University of Michigan Non-linear Active Beam Solver with Aerodynamics) was
developed by Thepvongs et al. [145] where the structural formulation captures plate-like
deformation in a geometrically-nonlinear beam-like framework. The aerodynamic model
was based on the 2D flexible airfoil theory and it includes 3D dynamic inflow model.
This code was also coupled with unstructured Reynolds Averaged Navier Stokes (RANS)
computational fluids dynamics (CFD) solver to obtain high fidelity aeroelastic solution
[146]. The comparison of aeroelastic loads predicted by the low order model and CFD
showed agreement in trimmed control setting, and some aspects of tip deflections and
fixed-frame hub loads.
1.4 Objectives of the Dissertation
In this study, further analysis has been done on the various on-blade approaches
available for vibration reduction with the aim of gaining further insight into the problem.
27
The specific areas of vibration reduction methodologies that this research focuses on are
highlighted in Figure 1-6.
Figure 1-6: Area of Focus for the Thesis
For the case of aeroelastic tailoring, there is a need to develop a high fidelity
multidisciplinary analysis framework which can model the structural properties of
realistic composite rotor blades. This framework can be used for design and analysis of
new rotor blade configuration with isotropic and orthotropic materials in their cross
section, which can be either active or passive. The design environment should combine a)
computational efficiency and speed of nonlinear 1D beam analysis with unsteady
aerodynamics, b) high-fidelity cross sectional analysis, and c) the ability to model
complicated topological details of a “realistic” composite rotor blade cross section.
With the use of composite material it is possible to design a rotor blade such that the
blade-twist due to active material in the blade is maximized. For the active twist rotor
blades, a tip twist of the order of ±2 deg at the actuation frequency is required to obtain
28
vibration reduction. In the preliminary analysis that was performed, it was observed that
the amplitude of dynamic twist is more directly related to the vibration reduction ability
of the active twist rotor blade, as compared to the static twist. Thus, the optimization
analysis should use dynamic twist as its objective function for optimum active rotor blade
design, which has not been done earlier. Also, the optimization framework must be
capable of working with both continuous and discrete design variables. This research
presents a new optimization strategy and framework for the design of a rotor blade with
active twist mechanism to enhance its capability for vibration reduction.
In most of the experimental and numerical studies that have been done till now to
analyze active flaps, the rotor blades are designed with low torsional stiffness such that
the effect of active flaps is enhanced. However, low torsional stiffness of the blade may
lead to detrimental effects like higher baseline vibrations (vibration in the absence of
active flap motion) and higher stresses in the blade root. In order to avoid these issues, it
is essential to design a blade with sufficient stiffness without compromising on the
effectiveness of active flaps for vibration reduction. This can be achieved by dynamically
tuning the blade frequencies near the expected actuation frequency. Thus, in the analysis
performed here, the amplitude of tip twist obtained due to the flap motion is used as the
objective function which is maximized.
The literature review in the previous section has highlighted that very few numerical
studies have been performed to explore the full potential of camber actuation for
vibration reduction and performance enhancement. Preliminary analysis done with lower
order model showed potential for vibration reduction and performance enhancement by
varying the amplitude and the phase of camber deformation along the blade span for a
29
scaled Bo105 rotor used in the HART II experiments. This research presents a detailed
study to explore the possibility of vibration reduction and performance enhancement
using camber actuation in forward flight conditions.
1.5 Outline of the Thesis
The main objective of this dissertation is to develop design strategies to enhance the
capabilities of active/passive rotor blades for vibration reduction. This objective is
achieved through the development of a high-fidelity aeromechanic analysis framework
and the use of appropriate optimization techniques. The achievements of this thesis and
the organization of chapters are given below:
1) Development of an aeromechanic analysis and design framework
Chapter 2 presents the development of an aeromechanic analysis environment for the
design of a composite rotor blade such that vibration reduction and performance
enhancement at the rotor hub is achieved using aeroelastic tailoring. This analysis
environment includes an advanced mesh generator for capturing the topological details of
a composite rotor blade cross section and for generating the 2D finite element mesh,
UM/VABS for the active cross-sectional analysis and comprehensive rotorcraft analysis
code for the aeroelastic analysis of rotor blade. The design environment was successfully
used to perform detailed parametric and optimization studies on the full scale model of a
UH-60 composite rotor blade [147] and on the passive version of ATR blade [148].
2) Optimization strategy for the design of active twist rotor
30
Chapter 3 introduces an optimization framework for the design and analysis of a
composite active twist rotor blade. The aeromechanic analysis environment described in
Chapter 2 was enhanced to include the effects of active plies in the cross section and it
was augmented with surrogate-based optimization technique to form an optimization
framework. In this study, the amplitude of dynamic twist has been proposed as the new
objective function for optimization studies for the design of active composite blade cross
section. It has been demonstrated using post-processing analysis that the dynamic twist is
a true indicator for vibration reduction capabilities of an active twist rotor. In this
framework, surrogate based optimization is included to explore the large design space
efficiently and to avoid the issues associated aeroelastic problems. (In these problems, the
runtime for each iteration is high (10-20 min) and some of the cases do not complete due
to failed convergence within the analysis). The optimum result obtained by maximizing
the dynamic twist amplitude is compared with optimum result obtained my maximizing
the static twist (the objective function used in all the studies discussed in literature
review) and advantages of the new strategy are highlighted [149]. Appendix A provides
mathematical expressions related to the development of surrogate models and Efficient
Global Optimization (EGO) algorithm used in the thesis.
3) Optimization at a range of actuation frequencies
Vibration reduction studies with the on-board active control devices have shown
that actuation frequency of (Nb-1)/rev, Nb/rev and (Nb+1)/rev are required for vibration
reduction in a rotor with Nb blades. Hence, the dynamic twist optimization performed in
Chapter 3 is carried out at a range of actuation frequencies and the active composite cross
31
section which is effective at a range of actuation frequencies simultaneously is
determined.
4) Active twist optimization with mixed design variables
In Chapter 4, the framework that was developed in Chapter 3 was extended to include
discrete design variables in the optimization study. The mixed-variable (with discrete and
continuous design variables) optimization described here is useful to obtain a realistic
optimum design and it highlights the effect of variable discretization on different
objective functions considered here. In the optimization studies performed in this chapter,
the ply thicknesses and ply angles are treated as discrete design variables. The modified
optimization framework includes both a genetic based optimizer and a gradient based
optimizer. The solution with mixed design variables is obtained using three different
techniques for comparison [150].
5) Design of a composite rotor blade with active flaps
The literature review highlighted that very few studies have been conducted to design
a composite rotor blade with active flaps that can be readily manufactured. Chapter 5
presents design studies for a composite rotor blade with active flaps such that the
authority of active flaps for vibration reduction is enhanced. In this study, a Mach-scaled
rotor blade which can be tested in the University of Michigan spin test stand is used as
the baseline rotor blade. In this study also, the mixed-variable surrogate-based
optimization framework described in Chapter 4 is used [151]. The aeromechanic analysis
was modified to account for the presence of active flaps on the rotor blade. Appendices C
32
to F present details about the design, fabrication and testing of a composite rotor blade
with dual flaps in a hover test stand.
6) Vibration reduction and performance enhancement in dynamics stall condition
using camber actuation
In Chapter 6, numerical studies are performed using UM/NLABS-A to study the
effect of camber actuation on vibratory loads at the hub and on rotor performance. The
rest of the surrogate based optimization framework is the same as in earlier chapters. In
the first study, the analysis is performed at an advance ratio of 0.24 where the quadratic
camber deformation shape function was used to obtain reduction in vibration and small
improvement in performance [63]. In the next step, a modified version of the ONERA
dynamic stall model was included in UM/NLABS-A for performing aeroelastic analysis
at high forward airspeeds. The vibration reduction and performance enhancement studies
in this section are carried out using both quadratic and cubic camber deformation shape
function [152] for μ = 0.33. Results obtained at the end show that the cubic camber
deformation shape function is more effective at reducing vibration and improving
performance of the rotor blade. Appendix B provides detailed description of the unified
aerodynamic model used in UM/NLABS-A.
Finally, Chapter 7 summarizes the work done in this thesis and presents
recommendations for the future work.
33
Chapter 2. A Multidisciplinary Design Environment for
Passive Composite Rotor Blades
The work presented in this chapter introduces a new design environment that
combines the computational efficiency and speed of 1D beam analysis with high-fidelity
accuracy approaching that of a 3D FE model for analyzing a composite rotor blade. The
environment contains a graphical modeling tool to rapidly define the cross-sectional
layup of a rotor blade or wing and a cross section mesh generator, both part of the
IXGEN pre-processing tool. It uses a cross-sectional beam analysis code (UM/VABS) to
determine the cross-sectional mass and stiffness properties, which it then feeds into a
comprehensive rotorcraft analysis code (RCAS). Using Phoenix Integration’s
ModelCenter as the optimization software, a full multidisciplinary design and
optimization environment for the preliminary design of composite rotor blades and wings
has been developed. As a test case, structural optimization case studies are presented
where the cross-sectional layup of the blade is determined that results in significant
vibration reduction at the rotor hub in forward flight condition for a model UH60 rotor
blade.
The implementation of the design environment was done in collaboration with
Advatech Pacific, Inc (San Bernardino, CA) as part of a SBIR project.
2.1 Introduction
34
The design of a composite rotor blade in a helicopter is inherently a multidisciplinary
problem involving aerodynamics, static and dynamic loads, aeroelasticity, materials,
fatigue life, manufacturing aspects, etc. Different stages of a product design and the
variation of estimated cost committed, design freedom and knowledge about the design is
shown in Figure 2-1 [100]. It also highlights the current design process and future trend
in product design. As indicate in the figure (by dark lines), during the preliminary design
stage, the knowledge about the design and the cost committed is the least while the
design freedom is the highest. As further progress is made in the design, there is an
increase in the cost and knowledge about the design and there is a decrease in the design
freedom. At the completion point, we have a complete product. The future trend
(indicated by gray lines) would be to shift the “knowledge about the design” curve
towards conceptual and preliminary design stage such that there is more design freedom
at less cost in the early stages. This strategy will help the designers to make better
exploration of the design variables in the early stages without significant cost. In order to
achieve this target, it is desired to obtain “mature” designs in the preliminary design stage
by making use of high-fidelity numerical tools for analysis. Thus, during the conceptual
design stage, there is a need to balance the fidelity of different models used in the
analysis with computational time requirements. Since making design changes in later
stages is far more expensive, it is essential to explore the complete design space in the
preliminary design stage itself.
35
Figure 2-1: Different Stages of Product Design [100]
For the detailed structural analysis of a rotor blade, a 3D Finite Element (FE) analysis
is required in order to capture all the topological details. However, the time-consuming
FE analysis is not suitable for conceptual design since a large design space needs to be
explored. Instead, due to the geometry of rotor blades, a “dimensional reduction” can be
performed that takes the original 3-D body and represents it as a 1D beam along a
predefined reference line. This can be done because one dimension (along the length) of a
rotor blade is much larger than the other two, and the structure is mostly uniform along
the span. For accurate dimensional reduction, following features are required:
36
a) Detailed modeling of the rotor blade cross section which includes isotropic and
orthotropic material properties and cross section topology.
b) Accurate representation of 3D blade properties along a reference line.
c) Non-linear aeroelastic analysis of the 1D blade.
The design environment presented here includes these features for rotor blade design
through the coupling of the following high-fidelity analysis tools in a design
environment:
a) IXGEN, the Intelligent Cross Section generator developed by Advatech Pacific,
Inc [153],
b) UM/VABS, the University of Michigan /Variational-Asymptotic Beam Sectional
analysis [70], and
c) RCAS, the Rotorcraft Comprehensive Analysis System [154].
2.2 Design Approach
The basic approach to the rotor blade design problem, which is adopted in this thesis,
is shown in Figure 2-2. It is based on the approach used for the design of a rotor blade
with active twist in [84, 86], however, that framework did not include the aeromechanic
analysis. In [84, 86], the failure analysis was done using the worst case loading obtained
a priori.
In this approach, important sections along the blade span are identified, where
geometrical or material properties do not change significantly. For each of these sections,
structural topology, layout of composite plies, materials and ply thicknesses are obtained
either through a user input or from the optimizer. This information is passed as input for
37
generating cross-sectional mesh and a UM/VABS input file using IXGEN. In the analysis
performed in this thesis, IXGEN is used as a mesh generator, though it has more
capabilities which are discussed in [153]. The cross-sectional analysis performed using
UM/VABS provides the mass and stiffness matrices which are used as input for 1D
nonlinear beam analysis. These matrices can also be used to determine the chordwise
location of shear center and center of gravity which act as constraints in the optimization
studies. The nonlinear beam finite element analysis of the composite beam is performed
in RCAS. It also includes aerodynamic models of different fidelities and various models
for determining inflow velocity and for capturing the dynamic stall effects. In a post-
processing step, the element loads resulting from the dynamic or aeromechanical analysis
are converted to equivalent stress and strain distributions in the individual cross sections.
Figure 2-2: 2D - 1D Design Approach.[148]
38
Figure 2-3 shows the Multidisciplinary Design Optimization (MDO) process that
implements the design approach described earlier for the design of a passive composite
rotor blade. The main pieces of software/codes used in this analysis are IXGEN, RCAS
and UM/VABS
interacts
User GUI
with
starts
Design Variables Integration Framework
- ply thicknesses Optimizer
- ply drops no
- spar location Termination
- ballast weight criteria met?
Cross
Section - etc.
yes
Mesh Sample Objectives
Generator and Constraints
End - Blade life
- Blade weight
2D Cross - Vibration levels
Stiffness Aeromechanic
Sectional - Noise levels
and Inertia Analysis
Analysis - CG location
Matrices (RCAS)
(UM/VABS) - EA location
- Aeroelastic stability
- Blade frequencies
3D - Local 3D stresses
Internal Loads
Influence and strains
Matrices
3D
stress/strain
(UM/VABS)
Figure 2-3: Multidisciplinary Optimization Process for Design of Passive Blade.
2.2.1 Blade Modeling Tool IXGEN
IXGEN [147, 148, 153] is a rotor blade and slender wing modeling environment that
lets the user quickly and easily define a rotor blade as a sequence of cross sections
stacked in the spanwise direction along a user-defined stacking axis. IXGEN has two
modes of operation – a GUI-driven mode for the designer to set up the blade, and a batch
mode for use in an automated design framework, where an optimizer or other type of
programmatic design driver modifies the defining parameters and regenerates the blade.
39
IXGEN contains a finite element mesh and UM/VABS model generator, and it has the
ability to execute UM/VABS directly from the UI. IXGEN has the capability to abstract
the definition of a rotor blade and its cross sections to a higher, feature based level. These
defining features, such as spar webs, spar caps, wrap layers, etc. are then parameterized,
and these parameters, in turn, can then be driven by an optimizer or a similar design
driver. IXGEN currently supports box, D and multi-cell spar concepts with spar webs
either perpendicular to the defining airfoil chord or at a slant angle off perpendicular.
Figure 2-4 shows several representative blade sections that have been modeled with
IXGEN. While the tool was developed for helicopter rotor blade design, it has also been
used to model typical wind turbine blade geometries.
Box Spar
Multi-Cell Spar
Slanted Spar Webs D Spar
Blade Root Section Typical Wind Turbine Cross Section
Figure 2-4: Representative Cross Sections Developed using IXGEN [148]
40
2.2.2 UM/VABS
UM/VABS [71] is a FORTRAN90 code developed at the University of Michigan
which solves the coupled equations of electro-thermo-elasticity in the cross section using
an asymptotic solution [155]. It includes cross-sectional analysis using different beam
theories: Euler-Bernoulli beam theory, Timoshenko beam theory, Vlasov beam theory,
and the original extended beam theory with finite section deformation modes. All these
models support the actuation effects in case active material is embedded in the cross-
sectional layup. As a post processing step, UM/VABS provides strain/stress influence
coefficients (SIC) which can be used to recover cross-sectional stress/strain and
displacements. The basic process of UM/VABS is shown in Figure 2-5.
In the UM/VABS analysis performed in this thesis, Timoshenko beam theory is used.
The analysis performed in Chapter 3 includes the effect of active plies used in the cross
section. As a result, the output produced by UM/VABS includes actuation forces and
strains in addition to the traditional mass and stiffness matrices. In Chapter 6, finite
section modes are used to model the effects of camber deformation and to determine the
corresponding higher-order mass and stiffness matrices.
41
Figure 2-5: Basic Process of UM/VABS [71]
2.2.3 RCAS
The Rotorcraft Comprehensive Analysis System (RCAS) is a software code developed
by U.S. Army Aeroflightdynamics Directorate (AFDD) and Advanced Rotorcraft
Technology (ART) to provide state-of-the-art rotorcraft modeling and analysis
technology for Government, Industry and Academia [156]. The current capabilities of
RCAS (RCAS v12.08) include large rigid-body motion, a nonlinear beam element valid
for large blade deformation, easier procedures for building complex finite-element
rotorcraft models, and various options for modeling unsteady aerodynamics, rotor inflow
and dynamic stall effects. RCAS is capable of modeling a complete range of complex
aircraft configurations operating in hover, forward flight, and maneuvering flight
conditions. RCAS is designed to perform a wide variety of rotorcraft engineering
analyses like vehicle performance, aerodynamics, aeroelastic stability, flight dynamics,
etc. It uses hierarchical finite-element modeling for the structure and airloads in order to
42
model complex rotorcraft configurations. In [154], a number of examples are presented to
demonstrate the unique and advanced modeling and analysis capabilities from RCAS.
2.2.4 ModelCenter Integration
ModelCenter is used for integrating all the numerical tools described above and for
performing parametric and optimization studies in this chapter. Any response parameter
produced by either UM/VABS or RCAS which is exposed by the MDO environment is
available to the optimizer as either a constraint or objective value. Figure 2-6 shows the
ModelCenter components implementing the rotor blade design. Detailed information
about each of the modules shown in Figure 2-6 is provided in [148].
Figure 2-6: ModelCenter-Based Blade Optimization Process.
43
This implementation of the rotor blade design process in ModelCenter facilitates a
wide variety of trade study and optimization scenarios. Design drivers can be wrapped
around either the entire process, or individual components, such as IXGEN-UM/VABS-
VABS Post-Processing, to optimize the structural properties of an individual cross
section without running the aeromechanical analysis, as demonstrated in [153].
ModelCenter’s parameter linking functionality can be used to link IXGEN parameters to
reduce the number of design variables and enforce continuity or manufacturing
constraints.
The development of IXGEN software and the integration of various analysis codes
into the ModelCenter were carried out by Advatech Pacific. Detailed analysis for the
verification and validation of the design environment and all the design studies presented
in this chapter were performed at the University of Michigan.
2.3 Application Example: Vibratory Hub Load Minimization
The MDO environment described in this chapter can be used to solve a variety of
optimization and design problems involving metallic and composite rotor blades. Any
parameter exposed in ModelCenter can be used as a design variable, part of a response,
constraint, or objective function. In order to demonstrate the capability of the design
environment described in the previous sections, a full-scale UH-60 rotor model given in
the RCAS examples is used as the baseline case for parametric and optimization studies.
44
2.3.1 Description of the Baseline Rotor Blade
The characteristics of the UH-60 rotor are listed in Table 2-1, and the top view of the
rotor blade is shown in Figure 2-7. For the purpose of analysis, the blade was subdivided
into three spanwise regions, as shown in Figure 2-7. These regions match with the airfoil
breaks existing in the actual UH-60 blade. The rotor blade consists of a SC1095 airfoil in
Section 1 and Section 3, while Section 2 has a SC1094R8 airfoil.
Table 2-1: Characteristics of UH-60 Rotor
Rotor Type Fully Articulated

Number of blades 4
Blade radius (R ) 26.83ft (8.18m)
Blade Chord (c ) 1.73 ft* (0.527m)
Solidity 0.0826
Airfoil Section SC1095/SC1094R8
Blade Pretwist -12 deg
Hinge Offset 1.25ft (0.381m)
Rotor Speed 258 RPM
CT 0.008
CQ 0.000354
Air density 0.00237 slugs/ft3 (1.225 kg/m3)
Advance ratio (µ) 0.24
Blade tip sweep 20 deg
* Average chord
Region 1 Region 2 Region 3
Figure 2-7: Top View of the UH-60 Rotor Blade [157]
45
The UH-60 rotor blade cross section as modeled in IXGEN is shown in Figure 2-8. It
consists of a boxed spar with overwrap plies wrapping around the whole airfoil. It also
includes an erosion strip, leading edge wrap, trailing edge tab and trailing edge fill that
are commonly observed in a typical rotor blade cross section. Finer details about the
cross section and material used in different regions of the airfoil are shown in Figure 2-9.
Among the plies used, E-Glass is bidirectional while S-Glass and IM7 plies are
unidirectional. The specific material properties for different materials used in the cross
section can be found in Table 2-2.
Figure 2-8: Cross Section of the UH-60 Rotor Blade
(A) Front Spar
(B) Middle Main Spar Region
46
(C) Trailing Edge Region
Figure 2-9: Cross Section Layup
Table 2-2: Material Properties
E-Glass IM7 Steel S-Glass Plascore

3
ρ (slugs/ft ) 3.34 3.01 15.13 3.61 0.09
E11 (ksi) 3002 23933 29736 6295 1.00
E22 (ksi) 3002 1276 29736 1740 20.02
E33(ksi) 3002 1276 29736 1740 1.00
G12 (ksi) 594 710 11169 522 3.48
G13 (ksi) 594 710 11169 522 1.00
G23 (ksi) 594 710 11169 522 5.80
ν12 0.15 0.34 0.3 0.28 0.01
ν13 0.15 0.34 0.3 0.28 0.3
ν23 0.3 0.3 0.3 0.3 0.01
The thickness of the plies used in the different regions is given in Table 2-3. These
thicknesses are the same as those obtained in [153] based on matching the cross-sectional
properties of an existing rotor blade. In [153], different sets of ply thicknesses were
obtained for Region 1, Region 2 and Region 3. However, for the optimization and
parametric studies shown in this chapter, it has been assumed that the cross-sectional
layup is the same in all the three regions to reduce the number of design variables,
thereby reducing the runtime required for optimization studies. The design environment
itself is capable of handling any number of design variables.
47
Table 2-3: Thickness of Plies used in the Layup
Ply Thickness Thickness (mil)

Spar Cap Ply 1 4.10
Spar Cap Ply 2 3.72
Spar Cap Ply 3 3.72
Spar Cap Ply 4 1.00
Overwrap Ply 1 0.50
Overwrap Ply 2 0.50
Overwrap Ply 3 0.50
Overwrap Ply 4 0.50
Erosion Strip 1.64
LE Fill 7.71
The structural frequencies of the blade at 100% RPM are listed in Table 2-4 and were
obtained using RCAS. The frequencies are slightly different (within ±6%) from those
obtained in [153] because of the assumption that cross-sectional layup is the same in all
the three regions of the blade. For the rotor aeroelastic analysis, the trim option (wind
tunnel trim) is used in the RCAS solution. The trim targets used in the analysis are: C T =
0.008, no longitudinal and lateral flapping angle for tip path plane (β1s = 0 and β1c = 0 for
the tip path plane), and the blade pitch settings are used as the trim variables. The mean
value of the hub loads and the amplitude of the 4/rev component for the baseline blade
are given in Table 2-5, where Fx, Fy, and Fz represent components of the hub force in the
non-rotating frame, while Mx, My, and Mz represent components of the moments at the
hub. In all the results shown in this chapter, an advance ratio of 0.24 is used in RCAS.
48
Table 2-4: Structural Frequencies of the Blade at 100% RPM
Mode Shape Frequencies

(/rev) (rad/s)
1st chordwise bending 0.27 7.26
1st flapwise bending 1.04 28
2nd flapwise bending 2.68 72.4
1st torsion 4.57 123.5
2nd chordwise bending 4.98 134.4
3rd flapwise bending 5.48 148
Table 2-5: Mean Value and Amplitude of 4/rev Vibratory Loads at the Rotor Hub
for µ = 0.24
Mean Values 4/rev Amplitude

Fx (lbf) 107.8 23.03
Fy (lbf) 559.9 18.74
Fz (lbf) 22586 212.1
Mx (ft-lbf) 3397 6696
My (ft-lbf) 1368 5458
Mz (ft-lbf) 26745 377.5
2.3.2 Definition of the Rotor Blade Optimization Problem
A different set of objective functions can be defined depending upon the problem
being solved. In principle, any output provided by RCAS or by UM/VABS can be
selected as the objective function. For this chapter, the following objective functions are
considered:
a) Minimization of 4/rev vertical vibratory load at the hub (min FZ4)
b) Minimization of combined vibratory load (from all the hub load components)
Similarly, any combination of outputs from RCAS and UM/VABS can be used to
form constraints. For the optimization studies presented herein, the following parameters
are constrained:
49
a) Chordwise location of blade cross-sectional center of gravity
b) Chordwise location of cross-sectional shear center
c) Blade fundamental rotating frequencies
d) Maximum allowable blade strain in the cross section
e) Mass per unit length for each section of the blade
The optimization problem can also include constraints on aeroelastic stability and
autorotation, however, these were not considered directly in the problems studied in this
thesis. The optimization problem can be solved using gradient based optimizer or non-
gradient based methods such as genetic algorithm and surrogate optimization. Finally, the
selected design variables for this study are:
 The thickness and lamination angle of spar cap plies in the cross section layup.
However, the material properties used in each ply are kept constant.
 The chordwise location of the vertical auxiliary spar web.
 Discrete ballast mass and its chordwise location in each of the sections
From the above discussion it can be seen that a large number of variables could be
used as design variables. However, since each run for a complete rotor analysis takes
between 15 to 30 minutes on a Windows machine (Intel Core2 QUAD CPU @ 2.39GHz
and 1.96GB of RAM), and the number of runs required for optimization increases
exponentially with the number of design variables, it is desirable to reduce the number of
design variables to the most influential ones. The variables which are most critical for the
design can be identified through parametric studies. Thus, as a first step, a parametric
study is performed with respect to different design variables.
50
2.3.3 Preliminary Parametric Study
In the first parametric study, the thickness of the plies used in the spar cap and the
fairing were considered independent variables. The thickness of each ply was varied
between ±25% of the baseline value, and the variation in blade properties was observed.
A total of 300 test cases were set up using the Latin Hypercube Method which spans the
whole design space. This example also helped to show the robustness of IXGEN in
generating the cross-sectional mesh for different kinds of layups. Table 2-6 shows the
variation observed for some of the critical responses during the parametric study with ply
thicknesses, where S11 is the axial stiffness, S44 is the torsional stiffness, S55 is the
flapwise bending stiffness, SC is the shear center, FZ4 is the amplitude of the 4/rev
vertical force, and MX4 is the amplitude of 4/rev rolling moment at the hub measured in
fixed system. Results obtained here show that the cross-sectional stiffness and 4/rev
vibratory moments at the hub are very sensitive to variation in ply thickness.
Table 2-6: Variation in Blade Parameters Observed During the Parametric Study
with Ply Thickness
Response Variation
S11 -11.30% to 10.72%
S44 -12.34% to 11.63
S55 -14.76% to 13.88%
SC -2.73% to 3.26%
1st Tor Freq -2.53% to 2.53%
FZ4 -11.78% to 9.77%
MX4 -61.17% to 83.97%
Further information about the influence of each design variable can be obtained by
observing the contribution of each design variable to the overall variation in the observed
response, as shown in Figure 2-10. In these plots, the Y-axis represents the percentage
51
contribution of each design variable while the X-axis lists each of the design variables
considered. (OW1 refers to thickness of Overwrap Ply 1, SC1 refers to thickness of Spar
Cap Ply 1, and so on.) The axial stiffness (S11) of the cross section is mainly influenced
by the thickness of Ply 1 in the spar cap since it is the unidirectional IM7 ply at 0 deg
angle. S55 follows the same trend as S11, and the effect of the thickness of the SC1 ply is
more apparent here. However, S44 is more influenced by the thickness of the SC2 and
SC3 plies since they are oriented at +450 and -450 angles, respectively. All plies in the
spar cap region and all plies in the overwrap region contribute equally to the variation in
mass per unit length, as expected. However, the contribution of the overwrap plies is
larger than the contribution of the spar cap plies since overwrap plies occupy a larger
fraction of the airfoil contour. The contribution of different ply thickness to the variation
in the first torsional frequency and FZ4 is less intuitive to predict, and in these cases, the
parametric studies are very useful in the overall understanding of the problem.
Similar parametric studies can also be done with respect to other design variables like
ballast mass and their chordwise locations, ply angle, and chordwise location of the main
and auxiliary spar web.
2.4 Optimization Studies for Vibration Reduction
In this section, results are presented for two optimization studies which were
performed to obtain a design with low vibration level using the design environment
described in the earlier section. The vertical component of 4/rev vibratory load (FZ4) is
used as the objective function for the first optimization study while in the second
52
optimization study, the combined vibratory load, which includes contribution from all the
three force and moment components, is reduced.
S11 M11
60 20
Contribution (%)
Contribution (%)
15
40
10
20
5
0 0
OW1OW2OW3OW4 SC1 SC2 SC3 SC4 OW1OW2OW3OW4 SC1 SC2 SC3 SC4
S55 S44
100 40
Contribution (%)
Contribution (%)
50 20
0 0
-50 -20
Tor Freq FZ4

20 30
Contribution (%)
Contribution (%)
20
0
10
-20
0
-40 -10
Figure 2-10: Contribution of each Ply Thickness to Overall Variation in the

Response
2.4.1 Minimizing Vertical Component of the 4/rev Vibratory Hub Load (min FZ4)
In the first optimization study, the amplitude of the 4/rev vertical hub force (FZ4) is
used as the objective function to be minimized. The design variables used for this study
are ply angles and ply thicknesses for the spar cap plies, chordwise location of the
auxiliary spar web and mass, and location of the ballast mass used in Section 1 and
Section 2. (Since Section 1 and Section 3 have same airfoil section and layup, the ballast
mass used in Section 1 and Section 3 were assumed to be the same.) Upper and lower
53
limits for these design variables are listed in Table 2-7. The lower limit used for the ply
thickness corresponds to 1/4th of the baseline ply thickness while the upper limit for the
ply thickness corresponds to two times the baseline value. The upper and the lower limit
used for the ply angles depend upon the nature of the material, whether it is unidirectional
or bidirectional. The ballast mass is allowed to vary between 0.05 slugs/ft and 0.15
slugs/ft while their chordwise location is allowed to vary between the leading edge and
the quarter chord of the airfoil section. Constraints used during the optimization are listed
in Table 2-8. The shear center (SC) and center of gravity (CG) of the cross section are
constrained to lie near the quarter chord of the airfoil to indirectly enforce stability
criteria. Mass per unit length of the cross section is allowed to vary between ±15% of the
baseline value. Maximum allowable axial strain along the material direction is limited to
6000 microstrain. The 1st torsion frequency is constrained between 3/rev and 6/rev.
Table 2-7: Design Variables used in the Optimization Study for Min FZ4
Design variables Baseline Min Max

Thickness SC Ply 1 (mil) 4.10 0.03 8.20
Angle SC Ply 1 (deg) 0.0 -90.0 90.0
Angle SC Ply 2 (deg) 45.0 -90.0 90.0
Angle SC Ply 3 (deg) -45.0 -90.0 90.0
Angle SC Ply 4 (deg) 90.0 -90.0 90.0
Aux Web Loc (%c) 6.60 2.5 15.0
Ballast mass 1 (slugs/ft) 0.10 0.05 0.15
Ballast mass 1 Loc (%c) 15.0 0.0 25.0
Ballast mass 2 (slugs/ft) 0.11 0.05 0.15
Ballast mass 2 Loc (%c) 16.0 0.0 25.0
54
Table 2-8: Constraints used in the Optimization Problem for Min FZ4
Constraints Baseline Min Max

CG Sec 1 (%c) 25.0 20.0 30.0
CG Sec 2 (%c) 25.0 20.0 30.0
M11 Sec 1 (slugs/ft) 0.212 0.180 0.244
M11 Sec 2 (slugs/ft) 0.223 0.256 0.189
SC Sec 1 (%c) 36.5 25.0 37.0
SC Sec 2 (%c) 35.9 25.0 37.0
Tor Freq (/rev) 4.53 3.0 6.0
Max Strain ε11 (µε) 2175.7 0.0 6000.0
For this optimization, the complete rotor blade analysis process is executed at each
run, which involves running IXGEN, UM/VABS and RCAS. During the parametric
studies, it was observed that some of the cases in RCAS did not reach a converged
solution for trim analysis, and hence, there were a few failed cases involved. For the
gradient-based optimization, it is required that none of the cases fail during the run. As a
result, the gradient-based algorithm was not used for this optimization. Among the
various options available in the ModelCenter Release 10 optimization tool package,
“Non-dominated Sorting Genetic Algorithm II” (NSGA) [158] was used since it allows
for failed runs in the optimization process. Parameters used for NSGA optimization are
listed in Table 2-9. The optimization process ran for approximately 48 hours and stopped
after exceeding the limit on maximum number of generations allowed. During this time, a
total of 236 complete iterations were performed. Although not the global optimum, the
result obtained at the end shows 52% reduction in FZ4 while satisfying all the constraints.
Details for the optimized case are listed in Table 2-10. The variation of the objective
function, design variables and constraints with generation are shown in Figure 2-11 to
Figure 2-14.
55
Table 2-9: Parameters used for NSGA II Algorithm
Population 24
Optimization Parameters for Binary Variables
Binary Crossover Probability 0.7
Binary Mutation Probability 0.5
Optimization Parameters for Real Variables
Crossover Probability 0.7
ηC (Index for crosssover) 15
ηM (Index for mutation) 20
Mutation Probability 0.167
Stopping Criteria
Convergence Generations 5
Convergence Threshold 0.001
Max Evaluations 1000
Max Generations 12
220
Objective Function FZ4 (lbf)
200
180
160
140
120
100
1 2 3 4 5 6 7
Generation Number
Figure 2-11: Variation of Objective Function (FZ4) with Generation Number
Among the ply thicknesses, Ply 1 which is oriented at 0 deg shows the maximum
increase of 34% in ply thickness. As compared to the baseline case, the most significant
variation is shown by all the ply angle variables. In the optimized case, all the plies are
oriented at approximately 76 deg. This is a direct result of the objective function being
purely the minimization of FZ4 (more in the next subsection). As a result of this, there is a
56
48% reduction in the axial stiffness (S11) and more than 60% reduction in torsional (S44)
and flapwise bending (S55) stiffness of the blade cross section, as shown in Table 2-11.
(The values listed in Table 2-11 are nondimensionalized with respect to the baseline case,
that is, ). The reduction in chordwise stiffness is relatively
small (11%). As shown in Figure 2-13, the variation in ballast masses and their chordwise
location is small, too. Also, the auxiliary spar web has moved back by 0.04c, resulting in
a further decrease in cross-sectional stiffness.
Table 2-10: Design Variables and Constraints for the Optimized Case
Design variables Baseline Min FZ4

Thickness SC Ply 1 (mil) 4.10 5.49
Angle SC Ply 1 (deg) 0.00 77.37
Angle SC Ply 3 (deg) -45.00 73.24
Aux Web Loc (%c) 6.60 10.27
Ballast mass 1 (lb/ft) 0.10 0.114
Ballast mass 1 Loc (%c) 15.00 18.50
Ballast mass 2 (lb/ft) 0.11 0.079
Ballast mass 2 Loc (%c) 16.00 0.226
Constraints Baseline Min FZ4

CG Sec 1 (%c) 24.14 25.52
CG Sec 2 (%c) 24.1 28.38
M11 Sec 1 (lb/ft) 0.207 0.2323
M11 Sec 2 (lb/ft) 0.223 0.197
SC Sec 1 (%c) 36.5 34.62
SC Sec 2 (%c) 35.9 34.07
Tor Freq (/rev) 4.53 3.71
Max Strain (µε) 2175.7 4901.0
57
The overall effect of these variations can be seen in the dynamic frequencies for the
optimized case, which are shown in Table 2-12, where the first torsion frequency, second
chordwise bending frequency, and third flapwise bending frequency have moved further
away from each other. Thus, the optimization study indicates that reducing the coupling
between these modes has resulted in lower vibration amplitude for FZ4.
Table 2-11: Ratio of Cross-Sectional Properties for the Optimized Case
Ratio Min FZ4 Ratio Sec 1 Sec 2

0.523 1.096 0.882
0.377 0.947 0.913
0.379 1.056 1.059
0.889 0.945 0.91
Table 2-12: Blade Structural Frequencies for the Optimized Case
Mode Shape Frequencies (/rev)

Baseline Optimized
1st chordwise bending 0.269 0.269
1st flapwise bending 1.0356 1.0356
2nd flapwise bending 2.679 2.59
1st torsion 4.57 3.72
2nd chordwise bending 4.979 5.019
3rd flapwise bending 5.48 4.44
Ply Thickness (mil)
10
Ply1
Ply2
5
Ply3
Ply4
0
1 2 3 4 5 6 7
Generation Number
Ply Angle (deg)
100
-100
1 2 3 4 5 6 7
Generation Number
Figure 2-12: Variation of Ply Thickness and Ply Angle with Generation Number
58
0.4
Bal Loc(c)
0.2
0
1 2 3 4 5 6 7
Section 1
0.2
Bal(slug/ft)
Section 2
0.1
0
1 2 3 4 5 6 7
0.2
Aux Loc(c) 0.1
0
1 2 3 4 5 6 7
Generation Number
Figure 2-13: Variation of Ballast Mass and Auxiliary Spar Web with Generation
Number
All the constraints used in the optimization problem are shown in Figure 2-14. The
constraints are non-dimensionalized using their maximum and minimum values such that
“0” represents the minimum value of the constraint while “1” represents the maximum
value of the constraints. The limits for the constraints are represented by solid red lines in
the plot. The results obtained indicate that only the constraint on mass per unit length for
Section 2 (M11-Sec 2) is closer to its lower limit while the rest of the constraints are well
within the boundaries. Besides this, an increase is observed in the maximum strain for the
cross section. This also indicates that the optimized solution obtained here may not be the
optimum solution, and there is a possibility of finding a better solution by increasing the
maximum number of generations allowed in the stopping criteria listed in Table 2-9.
Since, the aim of this study was to demonstrate the robustness of the aeroelastic design
environment for passive optimization studies, further iterations were not carried out.
59
1.2
CG-Sec1
CG-Sec2
1
Non-dimensional Constraints
SC-Sec1
SC-Sec2
0.8 Max 11
M11-Sec1
0.6
M11-Sec2
0.4 Tor Freq
0.2
1 2 3 4 5 6 7
Generation Number
Figure 2-14: Variation of Constraints with Generation Number
In the first optimization study, only the vertical component of the 4/rev vibratory force
at the rotor hub was minimized. As shown in Figure 2-15, vibration reduction in FZ4 is
accompanied by an increase in amplitude for FX4, FY4, MX4 and MZ4. In order to reduce
the vibratory loads for all the hub load components simultaneously, a different objective
function was selected in the second optimization study.
300
250
Change in Vibration (%)
200
150
100
50
0
Fx4 FY4 FZ4 MX4 MY4 MZ4
-50
-100
Load Component
Figure 2-15: Percentage Change in Amplitude of 4/rev Vibratory Hub Load

Components for the Optimized Case (min FZ4) with respect to the Baseline Case
60
2.4.2 Minimizing Combined Vibratory Hub Load (min FH4)
In this case, the contributions from all the load components at the hub are included in
the objective function (FH4). The new objective function is defined as:
where, R is the radius of the rotor blade.
The optimization problem is solved using the design variables, constraints, and
optimization parameters described in Table 2-7, Table 2-8 and Table 2-9, respectively. In
this case, as before, the optimization process stopped after exceeding the limit on the
maximum number of generations allowed. The final results show a 27% reduction in FH4,
as indicated in Figure 2-16. Table 2-13 shows the final optimized design for this case
along with the results obtained from the “min FZ4” case.
Among all the design variables, the most significant variation occurs in the thickness
of Ply 2, which has almost doubled. The effect of this can be seen in the cross-sectional
properties listed in Table 2-14 where the torsional stiffness of the cross section has
increased by almost 24%. In spite of the increase in ply thickness, the bending and axial
stiffness of the blade section has decreased due to a 15 deg change in ply angle for Ply 1.
Also, there is an approximately 15% increase in the flapwise-bending inertia.
61
540
520
Objective Function FH4 (lbf)

500
480
460
440
420
400
380
0 2 4 6 8 10
Generation Number
Figure 2-16: Variation of FH4 with Generation Number
Unlike the variation observed in the “min FZ4” case for the blade structural
frequencies, the dynamic frequencies for “min FH4” are very close to the baseline case.
As shown in Table 2-14 and Table 2-15 the increase in cross-sectional torsional stiffness
is accompanied by an increase in torsional inertia, due to which the resultant increase in
torsional frequency for the blade is very small. Thus, from the second optimization study,
it appears that a decrease in combined vibratory load at the rotor hub can be obtained by
increasing the torsional stiffness of the blade section without making any significant
changes in blade dynamic properties.
The percentage reduction in 4/rev vibratory loads at the rotor hub for the “min FZ4”
case and “min FH4” case are shown in Figure 2-17. For the “min FH4” case, lower
vibrations are observed for four of the six hub load components. The baseline vibration
shown in Table 2-4 clearly indicates that FZ4, MX4 and MY4 have the largest contribution
to the overall vibratory loads at the hub, and thus in order to reduce combined vibratory
loads, it is important to reduce vibration in these components.
62
Table 2-13: Design Variables and Constraints for Min FH4
Design variables Baseline Min FH4 Min FZ4

Angle SC Ply 1 (deg) 0.00 15.86 77.37
Angle SC Ply 2 (deg) 45.00 50.57 78.76
Angle SC Ply 3 (deg) -45.00 -45.00 73.24
Angle SC Ply 4 (deg) 90.00 90.00 76.12
Aux Web Loc (%c) 6.60 6.60 10.27
Bal Mass 1 (slugs/ft) 0.10 0.11 0.11
Bal Mass 1 Loc (%c) 15.00 12.30 18.50
Bal Mass 2 (kg/m) 0.11 0.10 0.08
Bal Mass 2 Loc (%c) 16.00 16.03 0.23
Constraints Baseline Min FH4 Min FZ4

CG Sec 1 (%c) 24.14 22.39 25.52
CG Sec 2 (%c) 24.1 24.42 28.38
M11 Sec 1 (lb/ft) 0.207 0.236 0.232
M11 Sec 2 (lb/ft) 0.223 0.226 0.197
SC Sec 1 (%c) 36.5 36.23 34.62
SC Sec 2 (%c) 35.9 35.72 34.07
Tor Freq (/rev) 4.53 4.58 3.71
Max Strain (µε) 2175.7 2471.0 4901
Table 2-14: Cross-sectional Properties for the Optimized Case (min FH4)
Sec 1 Sec 2 Sec 1

1.138 1.034 0.934
1.103 1.015 1.247
1.147 1.150 0.88
1.102 1.012 0.974
63
Table 2-15: Structural Frequencies for the Optimized Case
Mode Shape Frequencies (/rev)

Baseline Min FH4 Min FZ4
1st Chordwise bending 0.269 0.269 0.269
1st Flapwise bending 1.0356 1.036 1.0356
2nd Flapwise bending 2.679 2.650 2.59
1st Torsion 4.57 4.584 3.72
2nd Chordwise bending 4.979 4.866 5.019
3rd Flapwise bending 5.48 5.352 4.44
300.0
Min FH4
250.0 Min FZ4
200.0
Change in Vibration (%)
150.0
100.0
50.0
0.0
FX4 FY4 FZ4 MX4 MY4 MZ4 FH4
-50.0
-100.0
Vibratory Load Component
Figure 2-17: Percentage Change in Vibratory 4/rev Hub Loads for the Optimized
Cases
2.5 Concluding Remarks
A powerful, easy-to-use design environment has been developed to support conceptual
and preliminary rotor blade design. It integrates several well-established analysis codes
from different sources: UM/VABS for cross sectional analysis, RCAS for rotorcraft
simulation and ModelCenter for optimization. This design environment enables
conceptual/early preliminary multidisciplinary rotor blade design, allowing rapid design
64
trade studies early in the design process with realistic structural properties for modern
composite rotor blades. The tool supports multiple design scenarios. It can be used in a
cross-section-focused structural design problem in which one might want to find a
feasible structural concept and layup resulting in particular blade stiffness and mass
properties or location of the elastic axis. It can also be used in a more comprehensive
multidisciplinary environment including a full rotorcraft aeromechanical analysis, where
the entire rotor system can be optimized with respect to objectives such as rotor
performance, vibratory loads, etc. subject to aeroelastic and dynamic stability and other
design constraints.
The design environment was successfully used to perform detailed parametric and
optimization studies on the full scale model of a UH-60 composite rotor blade. The cross-
sectional design variables which can be easily modified during the composite rotor blade
manufacturing process were identified and a parametric study was conducted with each
of them. This study was useful in determining the influence of each design variable on
different objective functions and blade dynamic properties. Based on these studies, two
different optimization cases were set up to reduce 4/rev vibratory loads at the rotor hub in
the forward flight condition (µ = 0.24). These optimization problems required complete
cross-sectional and aeromechanic analysis of the rotor blade. The results obtained from
these studies showed:
a) 52% vibration reduction in FZ4 (Objective function: min FZ4)
b) 28% vibration reduction in FH4 (Objective function: min FH4)
where, FZ4 is the amplitude of the 4/rev vibratory vertical force at the hub and FH4 is the
amplitude of the combined 4/rev vibratory load at the hub.
65
Chapter 3. Optimization Framework for the Dynamic
Analysis and Design of Active Twist Rotors
This chapter presents the development of an optimization strategy/framework for the
aeroelastic analysis and design of active twist rotors. The active twist is generated by
piezoelectric material in the form of Active Fiber Composite (AFC) or Macro Fiber
Composite (MFC) embedded in the blade cross section. Proper tailoring of the blade
properties can lead to the maximization of the active twist authority under operating
conditions. Thus, using mathematical optimization, the cross-sectional layout is designed
for an active composite rotor blade to maximize the dynamic active twist while satisfying
a series of constraints on blade cross section parameters, stiffness and strength. The
dynamic twist is defined as the amplitude of twist obtained at the blade tip when the
active plies are actuated in rotating conditions. The optimization problem is solved using
a surrogate-based approach in which the “true” objective function and constraints are
replaced with computationally efficient functional relationships. Since approximation
errors can lead to sub-optimal solutions, the Efficient Global Optimization (EGO)
algorithm, which accounts for uncertainty in surrogate predictions, is employed.
The objectives of the work presented here are:
1) Develop the optimization strategy and framework for the dynamic analysis and
design of active twist rotor blades;
2) Demonstrate the impact of the new design strategy using existing results available in
literature for maximizing static twist per unit length;
66
3) Exemplify the optimization framework by maximizing the amplitude of tip twist for
4/rev actuation in hover conditions; and
4) Perform optimization at a range of actuation frequencies.
3.1 Optimization Framework
The basic flow diagram of the new optimization framework that implements the
strategy described above is shown in Figure 3-1. It consists of two main parts: a) the
ModelCenter-based structural/aeromechanical analysis of active twist rotors, and b) the
surrogate-based optimization with the Efficient Global Optimization (EGO) algorithm
[159]. By replacing the high-fidelity analyses with surrogates, a significant increase in the
robustness of the process is achieved. A description of each of these two main parts is
presented below.
Initial Test Points

Iterative Loop (from LHS)
ModelCenter Analysis
(Complete Aeroelastic Analysis)
SATISFIED
Stopping Criteria Final Result
NOT SATISFIED
Surrogate Model
(Kriging Interpolation)
Global Optimization with

EGO Algorithm
Infill Samples
Figure 3-1: Optimization Framework for Designing Active Twist Rotor Blades
67
3.1.1 High-Fidelity Analysis Framework
The aeromechanic analysis approach described in Figure 3-2 is a modified version of
the work presented Chapter 2 (in Figure 2-3), but now accounting for the presence of
active materials embedded in the blades. In the current analysis, UM/VABS also provides
actuation forces/moments produced by embedded active material to be used in blade
(beam) analysis. The magnitude of the active twisting moment determined using
UM/VABS is used as the amplitude of the external twisting moment applied to nodes of
the blade in the RCAS beam model. Although the active plies generate all the
components of forces and moments [70], it was observed during the preliminary analysis
that only twisting moment is the critical one. The frequency and phase of the twisting
moment are provided by the user or the optimizer. In turn, RCAS evaluates the blade
dynamic twist response for the prescribed frequency range, which will be used as the
objective function.
Figure 3-2: Analysis Framework for Active Twist Rotors
68
3.1.2 Surrogate-based Optimization
The goal of using surrogate method [160, 161] is to replace the true objective function
and constraints with smooth functional relationships of acceptable fidelity that can be
evaluated quickly. To form the surrogate, the objective function must first be evaluated
over an initial set of design points. The surrogate is then generated by interpolating the
initial design points. Although function evaluations coming from the expensive
aeroelastic simulation are needed to form the approximation, this initial investment of
computer time is significantly less than that needed in a global search using non-
surrogate based optimization methods. Once the surrogates have been created, they can
be used to replace the more expensive “true” objective function in the search process for
the global optimum. Moreover, experience shows that for some parameter combinations
of design variables, RCAS analysis does not converge. Therefore, few missing points in
the construction of the surrogates due to failed RCAS runs do not significantly impact
accuracy of the surrogates and the ability of the surrogate-based optimization process to
determine the optimum solution. The increased robustness of the process has a direct
impact on the ability to completely explore the entire design space. In this study, the
objective function and constraints used in the optimization are replaced by surrogates.
Detailed description of the surrogate based modeling technique used in this thesis is
provided in Appendix A.
The MATLAB’s Latin hypercube sampling function “lhsdesign” was used to generate
the space-filling design of experiments used in this study. The points in the Latin
hypercube represent design points at which complete aeroelastic helicopter simulations
are to be conducted. Once an initial set of fitting points has been produced, kriging
69
interpolation [162] is used to create the surrogate for the objective function and
constraints. Kriging interpolation is well suited to approximating nonlinear functions, and
does not require a priori assumptions on the form of the function that is to be
approximated. In kriging, the unknown function of interest, y(x), is assumed to be a
random variable of the form:
where g(x) is an assumed function (usually a low-order polynomial) and Z(x) is a
stochastic (random) process. The kriging surrogates were created with an available
MATLAB toolbox [163].
Once the surrogate objective function is created using kriging, a potential method for
finding the optimum is to optimize the surrogate directly, that is, the “one-shot”
approach. However, if the surrogate is not accurate everywhere in the design space, the
optimization may lead to local optima. Therefore, it is desirable to account for the
uncertainty in the surrogate model since promising designs could lie in regions where the
surrogate is inaccurate. After the first few iterations (2 to 4 iterations) during the
optimization process, it was observed that the EGO algorithm was not able to provide
further improvement to the objective function, and hence, the objective function
predicted by the surrogate model was maximized directly for the next two iterations.
3.1.3 EGO Algorithm
The Efficient Global Optimization (EGO) algorithm [159] is an alternative to the
“one-shot” approach which accounts for uncertainty in the surrogate and is more
efficient. The effectiveness of the EGO algorithm for passive design of helicopter rotors
70
for vibration reduction was demonstrated in [164]. In EGO, a small number of initial
design points are used to fit a kriging approximation. Based on the stochastic nature of
kriging, an expected improvement function (EIF) is created in order to facilitate the
selection of additional sample points (infill samples) where expensive computer
simulations are to be conducted. These sample points are chosen where there is a high
probability of producing a superior design over the current best design and/or where the
predictions of the surrogate are unreliable due to a high amount of uncertainty. These
infill samples represent a balance between the local consideration of finding an optimal
design based on the information in the surrogate, and the global consideration of
sampling in the design space where there is much uncertainty in the surrogate’s
predictions. Therefore, the EGO algorithm is able to adapt to potential errors in the
approximate objective function by sampling at points at which there is much uncertainty
in the surrogate’s predictions. The kriging model is revised after the additional sample
data is added to the initial data set, and the process of choosing additional sample points
is repeated until a user defined criterion is satisfied. In summary, the advantages of such a
method over the “one-shot” approach are: (1) a global search is conducted by sampling in
regions with high uncertainty in the surrogate, and (2) fewer expensive function
evaluations are required since a smaller initial sample set is used and additional sample
points are selected in a more “intelligent” manner, as opposed to starting with a larger
initial data set.
3.2 Numerical Studies
71
In the following sections, numerical studies of the new design strategy and framework
are presented. They are exercised using the original NASA/Army/MIT Active Twist
Rotor blade as the baseline case as described next.
The NASA/Army/MIT Active Twist Rotor (ATR) [65, 165] was originally designed
to study the effects of twist actuation on vibration and noise reduction and performance
improvement in helicopter rotors. The 9-ft-diameter, four-bladed rotor was tested at
NASA LaRC’s Transonic Dynamics Tunnel and was the first-of-a-kind system to
demonstrate vibration reduction using embedded AFC in open and closed loop forward
flight conditions [69]. This particular rotor blade was chosen for this study due to its
known properties and available experimental and computational results [68]. Figure 3-3
shows the planform view of the blade and its corresponding dimensions. The airfoil for
this blade is the NACA 0012 and it is uniform along the blade radius. The reference
cross-sectional layup is shown in Figure 3-4, while Table 3-1 lists the ply angles for all
the plies used in the cross section of the rotor blade. Among the plies used, E-Glass is
bidirectional while, S-Glass and AFC plies are unidirectional. The specific material
properties can be found in Table 3-2.
Figure 3-3: Planform View of the ATR Rotor Blade (Dimensions in Meters) [62]
72
The characteristic properties of the baseline ATR blade and its structural frequencies
at 100% RPM are listed in Table 3-3. Blade structural frequencies in vacuum were
obtained using RCAS. For the rotor dynamic analysis, the trim option (wind tunnel trim)
is included in the RCAS model. The trim targets used in the analysis are: C T = 0.0066, no
cyclic moments (Mx = 0 and My = 0), and the blade pitch settings are used as the trim
variables. (Note that the value of CT used in the numerical analyses performed here is the
same as that obtained in the experimental analysis presented in [69].) The mean value of
the hub loads and the amplitude of the 4/rev component of the hub loads in the fixed
system for the baseline blade (with no twist actuation) at an advance ratio of 0.24 are
given in Table 3-4 where Fx, Fy, and Fz represent the components of the hub force in the
non-rotating frame, while Mx, My, and Mz represent the components of the moments at
the hub.
SparPlies
Web Plies Ply 1: E-Glass
Ply 6: E-Glass Quarter Ply 2: AFC
Ply 7: E-Glass Chord Ply 3: E-Glass
Ply 4: AFC
Fairing Ply Spar web Ply 5: E-Glass
Ply 1: E-Glass Mass_2 Mass_1
Nose Plies
Ply 1: E-Glass
Ballast mass Location Ply 2a: S Glass
Spar End Location Ply 3: E-Glass
Ply 5: E-Glass
Figure 3-4: Cross-Sectional Shape of the Rotor Blade (NACA 0012 Airfoil)
Table 3-1: Baseline ATR Cross Section Ply Angles
Ply # Angle Ply # Angle

Ply 1 0/90 Ply 6 0/90
Ply 2 45 Ply 7 0/90
Ply 2a 0
Ply 3 ±45
Ply 4 -45
Ply 5 0/90
73
Table 3-2: Material Properties
E-Glass S-Glass AFC

Thickness (μm) 114.3 230 200
Density (kg/m3) 1720 1860 4060
E11 (GPa) 20.7 43.4 30.2
E22 (GPa) 20.7 12 14.9
E33 (GPa) 20.7 12 14.9
G12 (GPa) 4.1 3.6 5.13
G13 (GPa) 4.1 3.6 5.13
G23 (GPa) 4.1 3.6 5.13
ν12 0.13 0.28 0.454
ν13 0.13 0.28 0.454
ν23 0.3 0.3 0.3
Table 3-3: Characteristics of the Baseline ATR
Rotor Type Fully Articulated

Number of blades 4
Blade radius (R ) 1.397 m
Blade Chord (c ) 0.1077 m
Airfoil Section NACA 0012
Blade Pretwist -10 deg
Hinge Offset 0.0762 m
Rotor Speed 687.5 RPM
CT 0.0066
Air density 2.42 kg/m3
Mode Shape Frequency (/rev)

st
1 Chordwise bending 0.29
1st Flapwise bending 1.04
2nd Flapwise bending 2.78
3rd Flapwise bending 5.34
2nd Chordwise bending 5.76
1st Torsion 6.51
74
Table 3-4: Hub Loads for the Baseline ATR Case (μ = 0.24)
Mean Values 4/rev Amp

Fx (N) 8.85 1.38
Fy (N) 13.39 2.04
Fz (N) 990.7 23.56
Mx (Nm) 0.44 40.52
My (Nm) 1.19 36.24
Mz (Nm) 47.41 1.06
3.3 Optimization Results
Optimization studies were conducted to maximize the static twist per unit length (twist
rate obtained when a constant DC voltage is given to active plies) and to maximize the
dynamic tip twist amplitude (amplitude of tip twist obtained when a sinusoidal input
voltage is given to active plies at a fixed actuation frequency) in hover conditions (μ =
0.0) with wind tunnel trim. The cases considered in this chapter are similar to the static
twist optimization cases presented in [86]. This was done to verify the results obtained
for static case using the current framework. For all the active twist optimization studies
presented in this thesis, the amplitude of actuation voltage was fixed at 1000V.
For the results presented in this chapter, only six design variables were used, namely,
the chordwise location of the main spar web, the chordwise ending location of the
spar/AFC plies, and the magnitude (m1,m2) and location (x1,x2) of the ballast masses. The
design variables used in the current study and their upper and lower bounds are listed in
Table 3-5, while the constraints used are given in Table 3-6. For all the variables, the
initial value and the bounds used were the same as in [86], whenever they were available.
For the remaining cases, reasonable values were used for bounds such that the
optimization process was not affected.
75
Table 3-5: Design Variable Used for Optimization Study
Design variables Baseline Min Max

Main Spar Loc (c) 0.443 0.2 0.85
Spar End (c) 0.443 0.2 0.85
Ballast Mass 1 (m1) (kg/m) 0.222 0 0.5
Ballast Mass 1 Loc (x1) (c) 0.443 0 0.8
Ballast Mass 2 (m2) (kg/m) 0.23 0 0.5
Ballast Mass 2 Loc (x2) (c) 0.02 0 0.8
Table 3-6: Constraints used in Optimization
Constraints Baseline Max Min

SC (%c) 20.95 25 17
CG (%c) 24.3 28 20
M11 (kg/m) 0.677 0.65 0.72
1st Tor Freq (/rev) 6.51 4.5 8
Max |ε11| (με) 2747 6000 0
Max |ε12| (με) 3807 6000 0
3.3.1 Construction and Verification of the Surrogate Models
For each of the constraints and objective function considered, a surrogate model was
developed using MATLAB’s kriging toolbox described above. To test the surrogate
model, a sample of 30 test points (different from those used to construct the surrogate)
was generated. The complete aeroelastic simulation was performed for those 30 cases and
the results obtained were compared with those obtained from the surrogate model. The
values of average error, maximum error and standard deviation of the error obtained from
this analysis are shown in Table 3-7. Results show that the surrogate model prediction for
static twist, CG location, and mass per unit length are very accurate. For other cases, the
mean value of the error is less than 12%. However, for all the cases, it was observed that
the surrogate model was able to capture well the qualitative trends of the problem.
76
Table 3-7: Error Obtained in the Prediction of Design Variable and Constraints for
Surrogate models
Percentage Error (%)

Variable Mean Max Std Dev
θdyn,4/rev 11.77 43.06 9.34
θstat 0.93 2.01 0.55
CG 0.39 0.81 0.22
SC 8.65 46.78 10.67
1st Tor Freq 9.54 19.35 5.47
ε11 6.83 26.72 5.82
M11 0.01 0.05 0.03
ε12 10.17 34.78 9.41
3.3.2 Optimization Cases
3.3.2.1 Static Twist Optimization (Max θstat)
To verify the framework, one of the cases presented in [86] is studied first. In this
case, the static twist per unit length (θstat) obtained from the cross-sectional analysis was
defined as the objective function. Final results are compared with those in [86] and are
summarized in Table 3-8. It should be noted that the constraints shown in Table 3-8 are
non-dimensionalized such that a value of “0” represents their lower bound while a value
of “1” represents their upper bound.
The final result obtained for the “Max θstat” case is very close to that obtained in [86].
It can be seen that in both the cases, there is an increase in mass per unit length due to the
increase in length of the active plies used in the cross section. This is also accompanied
by a simultaneous increase in torsional stiffness of the cross section. However, the
increase in torsional stiffness is less when compared to the increase in torsional inertia
that occurs due to addition of the plies and the relocation of the ballast masses further
77
away from the reference axis (quarter chord). Thus, the optimized results in both cases
show an overall decrease in torsional frequency. The vertical spar web in both results is
located very close to the half chord. The only dissimilarity between the two sets of results
lies in the location of ballast masses and the chordwise location of the CG.
Table 3-8: Static Twist Optimization Result
Max θstat Ref. [86] Baseline

θstat (deg/m) 1.59 NA 1.34
% Increase 18.6 19.0 -
Non-Dimensional Constraints
st
1 Tor Freq 0.349 0.353 0.738
M11 0.965 0.714 0.486
SC 0.32 0.25 0.21
Max ε11 0.516 NA 0.458
Max ε12 0.690 NA 0.635
CG 0.655 0.475 0.493
Design Variables
m1 (kg/m) 0.001 NA 0.222
x1 (c) 0.793 0.418 0.443
m2 (kg/m) 0.311 NA 0.23
x2 (c) 0.006 0.045 0.02
Spar Web (c) 0.521 0.49 0.443
Spar End (c) 0.838 0.85 0.443
NA: Not Available
3.3.2.2 Dynamic Twist Optimization (Max θdyn,4/rev)
In this case, the amplitude of dynamic twist at the blade tip (θdyn,4/rev) was maximized
for a 4/rev actuation frequency. In the preliminary analysis that was carried out with the
baseline case, it was observed that the amplitude of dynamic twist does not vary
significantly with advance ratio. Thus to avoid unnecessary calculations, the flow was set
for hover conditions (μ = 0.0). In order to make sure that this is indeed the cases for the
optimized cases, aeroelastic analysis was performed in forward flight condition in Section
78
3.7. In the results shown in Table 3-9, it can be seen that the case with maximum static
twist does not coincide with the case with maximum dynamic twist. The case with
maximum static twist shows an increase of 57% in dynamic twist, while the case
optimized for maximum dynamic twist shows an increase of 63%. Also, the amount of
active material used in “Max θdyn,4/rev” case is 20% less than that used in “Max θstat” case.
In the “Max θdyn,4/rev” case, the spar plies (which includes AFC plies) end at 0.68c while
in case of “Max θstat”, spar plies extend to 0.85c. In both “Max θstat” and “Max θdyn,4/rev”
cases, there is a decrease in torsional frequency; however, the decrease is more
pronounced in “Max θdyn,4/rev” case. The shear center for “Max θdyn,4/rev” case is very close
to its lower limit, implying that it is a critical constraint for dynamic twist optimization.
The total ballast mass used in “Max θdyn,4/rev” case is more than that used in “Max θstat”
case and it is distributed further away from the reference axis, resulting in higher
torsional inertia.
Thus, in the “Max θdyn,4/rev” case, higher dynamic tip twist is obtained when compared
to the baseline ATR design due to increase in active ply coverage used in the cross
section and dynamic tuning of the blade properties. The shape of the blade cross section
for the baseline case and the optimized cases are shown in Figure 3-5. The ballast masses
used in the cross section are represented by “blue” and “red” circles.
79
Table 3-9: Dynamic Optimization Results
Max θstat Max θdyn,4/rev Baseline

θstat (deg/m) 1.59 1.46 1.34
% Increase 18.6 8.94 -
θdyn,4/rev (deg) 3.45 3.58 2.19
% increase 57.6 63.6 -
Non-Dimensional Constraints
st
1 Tor Freq 0.349 0.253 0.738
M11 0.965 0.513 0.486
SC 0.320 0.127 0.214
Max ε11 0.516 0.485 0.458
Max ε12 0.690 0.657 0.635
CG 0.655 0.976 0.493
Design Variables
m1 (kg/m) 0.001 0.066 0.222
x1 (c) 0.793 0.764 0.443
m2 (kg/m) 0.311 0.282 0.23
x2 (c) 0.006 0.005 0.02
Spar Web (c) 0.520 0.488 0.443
Spar End (c) 0.838 0.682 0.443
Figure 3-5: Baseline and the Optimized Cross Sections
80
3.3.3 Analysis of Optimized Results
3.3.3.1 Relation between blade twist and induced FZ4
To determine the relation between static tip twist, dynamic tip twist, and 4/rev vertical
hub shear (FZ4) induced by the twist actuation in hover condition, results obtained for all
the iterations (during the optimization process) that satisfy all the constraints were plotted
in order of increasing static twist in Figure 3-6, and in order of increasing dynamic twist
in Figure 3-7. Note that for the hover condition, “Induced FZ4” is a measure of control
authority of the active twist actuation. Hence, in order to maximize the control authority,
it is desired to maximize induced FZ4. The dashed black lines in both the figures
correspond to the baseline results. Here, it can be clearly seen that an increase in static tip
twist may not always result in an increase in dynamic tip twist and induced FZ4. However,
an increase in dynamic tip twist amplitude results in a proportional increase in F Z4
amplitude. Thus, the FZ4 induced at the rotor hub by actuation of the embedded active
material inside the blade cross section is proportional to the dynamic tip twist amplitude
of the blade.
81
1.6
stat (deg/m)
1.4
0 20 40 60 80 100 120 140
dyn,4/rev (deg) 4
0
0 20 40 60 80 100 120 140
Induced FZ4(N)
600
400
200
0 20 40 60 80 100 120 140
Sorted Iteration Cases
Figure 3-6: Objective Function Results in Order of Increasing Static Twist
1.6
stat (deg/m)
1.4
0 20 40 60 80 100 120 140
4
dyn,4/rev (deg)
0
0 20 40 60 80 100 120 140
Induced FZ4(N)
600
400
200
0 20 40 60 80 100 120 140
Sorted Iteration Cases
Figure 3-7: Objective Function Results in Order of Increasing Dynamic Twist
82
3.3.3.2 Effect of Frequency of Actuation
In the next step, frequency of actuation was varied for all three cases, namely, baseline
case, maximum static twist case (Max θstat), and maximum dynamic twist case (Max
θdyn,4/rev) at μ = 0.0. Results described in Figure 3-8 show that the “Max θdyn,4/rev” case
consistently provides maximum dynamic tip twist as compared to the other two cases,
except at the 5/rev actuation case. This is due to the fact that “Max θdyn,4/rev” case is
optimized for 4/rev twist actuation frequency. Figure 3-9 shows effect of frequency of
actuation on amplitude of vertical displacement at the blade tip, which shows similar
behavior as was observed for tip twist. The only difference is that all the cases here show
a peak close to 3/rev frequency that coincides with the blade second flapwise bending
frequency.
5
Baseline
Tip Twist Amplitude (deg)
Max stat
4 Max 
dyn
1
2 3 4 5
Frequency of actuation (/rev)
Figure 3-8: Effect of Frequency on Dynamic Twist
83
12
Baseline
Vertical Tip Amplitude (mm)

Max stat
10
Max 
dyn
8
2
2 3 4 5
Frequency of actuation (/rev)
Figure 3-9: Effect of Frequency on Amplitude of Vertical Tip Displacement
3.3.3.3 Effect on Vibratory Loads in Forward Flight Conditions for Zero Twist
Actuation
The blade designs obtained from the surrogate optimization for “Max θdyn,4/rev” and
“Max θstat” cases have different dynamic properties as compared to the baseline one.
Even though the optimized cases have higher twist actuation and, therefore, higher
control authority for vibration reduction using active twist actuation, it is desired to know
the vibratory characteristics of each design in the absence of any twist actuation. Thus, in
order to determine the baseline vibratory loads in forward flight conditions, all three
cases were run at an advance ratio of 0.24 with wind tunnel trim, and 4/rev vibratory
loads at the hub were determined. The percentage increase in vibratory loads for “Max
θdyn,4/rev” and “Max θstat” cases is shown in Figure 3-10. (Since the RCAS model used in
this analysis did not include the free wake mode for capturing BVI effects and a dynamic
stall model to account for the dynamic stall effects, an advance ratio of 0.24 was used to
minimize contribution from BVI and dynamic stall effects).
Results show a 10% increase in FZ4 and a 15% increase in MY4 for “Max θdyn,4/rev” case
and a 7-8% increase in FZ4 and MY4 for “Max θstat” case. Although the FY4 component
84
shows a higher percentage increase, results presented in Table 3-4 show that the
amplitude of vibration for the FY4 component is very small. The increase in vibration for
the optimized cases can be attributed to the decrease in torsional frequency.
40
Max stat
Max dyn
4/rev Vibratory loads (%)

30
20
10
-10
Fx Fy Fz Mx My Mz
Figure 3-10: Percentage Increase in 4/rev Vibratory Loads at the Rotor Hub with no
Actuation at μ = 0.24
3.3.3.4 Circle Plot for the Optimized Results
The circle plots for 4/rev vibratory loads at the rotor hub at µ=0.24 were obtained for
all the hub load components. In this analysis, the phase angle of actuation was varied,
while the frequency of actuation was kept constant at 4/rev. Figure 3-11 and Figure 3-12
show circle plots obtained for FZ4 and MY4 (4/rev component of the shear force and
lateral cyclic moment at the rotor hub, respectively). In these figures, non-actuated
vibratory loads (loads in the absence of twist actuation) for each of the cases are
represented by “x” while the origin is represented by “*”. The straight line drawn in these
plots joins the non-actuated vibratory load to vibratory load corresponding to 0-deg phase
actuation for each of the cases.
85
As it is shown in Figure 3-10, baseline vibratory loads for each of the cases considered
are very close to each other. From the obtained circle plots, it can be concluded that the
“Max θdyn,4/rev” case provides maximum control authority for vibration reduction. Since
the current analysis is done for µ =0.24, the baseline FZ4 load is small and very small
twist actuation is required to reduce FZ4 to zero. However, in case of the MY4 component,
larger twist actuation is required to minimize the vibratory loads. Similar analysis can be
conducted at different advance ratios.
600
Baseline
400 Max dyn
200 Max stat

FZ4,sin(N)
-200
-400
-600
-800
-1000 -500 0 500 1000
FZ4,cos(N)
Figure 3-11: Vibratory Hub Vertical Shear Force (μ = 0.24)
150
Baseline
Max dyn,4/rev
100 Max stat
M Y4,sin(Nm)
50
-50
-150 -100 -50 0 50
M Y4,cos(Nm)
Figure 3-12: Vibratory Hub Lateral Cyclic Moment (μ = 0.24)
86
3.3.3.5 Effect of Aeromechanic Analysis on Optimized Results
The analysis performed until now for dynamic twist optimization included trim option
(wind tunnel trim) for aeroelastic analysis. This was done so that the blade experiences
accurate aerodynamic loads. However, the trim analysis is very time consuming and each
run in RCAS takes 15-20 min for a complete aeroelastic analysis. As a result, following
two simplifications were considered to the analysis:
Case 1: Periodic Analysis: In this case, the pitch settings are kept constant and a periodic
solution is obtained. The rest of the analysis variables are kept the same. Thus, the blade
experiences similar aerodynamic stiffness (aerodynamic forces per unit blade twist) as in
the trim cases but the magnitude of aerodynamic loads is small since the initial pitch
settings used are very close to zero. The computation time (on an Intel Core 2 Quad
[email protected] GHz) required for a “Periodic Analysis” (~ 1 min) is an order of magnitude
less than the computational time required for the “Trim Analysis” (~15 min). The
“Periodic Analysis” can only be used to approximate the amplitude of blade deformation
due to actuation of on-blade active devices. The amplitude and mean value of tip twist for
active twist actuation at 3, 4 and 5/rev actuation frequencies for the baseline ATR blade
at μ = 0.0 are shown in Figure 3-13. As show in Figure 3-13, the amplitude of tip twist
predicted by “Trim Analysis” and “Periodic Analysis” are very close to each other,
however there is a significant difference in the mean value of tip twist predicted by the
two analyses.
87
Tip Twist Amp(deg)
3
2.5
2 Trim
Periodic
1.5
3 3.5 4 4.5 5
Mean Tip Twist (deg)

10
0
3 3.5 4 4.5 5
Actuation Frequency (/rev)
Figure 3-13: Variation of Mean Value and Amplitude of Tip Twist for twist
actuation at μ=0.0
Case 2: Low density analysis: In this case, a periodic solution is obtained, as in Case 1,
but the density of the medium is reduced significantly (from 2.42kg/m3 to 0.3kg/m3).
Thus, the aerodynamic stiffness is much smaller (almost an order of magnitude
reduction) in this case. This analysis was done to check if the optimization study can be
done in vacuum (without aerodynamic loads).
The design variables and the constraints used in the optimization are the same as in the
earlier study. In both these cases, the blade does not experience complete aerodynamic
loads as it would experience in a trim analysis. As a result, the constraints on the cross-
sectional strains were removed. As expected, the strains observed during these studies
were well below their upper bound for all the cases.
The values of dynamic twist obtained for each of the cases are shown in Table 3-10.
At the end of optimization process, it was observed that the optimized case obtained from
trim analysis and the optimized case obtained from periodic analysis is the same and is
referred as “Max θdyn,4/rev,T/P” case. The optimized result obtained from analysis at low
88
density medium is referred as “Max θdyn,4/rev,low ρ ” case, while the case corresponding to
maximum static twist per unit length is denoted “Max θstat ” as before. Results also show
how each case performs in different analysis conditions. (Note that the optimized solution
obtained for “Max θdyn,4/rev,T/P” and “Max θstat” cases in the current analysis are slightly
better than the optimized solution obtained for the Max θstat and Max θdyn,4/rev case in
Section 3.3.2.1 and Section 3.3.2.2, respectively.)
Table 3-10: Optimized Results
Static Twist Dynamic Twist

Trim Periodic Low ρ
Analysis Analysis Analysis
(deg) (deg) (deg) (deg)
Baseline Case 1.34 2.19 2.17 2.37
Max θdyn,4/rev,T/P Case 1.42 4.16 4.00 5.90
Percentage Increase (%) 5.82 90.20 84.42 148.97
Max θdyn,4/rev,low ρ Case 1.40 3.99 3.89 6.98
Max θstat Case 1.61 3.91 3.80 4.18
The maximum increase obtained in static twist per unit length is 20%. For this case,
the increase in dynamic twist is significant but less than the optimal. The “Max
θdyn,4rev,T/P” case shows a maximum increase of 90% in dynamic twist in aerodynamic
conditions with trim. Although the trim analysis and periodic analysis have the same
optimized result, the dynamic twist corresponding to these analyses are slightly different
(~4%). During the optimization, it was observed that the dynamic twist obtained from
trim analysis was consistently higher than that obtained from periodic analysis, but the
difference between them is small. The “Max θdyn,4rev, low ρ” case shows an increase of
194% in tip twist amplitude in low density analysis, which corresponds to approximately
89
80% increase in twist amplitude in trim and periodic analysis. Also, it should be noted
that the dynamically optimized cases have only 5-6% higher static twist per unit length as
compared to the baseline case. Thus, the increase in dynamic twist is due to tailoring of
the dynamic properties of the rotor blade.
Table 3-11: Optimization Constraints and Other Parameters
Baseline Max θdyn,4rev,T/P Max θdyn,4rev,low ρ Max θstat

Constraints Case Case Case Case
st
1 Tor Freq (/rev) 6.34 4.60 4.51 5.199
M11 (kg/m) 0.68 0.70 0.712 0.710
SC (%c) 18.71 20.75 17.49 17.01
CG (%c) 23.95 27.47 25.06 27.06
Other parameters
S44 (Nm2) 37.71 44.32 41.46 50.47
Act Mom Mx (Nm) 0.91 1.12 1.07 1.48
2nd Flap Freq (/rev) 2.76 2.79 2.79 2.82
The values of the constraints for the optimized cases are shown in Table 3-11. All the
cases show a decrease in the torsional frequency in spite of an increase in the cross-
sectional torsional stiffness. This occurs due to the redistribution of ballast masses such
that there is a net increase in the torsional inertia for the blade cross section. Also, for the
dynamically optimized cases, the first torsional frequency is very close to the lower
bound of the 1st torsion frequency and the actuation frequency of 4/rev. For the Max
θdyn,4/rev,lowρ case, the torsion frequency is exactly at the lower bound, implying that the
optimization at low density is driven mainly by the constraint on the 1st torsion frequency
and the optimizer tries to get the torsion frequency as close as possible to the actuation
frequency of 4/rev. All the cases show an increase in mass per unit length, which happens
due to the increase in the active plies used in the cross section. The chordwise location of
90
SC and CG for the optimized sections is close to its lower bound and upper bound,
respectively.
The optimized cases have higher cross-sectional torsional stiffness and they also
produce higher active twisting moment as compared to the baseline case. This occurs due
to the increase in the coverage of plies used in the cross section. The torsional stiffness
and active moment is highest for the Max θstat case, as expected, since it has the
maximum amount of active ply in the cross section.
Table 3-12: Design Variables for the Optimized Cases
Baseline Max θdyn,4rev,T/P Max θdyn,4rev,low ρ Max θstat

Case Case Case Case
Spar End ( c) 0.443 0.556 0.590 0.850
Spar Web Loc ( c) 0.443 0.501 0.460 0.481
Mass (m1) (kg/m) 0.23 0.109 0.095 0.075
Mass 1 Loc (x1) ( c) 0.02 0.831 0.830 0.002
Mass 2 (m2) (kg/m) 0.222 0.313 0.321 0.224
Mass 2 Loc (x2) ( c) 0.443 0.000 0.001 0.001
The design variables for the optimized cases are listed in Table 3-12. The chordwise
ending location of active plies is at its upper limit for the Max θstat case. As a result of
this, both the ballast masses for the Max θstat case are close to the leading edge. In the
case of dynamically optimized cases, the ballast masses are located on either side of the
quarter chord which results in an increase in torsional inertia. The optimum result
obtained from trim analysis and low density analyses are close to each other. The
difference lies in the fact that in the case of low density analysis, the torsional frequency
is the only driving factor whereas in the case of periodic/trim analysis, active moment
generated by active plies and aerodynamic stiffness also influence the results.
91
3.3.4 Trend Analysis
In order to observe the trends in the variation of the objective function and other
response variables, a set of 40 design points was created using the results obtained from
all the optimization studies. For these 40 points, all three of the analyses, namely: the trim
analysis, the periodic analysis, and the low density analysis was carried out. Figure 3-14
shows the variation of tip twist amplitude, torsional stiffness, and first torsion frequency
with iteration number when the data is sorted in order of increasing dynamic twist
obtained from low density analysis. Results show that the amplitude of dynamic twist
obtained from the low density analysis is directly related to the torsional frequency of the
blade. The amplitude of tip twist increase as the torsion frequency approaches actuation
frequency. The dynamic twist obtained from the trim analysis and periodic analysis does
increase with decrease in torsion frequency, however, the variation is not uniform and
thus, the torsion frequency is not the only critical parameter. Other variables shown in
Figure 3-15 do not show any consistent trend. It should be noted that the amplitude of
vertical displacement at the blade tip is highest for trim analysis for all the cases
considered since the trim analysis case experiences higher aerodynamic loads.
The results presented in this section can be used to conclude that it is sufficient to
carry out “periodic analysis” instead of the more time consuming “trim analysis” for
active twist optimization studies. However, a purely structural dynamic solution is not
sufficient and would not lead to an optimal solution as observed in the “Low ρ Analysis”.
92
7
dyn (deg)
5
3
0 5 10 15 20 25 30 35 40
60
50
GJ(Nm2)
40
30
0 5 10 15 20 25 30 35 40
Periodic Analysis
7 Trim Analysis
1st Tor Freq(/rev)
Low  Analysis
6
3
0 5 10 15 20 25 30 35 40
Sorted Iterations
Figure 3-14: Variation of Amplitude with Torsional Stiffness and 1st Torsion
Frequency (sorted with respect to amplitude obtained from low density analysis)
-3
x 10
8 1.6
Act Mom(Nm)
6 1.4
Zdisp(m)
4 1.2
Periodic Analysis
2 Trim Analysis 1
Low  Analysis
0 0.8
0 10 20 30 40 0 10 20 30 40
28 25
26
CG(%c)
SC(%c)
24 20
22
20 15
0 10 20 30 40 0 10 20 30 40
1 0.7
Spar Web(c)
Spar End(c)
0.8 0.6
0.6 0.5
0.4 0.4
0 10 20 30 40 0 10 20 30 40
Sorted Iterations Sorted Iterations
Figure 3-15: Variation with Iteration Number for Other Parameters (sorted with
respect to amplitude obtained from low density analysis)
93
3.4 Optimization at 3/rev Actuation Frequency (Max θ3/rev)
After completion of the optimization at 4/rev actuation frequency, optimization was
done at 3/rev actuation frequency. The constraints and design variables used were kept
the same as in the earlier optimization problems. The final results obtained are shown in
Table 3-13 and compared with results obtained with Max θstat and Max θdyn,4/rev (now
referred as θ4/rev) cases (optimized result corresponding to 4/rev actuation frequency).
Based on the observation made in the previous section, the “Periodic Analysis” is used
here.
Table 3-13: Optimized Results for 3/rev Actuation Frequency
Cases Max θstat Max θ3/rev Max θ4/rev Baseline

θstat (deg/m) 1.61 1.40 1.42 1.34
% Increase 20.02 4.75 5.82 -
θ3/rev (deg) 2.81 2.96 2.72 1.94
% Increase 44.97 52.59 40.18 -
Constraints
1st Tor Freq (/rev) 5.20 4.51 4.60 6.34
M11 (kg/m) 0.710 0.711 0.703 0.684
SC (%c) 17.00 17.03 20.75 18.71
CG (%c) 27.05 27.97 27.47 23.94
Design Variables
Spar End (c) 0.850 0.593 0.556 0.443
Spar Web Loc (c) 0.481 0.455 0.501 0.443
Mass m1 (kg/m) 0.224 0.098 0.109 0.222
Mass 1 Loc (x1) (c) 0.001 0.826 0.831 0.443
Mass m2 (kg/m) 0.075 0.315 0.313 0.230
Mass 2 Loc (x2) (c) 0.002 0.010 0.000 0.020
Other parameters
S44 (Nm2) 50.47 41.16 44.32 37.71
Active Moment (Nm) 1.48 1.07 1.12 0.91
2nd Flap Freq (/rev) 2.82 2.80 2.79 2.76
94
The optimized result shows a 52% increase in the dynamic twist at 3/rev actuation
frequency. The Max θstat and the Max θ4/rev cases show a 44% and 40% increase in
dynamic twist, respectively. As observed earlier for optimization at 4/rev frequency, the
increase in static twist for Max θ3/rev is only 4.75%. The Max θ3/rev and Max θ4/rev cases
show similar behavior for ballast masses and CG location. There are small variation in
location of SC, torsional stiffness, length of active plies, and location of vertical spar
web. As a result of this, the torsional frequency of the optimized blade is at its lower
bound of 4.5/rev. The trend observed for Spar End (chordwise location where the
spar/active plies end) shows that a higher twist amplitude can be obtained either through
an increase in the amount of active material in the cross section or by dynamic tuning of
the blade stiffness properties with significantly less active material. The vertical spar web
is located near the mid chord in the optimum cases. Between the ballast masses used,
ballast mass m2 is higher in magnitude and is located very close to the leading edge of the
blade cross section to get the CG close to the quarter chord. The ballast mass m1 is much
lower in magnitude and its location varies. For the dynamically optimized case, where the
higher dynamic twist is obtained by dynamic tuning, the ballast mass m1 is located aft of
mid chord.
3.5 Optimization at 5/rev Actuation Frequency (Max θ5/rev)
In the next step, the optimization was done at 5/rev actuation frequency. The design
variables and constraints used in the optimization were kept the same as in earlier studies.
Final results obtained are shown Table 3-14. The table also includes optimized results
corresponding to 3/rev and 4/rev actuation frequency and the dynamic twist obtained for
these cases at 5/rev actuation frequency.
95
Table 3-14: Optimization Results for 5/rev Actuation Frequency
Cases Max θ5/rev Max θ4/rev Max θ3/rev Max θstat Base
θstat (deg/m) 1.6 1.42 1.4 1.61 1.34
% Increase 19.48 5.82 4.75 20.02 -
θ5/rev (deg) 5.17 2.57 2.35 4.85 2.56
% Increase 101.83 0.24 -8.22 89.44 -
Design Variables
Spar End (c) 0.848 0.556 0.593 0.85 0.443
Spar Web Loc (c) 0.561 0.501 0.455 0.481 0.443
Mass 1 (m1) (kg/m) 0.01 0.109 0.098 0.075 0.222
Mass 1 Loc (x1) (c) 0.447 0.831 0.826 0.002 0.443
Mass 2 (m2) (kg/m) 0.299 0.313 0.315 0.224 0.23
Mass 2 Loc (x2) (c) 0.002 0 0.01 0.001 0.02
Constraints
1st Tor Freq (/rev) 5.47 4.6 4.51 5.2 6.34
M11 (kg/m) 0.718 0.703 0.711 0.71 0.684
SC (%c) 22.5 20.75 17.03 17 18.71
CG (%c) 27.9 27.47 27.97 27.05 23.94
Other parameters
S44 (Nm2) 55.43 44.32 41.16 50.47 37.71
Active Moment (Nm) 1.57 1.12 1.07 1.48 0.91
2nd Flap Freq(/rev) 2.82 2.79 2.8 2.82 2.76
The final optimized result shows a 101% increase in amplitude of dynamic tip twist
for 5/rev actuation frequency. Also, it is interesting to note that Max θ4/rev and Max θ3/rev
cases have poor twist amplitude at 5/rev actuation frequency, whereas Max θstat case
performs well even at 5/rev actuation frequency. This is due to the placement of 1st
torsion frequency for the Max θstat case. For the Max θ5/rev case, higher twist amplitude is
obtained due to the combination of both: higher amplitude of active twisting moment and
dynamic tuning. In the 3/rev and 4/rev actuation cases, the 1st torsion frequency of the
96
blade could not get very close to the twist actuation frequency due to the constraint on the
minimum value for the first torsion frequency. However, for 5/rev actuation case, the
actuation frequency is in the range of allowable torsional frequencies. As a result, the
percentage increase in dynamic twist amplitude for 5/rev actuation frequency is larger
than that obtained for 3/rev and 4/rev actuation cases. Another important observation
from the analysis is that the torsion frequency for the optimized cases is not exactly at
5/rev frequency. This can be attributed to aerodynamic forces which act against the
motion of the rotor blades.
Among the design variables, the value of 0.848c for design variable “Spar End”
implies that the coverage of the active region is close to the maximum allowable value,
indicating that the optimizer is trying to maximize the active twisting moment. It should
be noted that in Max θ5/rev case, both active twisting moment and torsional stiffness are
higher than that for Max θstat case even though the amount of active material used in the
cross section is the same in both cases. This is due to the difference in the location of the
vertical spar web for these cases. The increase in active twisting moment for the Max
θ5/rev case is offset by an even larger increase in torsional stiffness of the blade. As a
result, the static twist for Max θ5/rev case is smaller than that obtained for Max θstat case.
For optimization at 5/rev frequency, the mass per unit length and the chordwise location
of the CG are critical constraints and they both are closer to their upper limit.
3.6 Optimization at a Range of Actuation Frequencies (Max θ345/rev)
In this case, the objective is to maximize the amplitude of tip twist at a range of
actuation frequencies, which maybe required for vibration and noise reductions. In this
97
particular case, the objective function includes amplitude of dynamic twist at 3/rev, 4/rev
and 5/rev actuation frequencies since this a four-bladed rotor. During the optimization
studies at these actuation frequencies, it was observed that the amplitude of tip twist for
the optimized design for each of the cases was different. Thus, in order to remove the bias
towards a particular frequency, the amplitude corresponding to each frequency was non-
dimensionalized by the maximum amplitude obtained when the optimization was done at
that particular frequency. The objective function used is given by:
where, θ3/rev,max is the maximum amplitude of tip twist obtained from optimization at
3/rev actuation frequency, θ4/rev,max is the maximum amplitude of tip twist obtained from
optimization at 4/rev actuation frequency, and θ5/rev,max is the maximum amplitude of tip
twist obtained from optimization at 5/rev actuation frequency. The design variables and
constraints used in the optimization were kept the same as in the earlier studies.
The final result obtained from all the cases considered are shown in Table 3-15. The
columns in Table 3-15 show the value of non-dimensionalized tip twist amplitude for
different frequencies and the value of for all the optimum cases. As expected,
the Max θ345/rev case shows high twist amplitude at all the actuation frequencies. Max
θ3/rev and Max θ4/rev cases show high amplitude of twist for 3/rev and 4/rev actuation
frequencies, respectively, however their performance deteriorates at 5/rev actuation
frequency since the 1st torsional frequency for both these cases lies close to 4.5/rev (lower
bound for allowable torsion frequency). Also, results shown in Table 3-15 indicate that
98
Max θ345/rev case and Max θstat case are very close to each other and maximizing static
twist would have been sufficient to maximize the twist amplitude over a range of
actuation frequencies. However, this is only partially true, since the increase in dynamic
twist amplitude occurs due to both dynamic tuning and higher active twisting moment.
For the particular case being analyzed, the first torsion frequency for the Max θ stat case
lies very close to the first torsion frequency obtained for Max θ345/rev case as seen in the
results shown in Table 3-16.
Table 3-15: Results obtained from all the Optimization Cases
Cases  3/rev  4/rev  5/rev 345/rev

Max θ345/rev 0.95 0.94 0.97 0.96
Max θ5/rev 0.86 0.84 1.00 0.90
Max θ4/rev 0.92 1.00 0.50 0.80
Max θ3/rev 1.00 1.02 0.45 0.83
Max θstat 0.95 0.95 0.94 0.95
Baseline 0.65 0.54 0.50 0.56
3/ rev  4/ rev 5/ rev

where  3/ rev  ,  4/ rev  , and  5/ rev  .
3/ rev ,max  4/ rev ,max 5/ rev ,max
The final values of the objective function, constraints, and design variables for the
optimized case are shown in Table 3-16. As observed earlier, the Max θstat and Max
θ345/rev cases are very close to each other. Among the dynamically optimized cases, Max
θ5/rev case is the closest to optimum due to the location of the first torsional frequency.
The critical parameters for the optimization conducted at 3, 4 and 5/rev actuation
frequencies are mass per unit length and chordwise location of CG. Both these constraints
are close to their upper limit. The first torsion frequency of the blade approaches a value
99
between 5/rev and 5.5/rev. Also, the chordwise location of SC tends to be closer to its
lower limit.
Table 3-16: Optimization Results for 3, 4 and 5/rev Actuation Frequencies
Max Max Max Max Max

Cases θ345/rev θ5/rev θ4/rev θ3/rev θstat Base
θstat (deg/m) 1.6 1.6 1.42 1.4 1.61 1.34
% Increase 19.96 19.48 5.82 4.75 20.02 -
θ345/rev 0.956 0.901 0.804 0.826 0.946 0.564
% Increase 69.49 59.82 42.65 46.45 67.78 -
Design variables
Spar End (c) 0.849 0.848 0.556 0.593 0.85 0.443
Spar Web Loc (c) 0.482 0.561 0.501 0.455 0.481 0.443
Mass 1 (m1) (kg/m) 0.165 0.01 0.109 0.098 0.075 0.222
Mass 1 Loc (x1) (c) 0.017 0.447 0.831 0.826 0.002 0.443
Mass 2 (m2) (kg/m) 0.136 0.299 0.313 0.315 0.224 0.23
Mass 2 Loc (x2) (c) 0.001 0.002 0 0.01 0.001 0.02
Constraints
st
1 Tor Freq (/rev) 5.25 5.47 4.6 4.51 5.2 6.34
M11 (kg/m) 0.71 0.718 0.703 0.711 0.71 0.684
SC (%c) 17.07 22.5 20.75 17.03 17 18.71
CG (%c) 27.96 27.9 27.47 27.97 27.05 23.94
Other parameters
S44 (Nm2) 50.51 55.43 44.32 41.16 50.47 37.71
Active Moment (Nm) 1.48 1.57 1.12 1.07 1.48 0.91
2nd Flap Freq(/rev) 2.82 2.82 2.79 2.8 2.82 2.76
During the optimization studies, it was observed that the 1st torsion frequency is the
main driving parameter for optimization in dynamic conditions. In order to understand it
further, the amplitude of dynamic twist for different actuation frequencies is plotted as a
function of the first torsion frequency (see Figure 3-16). Results shown in Figure 3-16 are
non-dimensionalized as described in Table 3-15. The amplitude of tip twist for 3/rev and
4/rev actuations is high near the torsion frequency of 5.2/rev (due to large active twisting
100
moment) and near 4.5/rev frequency (due to the location of the first torsion frequency
near the actuation frequency). In the case of 5/rev actuation frequency, torsion
frequencies ranging from 5.2/rev to 5.6/rev provide high dynamic twist. In all the cases,
the amplitude of tip twist decreases significantly as the first torsion frequency moves
away from the actuation frequency. Since dynamically optimized cases are very sensitive
to the first torsion frequency, the optimum case at one frequency may not be optimum at
a different actuation frequency.
1
345/rev
0.5
0

4.5 5 5.5 6 6.5 7 7.5 8 8.5 9

1
3/rev
0.5

0
4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
1
4/rev
0.5

0
4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
1
5/rev
0.5

0
4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
1st Torsion Freq (/rev)
Figure 3-16: Variation of Dynamic Twist Amplitude with Torsion Frequency
3.7 Effect of Advance Ratio
The optimization studies for determining the optimum active cross section at different
actuation frequencies was performed in hover condition to simplify the analysis. Also,
preliminary analysis performed showed that the amplitude of dynamic twist does not vary
significantly with forward airspeed. In order to verify this for the optimized cases, the
aeroelastic analysis of the active twist blade was performed in forward flight condition at
different actuation frequencies and the variation in the amplitude of dynamic twist was
101
determined. Note that for the results presented in this section, the “Trim Analysis” option
was selected and hence the dynamic twist obtained for μ=0 in this section might not be
the same as that obtained in the earlier sections, however, their values are close to each
other. For each of the actuation frequencies, the result obtained for the dynamically
optimized case are compared with the result obtained by maximizing the static twist and
the result obtained for the baseline case.

4/rev comp
4.5
4
Max 4/rev Case
4/rev (deg)
3.5 Max stat Case

Baseline
3
2.5
2
0 0.1 0.2 0.3
Advance ratio ()
a) Results obtained for 4/rev actuation frequency
3/rev comp
3
2.8
2.6
3/rev (deg)
Max 3/rev Case

Baseline
2.2
1.8
0 0.1 0.2 0.3
Advance ratio ()
b) Results obtained for 3/rev actuation frequency
102
5/rev comp
5
4.5
4 Max 5/rev Case
5/rev (deg)
Baseline
3
2.5
2
0 0.1 0.2 0.3
Advance Ratio ()
c) Result obtained for 5/rev actuation frequency
Figure 3-17: Effect of Advance ratio on Dynamic Twist Amplitude
The results obtained for the variation of dynamic twist amplitude with advance ratio
are shown in Figure 3-17. It can be seen that, although there is a small variation in the
twist amplitude with advance ratio, the dynamically optimized cases consistently
provides higher dynamic twist than the Max θstat case and the baseline case. Also, all the
three cases follow similar trend in the variation with advance ratio. Thus, these results
justify the original assumption that the optimization studies for maximizing the amplitude
of dynamic twist can be performed in hover condition.
This chapter presented a new strategy and the corresponding framework for the
optimum design of active twist rotor blades. The design framework integrates different
codes: IXGEN for meshing composite rotor blade cross section, UM/VABS for the
analysis of active cross sections, RCAS for rotorcraft simulation, and ModelCenter for
integration. The optimization problem in the framework is solved using a surrogate-based
approach in which the “true” objective function and constraints are replaced with
103
computationally efficient functional relationships. The surrogate-based optimization
problem is solved in combination with the Efficient Global Optimization (EGO)
algorithm, which accounts for uncertainty in surrogate predictions. To demonstrate the
capability of the framework, three different optimization problems have been considered,
namely, a) maximizing static active twist per unit length, b) maximizing amplitude of
dynamic active twist at the blade tip at a fixed actuation frequency, and c) maximizing
dynamic twist at a range of actuation frequencies. In this chapter, 3, 4 and 5/rev actuation
frequencies were considered for optimization studies since these are most effective for
reduction of 4/rev vibratory loads at the hub in fixed frame. All the studies were
conducted using the same set of design variables and constraints. The design variables
considered in this study were: the chordwise location of the main spar web, the chordwise
ending location of the spar/AFC plies, and ballast masses and their chordwise location.
Departing from the NASA/Army/MIT Active Twist Rotor design, it was found that:
1) The optimized results for the blade cross section showed:
- 18.5% increase in static twist per unit length
- 63.6% increase in tip twist amplitude for 4/rev twist actuation.
- 52.6% increase in tip twist amplitude for 3/rev twist actuation
- 101% increase in tip twist amplitude for 5/rev twist actuation
- 71% increase in twist amplitude for actuation at a range of frequencies (3, 4 and
5/rev).
2) The optimum design corresponding to maximum dynamic active twist and the one
corresponding to the maximum static active twist are different from each other. Also,
the dynamic active twist amplitude is a direct measure of control authority associated
104
with twist actuation mechanism, while the same may not hold for the static active
twist.
3) When no twist actuation is used, the dynamically optimized case may result in a small
increase in vibratory loads in forward flight conditions as compared to the baseline
case
4) The circle plots show that the optimized design for maximum dynamic active twist
provides higher control authority for reducing vibrations in all the hub load
components when compared to the other designs.
5) Optimization studies for design of cross section with maximum dynamic twist can be
performed in hover condition using periodic analysis to reduce the computational
time and to improve efficiency.
6) The optimum design obtained by maximizing dynamic twist at a range of frequencies
is better for vibration reduction as compared to designs obtained by maximizing static
twist or dynamic twist a fixed actuation frequency.
7) Based on the optimization studies conducted, important factors that can be identified
for maximizing dynamic twist are: a) first torsional frequency of the rotor blade, b)
active moment generated by active material, and c) aerodynamic loads acting on the
rotor blade.
105
Chapter 4. Mixed-Variable Optimization for Design of Active
Twist Rotor Blades
In the previous chapter, preliminary optimization for maximizing the dynamic twist
amplitude was performed with a limited number of (six) design variables and it was
demonstrated that the dynamic twist obtained from twist actuation is the true measure of
control authority for vibration reduction. Optimization approaches suitable to deal with a
larger number and different types of design variables are needed to fully explore the
active blade design space and to obtain realistic designs. In addition to the design
variables used in the previous study, the thickness and ply angle of different plies used in
the cross section also need to be considered as design variables. The plies used in the
fabrication of composite rotor blades are made up of discrete layers, each with a
prescribed thickness (pre-preg composites). Therefore, the mixed-variable optimization
needs to be performed in order to design a manufacturable rotor blade.
In the case of optimization with (m + n) mixed design variables, some (m) of the
variables are continuous while the (n) remaining ones can take discrete values only. A
typical vector of (m+n) design variables is shown below:
XDV = [xc,1 xc,2 ….. xc,n xd,1 xd,2 …… xd,m]
where xc,i (1 ≤ i ≤ n) are n continuous design variables and xd,j (1 ≤ j ≤ m) are m discrete
design variables. In the optimization problem considered here, the ply thicknesses and ply
106
angles are treated as discrete design variables while the ballast masses, the chordwise
location of vertical spar web, and the chordwise location where the active plies end are
treated as continuous design variables. The ply thickness used in this chapter is the
multiple of the nominal prepreg ply thickness and it is referred to as “normalized ply
thickness” in rest of the thesis. The basic mixed-variable optimization problem in this
chapter is solved using the genetic algorithm in MATLAB 2012’s Global Optimization
Toolbox. It is based on special creation, crossover, and mutation functions which enforce
the variables to be integers, as described in [166]. In this chapter, the genetic optimization
process is combined with the gradient based optimization to obtain an optimum design
with continuous design variables and an optimum design with mixed design variables in
an efficient manner.
4.1 Architecture for New Mixed-variable Optimization Framework
The architecture of the framework used to obtain solution for a mixed-variable
optimization problem is shown in Figure 4-1. It is a modified version of the framework
described in the earlier chapter which efficiently accounts for:
a) Discrete design variables; and
b) Increased number of design variables in the optimization problem.
All the steps involved in the new mixed-variable optimization framework are
described below.
ModelCenter Analysis: In this part, the complete aeroelastic analysis of the active twist
rotor is performed as described in Section 3.1.1. In order to reduce the computational
time for the aeroelastic analyses required for optimization studies, the “Periodic
107
Analysis” is performed instead of using the “Trim Analysis”. All the cases are run in
hover condition, as explained earlier.
Initial Test Points

(from LHS)
Iterative Loop
1. ModelCenter Analysis
(Complete Aeroelastic Analysis)
SATISFIED
Stopping Criteria Preliminary Optimization
NOT SATISFIED
2. Surrogate Model Mixed Continuous

(Kriging Interpolation) Solution 1 Optimum
Global
Optimization 4a. Constrained 4b. Sequential/ Constrained
3a. Genetic
with EGO Mixed variable Gradient Based Optimization
Optimization with
Mixed Design Variable Genetic
3b. Gradient Based
Optimization Optimization
Mixed Variable
Infill Points 1. Model- 1. Model-
Center Analysis
Center Analysis
3b. Gradient Based
Optimization with
Continuous Design Variable Mixed Mixed
Solution 2 Solution 3
Continuous Infill
Points
3c. Remove Repeated Points
Infill Samples
Figure 4-1: Augmented Optimization Framework for Continuous/Discrete Design

Variables
Stopping Criteria: The stopping criteria can either be based on the maximum number of
iterations allowed or on the difference between the optimal value of the objective
function obtained from successive iterations. In the analysis performed here, the
optimization process was stopped after 4-6 iterations. It was observed during the
108
optimization process that the difference between the successive optimal point reduced
and the accuracy of the surrogate models improved with each iteration.
Surrogate Modeling: The surrogate modeling was performed using the DACE toolbox in
MATLAB as described in Section 3.1.2. Different correlation functions available in the
toolbox were used for different variables in order to reduce the error. The error was
calculated based on the process described in Section 3.3.1.
Global Optimization with EGO Algorithm: Global Optimization with EGO algorithm
was performed in multiple steps to account for: the mixed design variables and to reduce
the computational time.
In the first step, genetic optimization is performed with mixed design variables where
some of the design variables are continuous while the remaining ones are discrete. It was
observed that the genetic algorithm works faster when some of the variables are treated as
discrete instead of the case when all the design variables are continuous. Hence, the
genetic optimization process was used to obtain optimum results with mixed design
variables only. The results obtained from this analysis are referred to as “Mixed-variable
Infill Points.” It should be noted that multiple points (a set of best 5-10 points) are
selected at the end of each optimization and not just the one optimum point. These
multiple points represent different local minima in the design space and form a part of the
Infill Samples for the next round of iteration. These “Mixed-variable Infill Points” are
also used as the starting points for the gradient based optimization performed on the
surrogate models. The gradient based optimizer provides a set of continuous optimum
points. The gradient based optimization is performed using the “fmincon” function in
109
MATLAB. The set of points obtained at the end of the continuous optimization are
referred as “Continuous Infill Points.”
The set of best points obtained from genetic optimization and gradient based
optimization are used as Infill Samples. Before transferring these points to the next stage,
repeated points are removed from the analysis by checking the absolute distance between
the design points.
Iterative Loop: The complete aeroelastic analysis is performed again at the Infill Sample
points using the ModelCenter environment. The results obtained from new points are
used to update the surrogate models for all the constraints and the objective function. The
process of global optimization with genetic algorithm and gradient based optimization is
performed again. The iterative loop is repeated multiple times depending upon the
stopping criteria and each iteration is referred to as “SBO Iteration.”
Preliminary results: At the end of the iterative loop, the set of points which satisfy all the
constraints are sorted in the order of increasing objective function. The best point
obtained is referred as “Continuous Optimum” and it represents the best design point
with continuous design variables. Next, the points where the ply thicknesses (or ply
angles) have discrete values are sorted out of the group. The point with the best objective
function in this group is referred to as “Mixed Solution 1.” This point is the most
optimum solution obtained at the end of the iterative loop when the discrete design
variables have integer values only.
The mixed-variable solution can also be obtained in two other different ways using the
“Continuous Optimum” point obtained earlier. In the first method, the genetic
110
optimization for mixed design variable is used, while the second method involves the
usage of a gradient based method. These two methods are described in detail here.
a) Constrained Mixed-variable Genetic Optimization
This optimization is similar to what was performed in “Optimization with EGO
algorithm”, except that the bounds for discrete design variables are modified such
that a discrete solution is determined near the “Continuous Optimum” point. For
example, if the “Continuous Optimum” point gives a value of 1.36 for the
normalized ply thickness, then a lower bound of “1” and an upper bound of “2”
are used for this normalized ply thickness in the genetic optimization. The bounds
for a continuous design variable are kept unchanged during this process. A sample
case is shown in Table 4-1 where the optimization is performed with 12 design
variables. Of these 12 design variables, four are continuous while the remaining
eight can take discrete values only. The initial upper and lower bound for these
design variables (as used in “Optimization with EGO algorithm”) are shown by
the rows corresponding to Xupper and Xlower , respectively. The “Continuous
Optimum” solution obtained at the end of Preliminary Optimization is shown by
Xopt. Based on the optimum result obtained, the upper and lower bounds on the
design variables are modified to X’upper and X’lower, respectively. Note that in this
step, only the bounds for discrete design variables are modified while the bounds
on continuous design variables remain unchanged. The mixed-variable solution
obtained at the end of this optimization process is referred as “Mixed Solution 2.”
111
Table 4-1: Modified Bounds for Constrained Mixed-variable Genetic Optimization
xc1 xc2 xc3 xc4 xd1 xd2 xd3 xd4 xd5 xd6 xd7 xd8
Bounds for Original Mixed-variable Genetic Optimization
Xupper 0.85 0.85 0.5 0.5 5 5 5 5 5 5 5 5
X lower 0.2 0.2 0 0 0 0 0 0 0 0 0 0
Solution Obtained from Continuous Gradient-Based Optimization
X opt 0.85 0.84 0.29 0.012 0.10 4.93 1.16 0.10 1.31 0.10 0.54 0.64
Modified Bounds for Constrained Mixed-variable Optimization
X’upper 0.85 0.85 0.5 0.5 1 5 2 1 2 1 1 1
X’ lower 0.2 0.2 0 0 0 4 1 0 1 0 0 0
b) Sequential Constrained Gradient-based optimization
Another approach for obtaining a mixed-variable solution using the “Continuous
Optimum” design is the classical sequential optimization approach which can be
performed using a gradient based optimizer. In this approach, if any of the
discrete design variables in X opt has a value close to an integer, then the value for
that particular design variable is fixed to that integer value and it is not considered
a design variable anymore. For example, in the results shown in Table 4-1, the
value corresponding to xd2 is 4.93 in X opt. Since this value is very close “5”, the
value for this design variable is fixed to “5” and it is not considered a design
variable. Similarly, the value of design variables xd1, xd4 and xd6 is fixed to “0”.
The modified vector of design variables and their upper and lower bounds for the
next gradient-based optimization study are shown in Table 4-2. In the next step,
the value of one more discrete design variable is fixed to an integer value and the
process repeated till all the discrete design variables have been assigned an integer
value. In this particular case, the gradient based optimization had to be performed
112
four more times in order to get the final mixed-variable solution. The solution
obtained at the end of this method is referred as “Mixed Solution 3.”
Table 4-2: Modified Set of Design Variables for Sequential Gradient-Based

Optimization
xc1 xc2 xc3 xc4 xd3 xd5 xd7 xd8

Xupper 0.85 0.85 0.5 0.5 5 5 5 5
Xlower 0.2 0.2 0 0 0 0 0 0
This optimization technique has the advantage that it provides a mixed-variable
solution using a gradient-based optimizer. However, the optimization needs to be
performed multiples times depending upon the number of discrete design variable
in the problem. Every time, the time to convergence decreases as the size of the
problem decreases and the starting condition are very close to the optimum.
Hence, the Sequential Gradient Based Optimization approach may be time-
consuming.
Optimization Parameters: The optimization parameters used for the GA optimization
performed in steps 3a and 4a using the MATLAB’s Global Optimization Toolbox are
listed in Table 4-3. In most of the optimizations performed using GA, it was observed
that the process stopped after exceeding the limit on the maximum number of generations
allowed. Similarly, the optimization parameters used for GBO performed using the
fmincon function in MATLAB in steps 3b and 4b are listed in Table 4-4. It should be
noted that in the framework presented here (used in Chapter 4 and Chapter 5), both GA
and GBO optimizations are performed on the surrogate models for objective function and
constraints and thus, it is possible to perform a large number of iterations.
113
Table 4-3: Parameters used in GA Optimization
Parameter Value
Function Tolerance (def) 1.0e-06
Population Size 400
EliteCount 20
Generations 150
Crossover Fraction 0.6
Table 4-4: Optimization Parameters for GBO
Parameter Value
Tolerance (def) 1.0e-08
Maximum Fun Eval 10000
Maximum Iterations 500
4.2 Optimization with Normalized Ply Thickness
The baseline rotor blade used for the optimization studies in this chapter is the same
NASA/Army/MIT Active Twist Rotor (ATR), as described in Chapter 3. The cross
section shape and composite layup, planform of the rotor blade and rotor characteristics
are all described in Chapter 3. In the first study, the normalized ply thicknesses of
different plies used in the cross section are considered as design variables, along with the
variables described in Chapter 3. In order to make the rotor blade design more realistic,
the location of first ballast mass is fixed near the leading edge at x = 0.02c while the
second ballast mass is located just in front of the vertical spar web. (This is done to
ensure that the ballast mass is added in the region where passive plies can be used to
support it and thus prevent the ballast mass from flying out during the operation). Due to
these changes, there were small changes in the dynamic properties of the baseline case.
The set of design variables and their upper and lower bounds are given in Table 4-5. In
order to prevent the mesh generator from crashing, the lower bound on normalized ply
114
thickness is fixed at 0.1 instead of zero. A value of 0.1 for normalized ply thickness in an
optimum design implies that that particular ply is not required in the cross section and
should be removed in the next optimization. The constraints used in the optimization are
the same as those used in Chapter 3, except the lower bound on first torsional frequency.
As listed in Table 4-6, the lower bound for first torsional frequency was lowered to 3/rev
instead of 4.5/rev used earlier.
Table 4-5: Design Variables and their Bounds
Design variables Baseline Lower Upper Ply Type

1 Spar Web Loc (c) 0.443 0.2 0.85
2 Spar End (c) 0.443 0.2 0.85
3 Ballast Mass 1 (m1) (kg/m) 0.23 0 0.5
5 Nor Thickness of Ply 1 1 0.1 5 E-Glass
6 Nor Thickness of Ply 2a 1 0.1 5 S-Glass
7 Nor Thickness of Ply 2 1 0.1 5 AFC
9 Nor Thickness of Ply 4 1 0.1 5 AFC
Table 4-6: Constraints for Optimization Problem
Constraints Min Max

SC (%c) 17 25
CG (%c) 20 28
M11 (kg/m) 0.65 0.72
st
1 Tor Freq (/rev) 3.0 7
Objective functions which are considered for optimization studies are listed below:
1) Maximize static twist per unit length (Max θstat)
2) Maximize amplitude of twist for 3/rev actuation (Max θ3/rev)
3) Maximize amplitude of twist for 4/rev actuation (Max θ4/rev)
115
4) Maximize amplitude of twist for 5/rev actuation (Max θ5/rev )
5) Maximize amplitude of twist at 3,4 and 5/rev actuation simultaneously (Max
θ345/rev)
. For all the active twist optimization studies presented in this chapter, the amplitude
of actuation voltage was fixed at 1000V. Final results obtained for the objective functions
at the end of optimization are shown in Table 4-7. The results show the optimum value of
objective function when all the design variables are treated as continuous and when the
normalized ply thicknesses are treated as discrete (obtained from all the three mixed-
variable optimization techniques described in Figure 4-1).
Table 4-7: Final Result obtained from Optimization Studies
Max θstat Max θ3/rev Max θ4/rev Max θ5/rev Max θ345/rev
(deg/m) (deg) (deg) (deg)
Continuous Optimum 2.59 5.69 6.56 7.97 0.89
Mixed Solution 1 2.56 4.24 5.88 7.79 0.87
Mixed Solution 2 2.55 4.19 6.01 7.93 0.89
Mixed Solution 3 2.55 4.18 5.85 7.98 0.89
Baseline 1.34 1.85 2.06 2.34 0.31
The results show that the value of objective function corresponding to optimization
with continuous design variables is always better than those obtained for the cases with
mixed design variables. In general, the results obtained from the three mixed-variable
optimization techniques are close to each other. The most interesting aspect of these
results is the difference between the value of the objective function when all the variables
are treated as continuous and when the variables are of mixed type. The percentage
difference between the value of objective function for the continuous variable case and
the average value of objective function for the mixed-variable cases is shown in Table
116
4-8. The percentage difference is less than 1.5% for Max θstat, Max θ5/rev and Max θ345/rev
cases, while it is highest for the Max θ3/rev case.
Table 4-8: Percentage Difference between the Objective Function for Continuous
Variable Optimization and Mixed-variable Optimization
Difference (%) 1.42 26.13 9.86 0.88 1.20
4.2.1 Optimization with Continuous Design Variables
The value of design variables and constraints for the optimization cases with
continuous design variables and for the baseline case is shown in Table 4-9. As observed
earlier, the most critical parameter for maximizing the dynamic twist is the first torsion
frequency of the blade. The optimizer tries to bring the first torsion frequency of the
blade closer to the actuation frequency. The chordwise location of CG for all the cases is
closer to the aft constraint limit on CG location. This can be attributed to the increase in
the value of design variable “Spar End” which is at its upper limit. By increasing the
chordwise coverage of the spar/active plies, higher active twisting moment can be
obtained, which would also result in an increase in the dynamic twist. The chordwise
location of the vertical spar web is very close to the “Spar End” value for all the
optimized cases. This results in a box-type spar for all the optimized cases. The increase
in the chordwise coverage of plies in the cross section leads to an increase in the torsional
stiffness. For all the optimized cases (except the Max θ3/rev case), the torsional stiffness of
the optimum blade is higher than that for the baseline case, even though the first torsion
frequency is lower. The placement of the first torsion frequency for the optimized cases is
117
controlled by manipulating the values of two ballast masses. The amount of ballast mass
used in the cross section is highest for the Max θ3/rev case and it is least for the Max θstat
case. Thus, the two ballast masses play an important role in varying the first torsional
frequency of the blade.
Table 4-9: Constraints and Design Variables for Optimization with Continuous
Design Variables
Baseline Max θstat Max θ3/rev Max θ4/rev Max θ5/rev Max θ345/rev
Constraints
1st Tor Freq (/rev) 6.53 5.9 3.71 4.86 5.6 5.09
M11 (kg/m) 0.682 0.701 0.7 0.719 0.717 0.719
SC (%c) 18.71 23.5 17.07 19.12 24.82 18.98
CG (%c) 21.64 27.22 26.46 27.92 27.41 27.8
Continuous Variables
Spar End (c ) 0.443 0.85 0.818 0.85 0.842 0.85
Spar Web (c ) 0.443 0.84 0.813 0.85 0.834 0.85
m1 (kg/m) 0.23 0.299 0.397 0.346 0.32 0.334
m2 (kg/m) 0.22 0.012 0.123 0.06 0.034 0.047
Discrete Variables (Normalized Ply Thickness)
Ply 1 1 0.1 0.16 0.1 0.11 0.1
Ply 2a 1 4.93 2.40 5 4.84 5
Ply 2 (AFC) 1 1.16 0.41 0.84 1.03 0.98
Ply 3 1 0.1 0.10 0.1 0.10 0.1
Ply 4 (AFC) 1 1.31 0.68 1.09 1.27 1.12
Ply 5 1 0.1 0.17 0.1 0.1 0.1
Ply 6 1 0.59 0.26 0.43 0.54 0.59
Ply 7 1 0.64 0.10 0.25 0.82 0.1
Other Parameters
S44 (Nm2) 37.7 62.4 28.8 49.2 60.1 52.3
Act Mom (Nm) 0.91 2.83 1.16 2.21 2.70 2.37
nd
2 Flap Freq (/rev) 2.76 2.75 2.67 2.72 2.74 2.73
rd
3 Flap Freq (/rev) 5.26 5.33 4.80 5.14 5.26 5.19
118
Among the ply thickness design variables, the normalized ply thickness of all passive
plies (Ply 1, Ply 3 and Ply 5) in the spar region have been reduced to their minimum
allowable value. This was expected since they do not contribute to the active twist.
However, the nose ply (Ply 2a) is very important for obtaining higher active twisting
moment and hence all the optimized cases show an increase in the normalized thickness
of nose ply. The plies in the vertical spar web (Ply 6 and Ply 7) need to have sufficient
thickness in order to control the chordwise location of the shear center. Hence, even
though these are passive plies, the normalized ply thickness for the spar web plies is not
close to zero.
An increase in the normalized thickness of active plies is also accompanied by a
corresponding increase in the torsional stiffness for the cross section. Hence, the
optimized cases have different values for the normalized thickness of active plies (Ply 2
and Ply 4), depending upon the actuation frequency. The thickness of active plies is
highest for the Max θstat case while it is the least for the Max θ3/rev case. The results
obtained for normalized ply thickness also demonstrate that, for the fixed amount of
active material available, it is preferable to increase the chordwise coverage of active
material as compared to increasing the thickness of active plies in order to get a higher
dynamic twist amplitude. Another important trend observed is the direct correlation
between torsional stiffness (GJ) of the cross section and the active twisting moment
generated by the embedded active plies. For all the optimized cases, the normalized
thickness of the inner active ply (Ply 4) is higher than that of the outer active ply (Ply 2).
The convergence of the optimum results obtained with continuous design variables is
shown in Figure 4-2. The X-axis in the plot represents the number of Surrogate Based
119
Optimization (SBO) iterations, as described in Section 4.1. Results show that for some of
the cases, the optimized result is obtained in the first 1-2 iterations. The variation of
constraints and design variables for the Max θ4/rev case with SBO iterations is shown in
Figure 4-3 and Figure 4-4, respectively. Results presented here show that a local
optimum results is obtained in the 1st iteration where the first torsion frequency of the
blade is very close to the actuation frequency of 4/rev. For this case, the chordwise
location of vertical spar web and spar end are close to each other and near the maximum
value allowed for these two design variables. Also, the normalized thickness of both the
active plies is less than “1”. Thus, the optimizer is trying to tune the first torsion
frequency to obtain higher amplitude. However, the best result obtained in Iteration 5
shows an increase in the thickness of active plies and a corresponding increase in
torsional frequency. Thus, the best case tries to maximize the active twisting moment
generated at the cost of higher torsional frequency. The increase in cross-sectional mass
due to the increase in thickness of active plies is balanced by reducing the ballast masses
used. This also shifts the CG of the cross section closer to its upper bound.
3 6
stat (deg/m)
3/rev (deg)
4
2
2
1 0
0 2 4 6 0 2 4 6
SBO Iteration SBO Iteration
8 8
4/rev (deg)
5/rev (deg)
6 6
4 4
2 2
0 2 4 6 0 2 4 6
Figure 4-2: Variation of objective function with Iteration number for optimization
with Continuous Design Variables
120
1st Tor Freq (/rev)
7 0.75
M11 (kg/m)
6
0.7
5
4 0.65
3
0 2 4 6 0 2 4 6
26
28
24
CG (%c)
SC (%c)
26
22
24
20
22
18
20
0 2 4 6 0 2 4 6
Figure 4-3: Variation of Constraints for Max θ4/rev Optimization
1 1
Spar Web(c)
Spar End(c)
0.5 0.5
0 0
0 2 4 6 0 2 4 6
1 1
m1(kg/m)
m2(kg/m)
0.5 0.5
0 0
0 2 4 6 0 2 4 6
Design Variables (A)

5
Ply 1
Ply 2a
4 Ply 2
Normalized Thickness (#)
Ply 3
Ply 4
3 Ply 5
Ply 6
Ply 7
2
0
0 1 2 3 4 5 6
SBO Iteration
Design Variables (B)
Figure 4-4: Variation of Design Variables for Max θ4/rev Optimization
121
The performance of the optimized cases at different actuation frequencies is shown in
Table 4-10. Each column represents one of the optimized cases as listed in Table 4-9. The
tip-twist values listed in Table 4-10 are non-dimensionalized by the maximum value
obtained for that objective function during the optimization study (except for 345/rev ). The
results show that the value of static twist is very close to the maximum value that can be
obtained for Max θ4/rev, Max θ5/rev and Max θ345/rev cases. This table also highlights that
the optimum solution obtained at one actuation frequency may not be optimum at a
different actuation frequency, and hence the optimization needs to be performed at a
range of actuation frequencies. The solution obtained by maximizing 345/rev shows high
values of dynamic twist for all the actuation frequencies considered.
Table 4-10: Performance of Optimized Cases at other Actuation Frequencies
Cases Baseline Max θstat Max θ3/rev Max θ4/rev Max θ5/rev Max θ345/rev
 stat 0.52 1 0.87 0.98 0.99 0.99
 3/rev 0.33 0.68 1 0.81 0.69 0.77
 4/rev 0.31 0.73 0.55 1 0.78 0.93
 5/rev 0.29 0.91 0.16 0.81 1 0.97
345/rev 0.31 0.77 0.57 0.87 0.82 0.89
 stat 3/ rev  4/ rev 5/ rev

where,  stat  ,  3/ rev  ,  4/ rev  and  5/ rev  ;
 stat ,max 3/ rev ,max  4/ rev ,max 5/ rev ,max
and θstat,max = 2.59, θ3/rev,max = 5.69, θ4/rev,max = 6.56, and θ5/rev,max = 7.97
4.2.2 Optimization with Mixed Design Variables
In this section, the results obtained from the optimization with continuous design
variables are compared with those obtained using mixed-variable for each of the
objective function described above. As discussed earlier, in the case of mixed design
122
variable optimization, four of the twelve design variables are as continuous while the
remaining eight are discrete and can take integer values only. In this case also, the lower
bound on the normalized ply thickness was fixed at 0.1 instead of zero to prevent the
mesh generator from crashing.
4.2.2.1 Maximizing θstat
The final results obtained from maximizing θstat using the optimization process
described in Section 4.1 are shown in Table 4-11. For this objective function, the
difference in the value of objective function between the optimization with continuous
design variables and the optimization with mixed design variables is less than 1.5%.
Although, the final values of the objective function for the optimized cases are close,
there is a noticeable difference between the optimum designs. Also, the difference
between the results obtained from the three different techniques used for optimization
with mixed design variables is small.
The biggest difference between the continuous variable and mixed-variable
optimization lies in the value of first torsion frequency for the optimized cases. In the
continuous variable case, the active plies, Ply 2 and Ply 4, have thickness 16% and 30%
higher than those for the mixed-variable case, respectively. Due to this, the optimum
design with mixed design variables has less torsional stiffness and the embedded active
plies generate less active twisting moment. This also highlights that multiples local
minima exist in the design space being considered. The mixed-variable cases also show
an increase in the magnitude of leading edge ballast mass and a corresponding increase in
the mass per unit length for the cross section.
123
Table 4-11: Optimization Results for Maximizing θstat
Objective Function Continuous Mixed Mixed Mixed

Optimum Solution 1 Solution 2 Solution 3
θstat (deg/m) 2.59 2.56 2.57 2.57
Constraints
Tor Freq (/rev) 5.90 5.14 5.33 5.11
M11 (kg/m) 0.701 0.712 0.720 0.719
SC (%c) 23.50 24.52 23.43 23.43
CG (%c) 27.22 27.28 20.55 27.72
Spar End (c ) 0.850 0.847 0.850 0.850
Spar Web (c ) 0.840 0.844 0.850 0.850
m1 (kg/m) 0.299 0.342 0.361 0.342
m2 (kg/m) 0.012 0.052 0.035 0.054
Ply 1 0.10 0.1 0.1 0.1
Ply 2a 4.93 4 5 5
Ply 2 (AFC) 1.16 1 1 1
Ply 3 0.10 0.1 0.1 0.1
Ply 4 (AFC) 1.31 1 1 1
Ply 5 0.10 0.1 0.1 0.1
Ply 6 0.54 0.1 1 0.1
Ply 7 0.64 1 0.1 1
4.2.2.2 Maximizing θ3/rev
The results obtained by maximizing θ3/rev using continuous and mixed design variables
are shown in Table 4-12. For this case, while the three results obtained with mixed
design variables are close to each other, there is a 26% difference between the optimum
values of objective function as compared to the continuous variable case. The main
reason for this is the discretization of normalized thickness for the active plies. In order to
reduce the torsional frequency (and torsional stiffness) of the blade, the normalized
thickness of active plies in the cross section is well below “one” for the continuous
124
design variable case. But when the normalized thickness of active plies is rounded to
“one” for the mixed-variable cases, there is a significant increase in the torsional stiffness
of the cross section which could not be completely offset by adding more ballast masses.
As a result, all the cases with mixed design variables show a higher torsional frequency
and thus lower amplitude for the dynamic twist.
Table 4-12: Optimization Results for Maximizing θ3/rev

θ3/rev (deg) 5.69 4.24 4.19 4.18
Constraints
st
1 Tor Freq (/rev) 3.71 5.15 4.81 5.11
M11 (kg/m) 0.700 0.694 0.683 0.718
SC (%c) 17.07 19.91 17.27 23.74
CG (%c) 26.46 26.93 27.35 27.61
Spar End (c ) 0.818 0.845 0.828 0.816
Spar Web (c ) 0.813 0.846 0.599 0.816
m1 (kg/m) 0.397 0.326 0.309 0.343
m2 (kg/m) 0.123 0.046 0.072 0.068
Ply 1 0.16 0.1 0.1 0.1
Ply 2a 2.40 5 2 4
Ply 2 (AFC) 0.41 1 1 1
Ply 3 0.10 0.1 0.1 0.1
Ply 4 (AFC) 0.68 1 1 1
Ply 5 0.17 0.1 0.1 0.1
Ply 6 0.26 1 0.1 0.1
Ply 7 0.10 0.1 1 1
Small differences can be observed among the three results obtained with mixed design
variables. In the “Mixed Solution 1”, five plies are used in the nose region which gives
higher active twisting moment. Thus, the “Mixed Solution 1” provides the maximum
125
dynamic twist amplitude inspite of having the highest torsional frequency. In the case of
“Mixed Solution 2”, the vertical spar web is located near mid chord and the first torsional
frequency is closer to the actuation frequency of 3/rev. Thus, the optimizer is trying to
increase the amplitude of dynamic twist by reducing the first torsion frequency.
The optimization results obtained by maximizing θ4/rev with continuous and mixed
design variables are shown in Table 4-13.

θ4/rev (deg) 6.56 5.88 6.01 5.85
Constraints
st
1 Tor Freq (/rev) 4.86 5.13 5.11 5.10
M11 (kg/m) 0.719 0.714 0.719 0.719
SC (%c) 19.12 24.50 23.43 23.80
CG (%c) 27.92 27.54 27.79 27.81
Spar End (c ) 0.850 0.849 0.850 0.819
Main web (c ) 0.850 0.842 0.850 0.819
m1 (kg/m) 0.346 0.342 0.341 0.343
m2 (kg/m) 0.060 0.053 0.054 0.068
Ply 1 0.10 0.10 0.10 0.10
Ply 2a 5.00 4 5 4
Ply 2 (AFC) 0.84 1 1 1
Ply 3 0.10 0.10 0.10 0.10
Ply 4 (AFC) 1.09 1 1 1
Ply 5 0.10 0.10 0.10 0.10
Ply 6 0.43 1 1 0.10
Ply 7 0.25 0.10 0.10 1
126
In this case, the difference between the optimum value of the objective function
obtained using continuous design variables and mixed design variables is 9.9%. Here, the
normalized thickness of active plies is closer to their discrete value, than they were in the
case of “Maximizing θ3/rev.” Besides the thickness of active plies and vertical spar web
plies, there is very small difference in the optimum design obtained with continuous
design variables and mixed design variables. Among the different results with mixed
design variables, the “Mixed Solution 2” gives the best result since it has more plies in
the nose region that result in a higher active twisting moment.
The results obtained for maximizing the amplitude of dynamic twist at 5/rev actuation
frequency with continuous and mixed design variables are shown in Table 4-14. Unlike
the results obtained for “Maximizing θ3/rev” and “Maximizing θ4/rev” cases, the difference
between the optimum value of the objective function obtained using continuous design
variables and mixed design variables is very small. In the optimization with continuous
design variable, the normalized thickness of active plies is more than 1 in order to obtain
higher active twisting moment. However, in the case of mixed design variables, the
dynamic twist is maximized by placing the first torsion frequency closer to the actuation
frequency. Also, the results with mixed design variables show slightly heavier ballast
mass in the spar region to increase the torsional inertia for the cross section and to further
reduce the torsion frequency as compared to the continuous design variable case.
127
Table 4-14 Optimization Results for Maximizing θ5/rev

θ5/rev (deg) 7.97 7.79 7.93 7.98
Constraints
st
1 Tor Freq (/rev) 5.60 5.20 5.10 5.22
M11 (kg/m) 0.717 0.704 0.720 0.698
SC (%c) 24.8 23.2 23.5 23.4
CG (%c) 27.4 27.4 28.0 27.9
Spar End (c ) 0.842 0.850 0.850 0.848
Spar Web (c ) 0.834 0.833 0.850 0.841
m1 (kg/m) 0.320 0.331 0.342 0.326
m2 (kg/m) 0.034 0.050 0.055 0.050
Ply 1 0.11 0.1 0.1 0.1
Ply 2a 4.84 5 5 5
Ply 2 (AFC) 1.03 1 1 1
Ply 3 0.10 0.1 0.1 0.1
Ply 4 (AFC) 1.27 1 1 1
Ply 5 0.10 0.1 0.1 0.1
Ply 6 0.54 0.1 0.1 0.1
Ply 7 0.82 1 1 1
For this particular optimization study, the optimum design shown in “Mixed Solution
3” is slightly better than that obtained for Continuous Optimum, which contrary to the
expected trend. This implies that the result obtained with continuous design variables is
not optimum solution and it should be possible to find a better solution. However, the
difference between optimum values predicted by “Continuous Optimum” and “Mixed
Solution 3” is very small and is within the error in the prediction of θ5/rev by the surrogate
model.
128

θ345/rev 0.89 0.87 0.88 0.89
Constraints
st
1 Tor Freq (/rev) 5.09 5.13 5.08 5.11
M11 (kg/m) 0.719 0.714 0.719 0.719
SC (%c) 18.98 24.50 24.51 23.43
CG (%c) 27.80 27.54 27.99 27.82
Spar End (c ) 0.850 0.849 0.848 0.850
Spar Web (c ) 0.850 0.842 0.848 0.850
m1 (kg/m) 0.334 0.342 0.346 0.341
m2 (kg/m) 0.047 0.053 0.056 0.054
Ply 1 0.10 0.1 0.1 0.1
Ply 2a 5.00 4 4 5
Ply 2 (AFC) 0.98 1 1 1
Ply 3 0.10 0.1 0.1 0.1
Ply 4 (AFC) 1.12 1 1 1
Ply 5 0.10 0.1 0.1 0.1
Ply 6 0.59 1 0.1 0.1
Ply 7 0.10 0.1 1 1
Finally, the result obtained for maximizing θ345/rev using continuous and mixed design
variables are shown in Table 4-15. In this case also, the optimum result obtained from
optimization with mixed design variables is very close to that obtained using continuous
design variables.
In this section, the optimization studies were conducted with twelve design variables,
where four of the design variables were continuous while the remaining eight were
discrete. Here, the optimum solution was obtained using continuous design variables and
129
mixed design variables in order to compare the two results and obtain a feasible design,
which can be readily manufactured. The results showed that in some cases it is possible
to get results with mixed design variables very close to those obtained with continuous
design variables, depending upon the objective function.
4.3 Optimization Study with 8 Design Variables
In order to prevent the mesh generator from crashing, the minimum allowable
normalized ply thickness was fixed to “0.1” instead of using “0” for the optimization
studies performed in Section 4.2. The optimization results obtained showed that the
optimizer tried to reduce the normalized thickness of all passive plies in the spar region
(Ply 1, Ply 3 and Ply 5) to 0.1 indicating that these plies do not contribute to the dynamic
twist amplitude and hence, these plies should be removed from the cross section.
Therefore, in the optimization study presented in this section, Ply 3 and Ply 5 are
removed from the analysis. Since Ply 1 is the outermost ply, it cannot be removed from
the cross section. Thus, the normalized thickness of Ply 1 is fixed to minimum possible
thickness, which is “one”, for all the studies presented in this section. Also, in order to
reduce the number of design variables, both the plies in the vertical spar web region (Ply
6 and Ply 7) are grouped and it is treated as one equivalent ply (Ply 6) whose thickness is
a design variable. The modified cross section which is used as the baseline case is shown
in Figure 4-5 and is referred to as “Baseline 2” in rest of the thesis. The final set of design
variables used in this study and their upper and lower bounds are listed in Table 4-16.
The constraints used in this study are the same as those listed in Table 4-6.
130
Figure 4-5: Modified Baseline Case (Baseline 2)
Table 4-16: Design Variable and their Bounds

1 Main Spar Web Loc (c) 0.443 0.2 0.85
2 Spar End (c) 0.443 0.2 0.85
Normalized Ply Thickness
5 Ply 2a Thickness 1 1 5 S-Glass
6 Ply 2 Thickness 1 1 5 AFC
8 Spar Web Ply 6 Thickness 1 1 10 E-Glass
The summary of the results obtained for objective function at the end of optimization
with continuous design variables and mixed design variables is shown in Table 4-17.
Comparing the results obtained for the optimized cases with those obtained earlier in
Table 4-7 and Table 4-9, it can be seen that the final value of objective function is smaller
in this optimization study, for all the objective functions considered. The main reason for
this is the fact that the normalized thickness of the outermost passive ply (Ply 1) in the
cross section is fixed to “1” whereas, in the previous case, the optimizer had the freedom
to reduce the normalized thickness of this passive ply to the minimum allowable value
131
which was fixed at “0.1.” Also, the percentage difference in the optimum result obtained
with continuous design variables and optimum results obtained with mixed design
variables is different in this optimization (as compared to the percentage differences
observed in Table 4-8).
Table 4-17: Results obtained for Optimization with 8 Design Variables
(deg/m) (deg) (deg) (deg)
Continuous Optimum 2.49 4.9 5.94 7.77 0.894
Best Mixed Solution 2.41 4.6 5.41 7.63 0.889
% Difference 3.21 6.12 8.92 1.80 0.56
4.3.1 Optimization Results with Continuous Design Variables
The results obtained for all the cases with continuous design variables using the
framework described in Section 4.1 are shown in Table 4-18. For the Max θstat case, three
of the constraints namely, mass per unit length and chordwise location of CG and SC are
close to their upper bound. This occurs, because there is an increase in the thickness of all
the plies used in the cross section and the chordwise coverage of active plies is at the
maximum allowable value. Here, only the leading-edge ballast mass is used to get the
chordwise location of CG within the bounds required. Among the ply thicknesses, the
thickness of the nose ply is very close to the maximum allowable value since it results in
a higher active twisting moment. There is an increase of 23% in the normalized thickness
of active plies, namely, Ply 2 and Ply 4. Among all the optimized cases, the Max θ stat case
has the highest cross-sectional stiffness.
132
Table 4-18: Results Obtained with Continuous Design Variables
Max Max Max Max Max

Cases θstat θ3/rev θ4/rev θ5/rev θ345/rev Baseline 2
Objective Function
θstat (deg/m) 2.49 2.25 2.24 2.42 2.41 1.67
θ3/rev (deg) 3.67 4.9 4.66 3.86 3.91 2.4
θ4/rev (deg) 4.48 5.92 5.94 5.21 5.34 2.54
θ5/rev (deg) 6.81 3.6 4.31 7.77 7.67 2.38
θ345/rev 0.794 0.82 0.835 0.888 0.894 0.408
Constraints
1st Tor Freq (/rev) 6.01 4.53 4.69 5.37 5.29 5.48
M11 (kg/m) 0.72 0.719 0.719 0.72 0.72 0.642
SC (%c) 24.8 17.17 17.14 24.95 23.93 19.01
CG (%c) 27.75 27.79 27.75 28 28 21.87
Design Variables
Spar End (c) 0.85 0.85 0.85 0.85 0.85 0.443
Spar Web (c) 0.85 0.309 0.319 0.85 0.85 0.443
m1 (kg/m) 0.299 0.309 0.307 0.328 0.33 0.23
m2 (kg/m) 0 0 0 0.033 0.036 0.22
Ply 2a (E-Glass) 4.88 1 1.54 4.75 4.85 1
Ply 2 (AFC) 1.23 1.37 1 1.04 1 1
Ply 4 (AFC) 1.24 1 1.36 1 1 1
Ply 6 (E-Glass) 1.2 10 10 1.15 1 1
Other Parameters
S44 (Nm2) 66.2 35 37.9 57.9 56.4 25.1
Act Mom (Nm) 2.87 1.5 1.61 2.45 2.38 0.76
nd
2 Flap Freq(/rev) 2.79 2.79 2.79 2.76 2.76 2.72
3rd Flap Freq (/rev) 5.51 5.49 5.51 9.42 5.37 5.07
Max ε11 (µε) 2478 3872 3547 2658 2646 3935
Max ε12 (µε) 4347 5482 4968 4288 4816 5591
In the Max θ3/rev case, the presence of the outermost passive ply does not permit
significant reduction in the first torsional frequency, as it was possible in the previous
optimization study presented in Section 4.2. Thus, the first torsion frequency and cross-
sectional torsional stiffness obtained for the Max θ3/rev case in Table 4-18 is higher than
133
that obtained for the Max θ3/rev case in Table 4-9. As a consequence of this, the value of
objective function for the optimized case in Table 4-18 is significantly lower than that
obtained in Table 4-9. The vertical spar web is located near the quarter chord due to
which the chordwise location of shear center is closer to its lower limit. Unlike the Max
θstat case, the normalized thickness of the nose ply, Ply 2a, is at its minimum value while
the normalized thickness of vertical spar web ply, Ply 6, is at the maximum allowable
value. Thus, in the Max θ3/rev case, the optimizer is trying to lower the torsional stiffness
as much as possible in order to get the first torsion frequency closer to the actuation
frequency.
The results obtained for Max θ4/rev and Max θ3/rev cases are very close to each other.
This is specific to this problem and it can be attributed to the bounds used for constraints
and design variables in the optimization problem definition. The only noticeable
difference between the Max θ4/rev case and Max θ3/rev case is in the thickness of active
plies.
As observed in Section 4.2, the result for Max θ5/rev case is similar to the result
obtained for Max θstat case since their first torsion frequencies are close to each other. In
this case, the second ballast mass is also used to tune the first torsion frequency of the
blade. The total ballast mass used in the Max θ5/rev case is higher than that used in the
cases discussed above. Thus, for the Max θ5/rev case, the optimizer takes advantage of
both, the higher active twisting moment and dynamic tuning, to obtain large amplitude of
oscillation at the blade tip. The result obtained for the Max θ345/rev case is close to the
result for Max θ5/rev case, but with a slightly lower first torsion frequency to improve the
amplitude of twist oscillation for all the actuation frequencies.
134
In the optimization studies presented in this section, the cross-sectional strains are not
included as part of the constraints. The results obtained here show that the maximum
value of ε11 and ε12 in the cross section for all the optimized cases is approximately equal
to or less than that obtained for the baseline case. Thus, the blade designs obtained from
these optimization studies have sufficient strength to withstand the large centrifugal
loads.
4.3.2 Optimization Results with Mixed Design Variables
For the results presented in this section, the normalized ply thicknesses are treated as
discrete design variables. In the previous section, it was shown that the optimization with
mixed design variables can be performed in three different ways. The mixed solutions,
“Mixed Solution 1” and “Mixed Solution 2”, are obtained using the genetic mixed-
variable optimization while the “Mixed Solution 3” is obtained using a gradient based
optimizer only. The results obtained in Section 4.2 showed that the final results obtained
for the objective function with different mixed design variables optimization techniques
are close to each other. Also, it was observed that obtaining the “Mixed Solution 3”
required significant computational time since the optimization is performed in a recursive
manner. Hence, in this section, the mixed-variable optimization is performed to obtain
“Mixed Solution 1” and “Mixed Solution 2” only. The final results presented here in
Table 4-19 show only the best result obtained with mixed design variables.
135
Table 4-19: Optimization Results with Mixed Design Variables
Max Max Max Max Max

Objective Function
θstat (deg/m) 2.41 2.22 2.21 2.41 2.39 1.67
θ3/rev (deg) 3.36 4.6 4.34 3.89 3.88 2.4
θ4/rev (deg) 4.3 5.28 5.41 5.3 5.31 2.54
θ5/rev (deg) 5.73 3.86 4.81 7.63 7.62 2.38
θ345/rev 0.716 0.775 0.806 0.889 0.889 0.408
Constraints
1st Tor Freq (/rev) 5.71 4.71 4.93 5.31 5.28 5.48
M11 (kg/m) 0.694 0.688 0.697 0.718 0.72 0.642
SC (%c) 23.81 17.05 17.19 23.81 24.77 19.01
CG (%c) 20.33 27.92 27.79 27.9 28 21.87
Design Variables
Spar End (c) 0.85 0.849 0.85 0.849 0.846 0.443
Spar Web (c) 0.85 0.303 0.318 0.849 0.843 0.443
m1 (kg/m) 0.332 0.271 0.273 0.328 0.334 0.23
m2 (kg/m) 0.009 0.063 0.06 0.036 0.038 0.22
Ply 2a (S-Glass) 5 1 2 5 4 1
Ply 2 (AFC) 1 1 1 1 1 1
Ply 4 (AFC) 1 1 1 1 1 1
Ply 6-7 (E-Glass) 1 9 10 1 1 1
Other Parameters
S44 (Nm2) 56.5 32.9 37 56.5 55.8 25.1
Act Mom (Nm) 2.38 1.37 1.54 2.38 2.33 0.76
nd
2 Flap Freq (/rev) 2.77 2.77 2.78 2.76 2.76 2.72
3rd Flap Freq (/rev) 5.4 5.42 5.44 5.38 5.36 5.07
Max ε11 2489 3994 3364 2601 2896 3935
Max ε12 4220 5600 4748 4798 4711 5591
The results obtained with mixed designs variables show the similar trend as it was
observed in the results with continuous design variables. The optimum value of the
objective function for the optimized cases obtained using mixed design variables is
136
always lower than that obtained with continuous design variables, however the difference
between the results depends on the objective function.
4.4 Optimization Studies with Ply Angles and Ply Thicknesses
In the results presented in this section, the ply angles are also included as the design
variables. The “Baseline 2” case, shown in Figure 4-5 and described in Section 4.3, is
used as the baseline case. The bounds for design variables and their baseline values are
listed in Table 4-20. The bounds for normalized ply thicknesses are the same as that
shown in Table 4-16. The bounds used for ply angle depends on the nature of the prepreg.
For the unidirectional plies, the ply angle varies from -90 to +90 degrees, whereas for the
bidirectional plies, the ply angle varies from 0 to 90 degrees. Even though the ply angle
can be treated as a continuous design variable, it is difficult to accurately manufacture a
composite structure where the ply angle has a real value. Hence, in the mixed-variable
optimization performed here, the ply angles are treated as discrete design variables for the
ease of manufacturing. In some of the earlier work [99], ply angles are discretized in
multiples of 5 or 10 degree. The framework presented here is also capable of working
with this discretization, however for the analysis presented in this section; the ply angle is
allowed to take any integer value within the bounds specified. The constraints used in the
analysis are the same as those described in Table 4-6.
137
Table 4-20: Design Variables for Optimization with Ply Thicknesses and Ply Angles
Design variables Baseline2 Lower Upper Ply Type

2 Spar End (c) 0.443 0.2 0.85
5 Ply 2a Thickness 1 1 5 S-Glass
8 Spar Web Ply 6 Thickness 1 1 10 E-Glass
Ply Angles
9 Ply 1 Angle 0 0 90 E-Glass
10 Ply 2a Angle 0 -90 90 S-Glass
11 Ply 2 Angle 45 -90 90 AFC
12 Ply 4 Angle -45 0 90 AFC
13 Spar Web Ply 6 Angle 0 0 90 E-Glass
4.4.1 Optimization Results with Continuous Design Variables
The results obtained, when all the design variables listed in Table 4-20 are treated as
continuous design variables, are shown in Table 4-21. The optimization study performed
in Section 4.3 is a subset of the analysis performed in this section. For some of the
objective functions, it was observed that the results obtained in Section 4.3 are the
optimal results and it is not possible to obtain further improvement in the optimum value
of the objective function by including ply angles as the additional design variables. This
is true for the Max θstat and Max θ4/rev cases shown in Table 4-21 and Table 4-18.
138
Table 4-21: Results for Optimization with Continuous Design Variables
Max Max Max Max Max

Objective Function
θstat (deg/m) 2.49 -2.22 -2.24 2.4 2.4 1.67
θ3/rev (deg) 3.67 4.91 4.66 3.9 3.9 2.4
θ4/rev (deg) 4.48 5.89 5.94 5.29 5.29 2.54
θ5/rev (deg) 6.81 3.53 4.31 8.02 8.02 2.38
θ345/rev 0.784 0.811 0.828 0.895 0.895 0.408
Constraints
st
1 Tor Freq(/rev) 6.01 4.51 4.69 5.32 5.32 5.48
M11 (kg/m) 0.72 0.719 0.719 0.72 0.72 0.642
SC (%c) 24.8 17.29 17.14 22.16 22.16 19.01
CG (%c) 27.75 27.79 27.75 27.99 27.99 21.87
Design Variables
Spar End (c) 0.85 0.85 0.85 0.85 0.85 0.443
Spar Web (c) 0.85 0.309 0.319 0.85 0.85 0.443
m1 (kg/m) 0.299 0.309 0.307 0.33 0.33 0.23
m2 (kg/m) 0 0 0 0.036 0.036 0.22
Normalized Ply Thicknesses
Ply 2a (S-Glass) 4.88 1 1.54 4.91 4.91 1
Ply 2 (AFC) 1.23 1.37 1 1 1 1
Ply 4 (AFC) 1.24 1 1.36 1 1 1
Ply 6-7 (E-Glass) 1.2 10 10 1 1 1
Ply Angles
Ply 1 (E-Glass) 0 0 0 0 0 0
Ply 2a (S-Glass) 0 1.9 0 62.2 62.2 0
Ply 2 (AFC) 45 43.5 45 -42 -42 45
Ply 4 (AFC) -45 -54.9 -45 47.1 47.1 -45
Ply 6-7 (E-Glass) 0 0 0 0 0 0
Other Parameters
2
S44 (Nm ) 66.2 34.8 37.9 57.4 57.4 25.1
Act Mom (Nm) 2.87 1.48 1.61 2.42 2.42 0.76
nd
2 Flap Freq (/rev) 2.79 2.78 2.79 2.76 2.76 2.72
rd
3 Flap Freq (/rev) 5.51 5.47 5.51 9.31 9.31 5.07
Max ε11 (με) 2478 3906 3547 4508 4508 3935
Max ε12 (με) 4347 5263 4968 6201 6201 5591
139
The results for Max θ3/rev case show a very small improvement with ply angles, as
compared to the results shown in Table 4-18. The improvement is obtained by changing
the ply angle for active plies away from ±45 degrees. Although the active twisting
moment generated is reduced due to the ply angle changes, the lowering of the torsional
frequency results in a higher dynamic twist at the blade tip. In the Max θ5/rev case also,
small changes are observed in the ply angle for active ply. But the most noticeable
change occurs in the ply angle for nose ply, which changes to 62.2 degrees. Similar to the
Max θ3/rev case, the changes in ply angle result in lower active twisting and also lower
torsional stiffness and first torsion frequency. The result obtained for Max θ345/rev case is
the same as that obtained for Max θ5/rev case.
4.4.2 Optimization with Mixed Design Variables
The results obtained, when the normalized ply thicknesses and ply angles listed in
Table 4-20 are treated as discrete design variables, are shown in Table 4-22. As observed
in the results with continuous optimization, for some of the objective functions
considered, it was not possible to find a better solution by including ply angles as the
design variables. For the mixed-variable optimization performed here, the results
obtained for Max θstat, Max θ3/rev, Max θ4/rev, Max θ345/rev could not be improved further.
For the Max θ5/rev case, the changes in ply angle for the nose ply (Ply 2a) and the
outermost ply (Ply 1) result in higher twisting moment, and thus larger dynamic twist as
compared to the optimum design shown in Table 4-19.
140
Table 4-22: Results Obtained for Optimization with Mixed Design Variables
Max Max Max Max Max

Objective Function
θstat (deg/m) 2.41 -2.22 -2.21 2.39 -2.39 1.67
θ3/rev (deg) 3.36 4.6 4.34 3.81 3.88 2.4
θ4/rev (deg) 4.3 5.28 5.41 5.01 5.31 2.54
θ5/rev (deg) 5.73 3.86 4.81 7.87 7.62 2.38
θ345/rev 0.708 0.769 0.798 0.867 0.878 0.408
Constraints
st
1 Tor Freq (/rev) 5.71 4.71 4.93 5.48 5.28 5.48
M11 (kg/m) 0.694 0.688 0.697 0.697 0.72 0.642
SC (%c) 23.81 17.05 17.19 21.51 24.77 19.01
CG (%c) 20.33 27.92 27.79 27.95 28 21.87
Design Variables
Spar End (c) 0.85 0.849 0.85 0.839 0.846 0.443
Spar Web (c) 0.85 0.303 0.318 0.839 0.843 0.443
m1 (kg/m) 0.332 0.271 0.273 0.313 0.334 0.23
m2 (kg/m) 0.009 0.063 0.06 0.034 0.038 0.22
Ply 2a (E-Glass) 5 1 2 5 4 1
Ply 2 (AFC) 1 1 1 1 1 1
Ply 4 (AFC) 1 1 1 1 1 1
Ply 6-7 (E-Glass) 1 9 10 1 1 1
Ply Angles
Ply 1 (E-Glass) 0 0 0 87 0 0
Ply 2a (S-Glass) 0 0 0 -28 0 0
Ply 2 (AFC) 45 45 45 -45 45 45
Ply 4 (AFC) -45 -45 -45 45 -45 -45
Ply 6-7 (E-Glass) 0 0 0 1 0 0
Other Parameters
2
S44 (Nm ) 56.5 32.9 37 58.1 55.8 25.1
Act Mom (Nm) 2.38 1.37 1.54 2.44 2.33 0.76
nd
2 Flap Freq (/rev) 2.77 2.77 2.78 2.76 2.76 2.72
rd
3 Flap Freq (/rev) 5.4 5.42 5.44 5.35 5.36 5.07
Max ε11 (με) 2489 3994 3364 1340 2896 3935
Max ε12 (με) 4220 5600 4748 6765 4711 5591
141
The shape of the cross section for the optimized cases is shown in Figure 4-6. In these
section, the leading edge ballast mass is presented by a red circle while the ballast mass
used near the vertical spar web is represented by a blue circle.
Figure 4-6: Cross Section for the Optimized Cases obtained with Mixed Design
Variables
4.5 Post Processing of Optimization Results
The final results obtained at the end of optimization process with mixed design
variables, as shown in Table 4-22, are analyzed further in order to check their validity.
142
Here three different kinds of analyses are performed. In the first check, the variation of
vibratory loads in forward flight condition is analyzed when no flap actuation is applied
in order to make sure that the optimized designs do not lead to higher baseline vibration
(vibration level in the absence of twist actuation). In the second analysis, variation of the
amplitude of dynamic twist with advance ratio is determined for different actuation
frequencies. And finally, circle plots are generated for each of the optimized cases in
forward flight condition at different actuation frequencies in order make sure that the
optimized results do provide higher authority for vibration reduction at the hub.
4.5.1 Effect on Baseline Vibration
In this case, the aeroelastic analysis for each of the optimized cases and baseline case is
performed at µ = 0.24 using the “Trim Analysis” (wind tunnel trim). When the trim
condition is reached, the amplitude of 4/rev vibratory load at the hub in fixed system is
recorded. The percentage difference in the vibratory loads for Fz, Mx and My components
with respect to the baseline case is shown in Figure 4-7. The results obtained show that
the increase in baseline vibration is less than 13% for all the optimized cases. Among all
the cases, the “Max θ5/rev” case shows maximum vibration.
143
15
FZ4 (%)
10
15
MX4 (%) 10
10
MY4 (%)
-5
Max Stat Max 3/rev Max 4/rev Max 5/rev
Figure 4-7: Percentage Increase in Vibratory Loads
4.5.2 Effect of Advance Ratio
In this section, aeroelastic studies with “Trim Analysis” were performed for each of
the optimized cases at different forward flight speeds. This study was performed to verify
the original assumption that there is no significant change in the amplitude of tip twist
with forward flight speed. The results obtained for actuation frequencies of 3, 4 and 5/rev
are shown in Figure 4-8, Figure 4-9 and Figure 4-10, respectively. The results obtained
show that the variation in the amplitude of dynamic tip twist with advance ratio is small.
Since the results presented here include “Trim Analysis”, they do not match exactly the
results shown in Table 4-22 where “Periodic Analysis” is used. For each of the actuation
frequency, the corresponding case provides maximum dynamic twist at all the advance
ratios considered.
144
3/rev comp
4.5
Amplitude of Tip Twist (deg)

4
3.5
Max 3/rev Case

3
Max 4/rev Case
Max 5/rev Case
2.5
Max stat Case
Baseline
2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Advance Ratio ()
Figure 4-8: Effect of advance ratio at 3/rev actuation frequency
4/rev comp
6
5.5
4.5
4 Max 3/rev Case

Max 4/rev Case
3.5
Max 5/rev Case
3 Max stat Case

Baseline
2.5
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Advance Ratio ()
Figure 4-9: Effect of Advance Ratio at 4/rev Actuation Frequency
145
4/rev comp
8

Max 3/rev Case
6 Max 4/rev Case

Max 5/rev Case
5 Max stat Case
Baseline
4
2
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Advance Ratio ()
Figure 4-10: Effect of Advance Ratio at 5/rev Actuation Frequency
4.5.3 Circle Plot for Optimized Cases
In order to generate the circle plot for each of the optimized cases and the baseline
case, the twist actuation is provided at a fixed frequency and the phase of actuation is
varied from 0 to 360 degree in the intervals of 30 degree. Once the response for each of
the hub loads in the fixed system is obtained, FFT is used to determine the sine and
cosine component of the response corresponding to 4/rev frequency. The circle plots
generated for 3/rev, 4/rev and 5/rev actuation frequencies for vertical component of the
force at the hub (Fz) are shown in Figure 4-11, Figure 4-12 and Figure 4-13, respectively.
146
Max 4/rev Case
200 Max 3/rev Case
Max 5/rev Case
100 Max stat Case
Baseline Case
FZ4,sin(N)
0
-100
-200
-300
-300 -200 -100 0 100 200 300 400
FZ4,cos(N)
Figure 4-11: Circle Plot for 3/rev Actuation Frequency
Since the optimum result obtained for the “Max θ3/rev” and “Max θ4/rev” cases are close
to each other, the circle plots corresponding to these cases for 3/rev and 4/rev actuation
frequencies are close to each other. As shown in Figure 4-11 and Figure 4-12, the circle
plot corresponding to “Max θ5/rev” case has larger size than that corresponding to “Max
θstat” case. Thus, each of the dynamically optimized cases performs better that statically
optimized case for 3/rev and 4/rev actuation frequency.

Max 3/rev Case
600
Max 5/rev Case
400
Max stat Case
200 Baseline Case
FZ4,sin(N)
-200
-400
-600
-800
-1000 -500 0 500 1000

FZ4,cos(N)
147
Max 4/rev Case
400
Max 3/rev Case
200 Max stat Case
Baseline Case
FZ4,sin(N)
100
-100
-200
-300
-400
-400 -200 0 200 400 600
FZ4,cos(N)
In case of the circle plot obtained at 5/rev actuation frequency, the “Max θ5/rev” case is
the most effective for vibration reduction as shown in Figure 4-13. Since, the optimum
design for “Max θstat” case is close to that for “Max θ5/rev” case, the “Max θstat” case
outperforms the “Max θ3/rev” and “Max θ4/rev” cases for vibration reduction at 5/rev
actuation frequency.
The results presented in this section highlight the original assumption that authority of
an active twist rotor to reduce vibratory loads at the hub can be increased by maximizing
the amplitude of dynamic twist obtained from twist actuation.
The use of prepreg material for manufacturing composite aerospace structures leads to
discrete design variables in the design and optimization studies. In order to obtain a
realistic and manufacturable design at the end of optimization, the ply thicknesses and ply
angles should be treated as discrete design variables. This chapter presented the
148
architecture of a design framework which can be used to perform optimization studies
with mixed design variables for designing a composite active twist rotor blade. In the
proposed framework, the optimum solution with mixed design variables is obtained using
three different methods, in addition to the optimum design when all the variables are
treated as continuous. This facilitates the designer to estimate the loss due to
“discretization” and make necessary changes to improve the design.
The mixed design variable optimization framework was successfully used to design
the cross section of a composite rotor blade with embedded active material. In the first
case, ply thicknesses were considered as discrete design variables, in addition to the
continuous design variables like the chordwise location of vertical spar web, ballast
masses and chordwise location where the active plies end. In this case, the minimum
allowable normalized thickness of prepreg plies was fixed at “0.1” instead of “0” to
prevent the mesh generator from crashing. The results obtained from these studies
showed that some of the plies had normalized ply thickness as “0.1” in the optimum
results, indicating that these plies should be removed from the analysis. In the next step,
these passive plies were not considered as design variable and the minimum allowable
normalized thickness was fixed to “1”. And the third case considered included the ply
angles as discrete design variables in addition to the ply thicknesses.
The final results obtained showed that:
1) The difference between the results obtained from continuous and mixed-variable
optimization depends on the objective function being considered.
149
2) The mixed design variable results obtained using three methods are close to each
other. And it is sufficient to obtain only the “Mixed Solution 1” and “Mixed
Solution 2” to predict the optimum solution with mixed design variables since the
“Mixed Solution 3” is very time consuming.
3) While maximizing the static and dynamic twist, the optimum design obtained
always led to a stiffer cross section and thus most of the optimum designs had
lower cross sectional strains.
4) A thick prepreg layer is required near the leading edge (Ply 2a) to obtain higher
active twisting moment, but it may increase the torsional stiffness which may
cause the dynamic twist performance to deteriorate.
5) Increasing the chordwise coverage of active plies is better than increasing the
thickness of active plies, in order to get higher static and dynamic twist. Also,
boxed-shaped spar design, in which the chordwise location where the spar plies
end and the chordwise location of vertical spar web are close to each other, is
suitable for maximizing the dynamic twist amplitude.
6) There is a significant difference in the optimum design obtained for different
actuation frequencies.
7) For the results obtained for this particular case, the optimum design obtained by
maximizing amplitude of dynamic twist at 3, 4 and 5/rev actuation frequencies
tends to be closer to result obtained for maximizing amplitude for 5/rev actuation
150
frequency. This can be attributed to the higher cross-sectional stiffness in Max
θ5/rev case which in turn results in higher active twisting moment.
151
Chapter 5. New Strategy for Design of Composite Rotor
Blade with Active Flaps
The optimization framework and strategy developed for the design of rotor blades
with active twist is extended for the design of composite rotor blade with active flaps in
this chapter. The optimum blade design must aim to maximize the flap authority for
vibration reduction while satisfying the constraints on the chordwise location of cross-
sectional center of gravity (CG) and shear center (SC), blade mass per unit length,
torsional frequency, etc. Since the vibration reduction in rotors with active flaps is
obtained through the servo-flap effect, the amplitude of tip twist obtained from flap
actuation is a good metric for estimating the flap authority. Thus, in the analysis
performed here, the amplitude of tip twist obtained due to the flap motion is used as the
objective function which is maximized. The objectives of this chapter are:
a) To demonstrate the use of new optimization strategy with high fidelity analysis
tools for designing a rotor blade with active flaps
b) To design a realistic composite rotor blade to enhance the effect of active flaps for
vibration reduction.
5.1 Aeroelastic Analysis with Active Flaps
The aeroelastic analysis performed for analyzing a rotor blade with active flaps is
similar to that described in Chapter 3 for analyzing an active twist rotor blade. The
aeroelastic analysis performed using RCAS has been modified to account for the
152
presence of active flaps. The aerodynamic analysis for active flaps is performed in RCAS
using the flexible airfoil theory [167] and 3-D dynamic inflow [168] model. The table
lookup required for analyzing a rotor blade with active flaps in RCAS is generated using
XFOIL.
5.2 Baseline Rotor Blade
A composite rotor blade that can be tested in the University of Michigan spin test
stand (see Figure 5-1) is used as the baseline rotor blade. This test stand was used earlier
for testing a 1/6th Mach-scaled CH-47D rotor blade with active twist [58] and active flap
[112]. The properties of the test stand and baseline rotor blade are given in Table 5-1. The
rotor has a 10 ft diameter and it has articulated configuration with a root offset of 0.15R.
The nominal operating RPM is 1336 which corresponds to Mach number of 0.6 at the
blade tip in hover condition.
Figure 5-1: University of Michigan Spin Test Stand
153
The composite rotor blade is made up of E-Glass and IM7 graphite plies as shown in
Figure 5-2. The baseline blade has leading edge ballast mass to bring the CG of the cross
section close to the quarter chord. The D-shaped cross section consists of 2 layers of E-
Glass plies oriented at ±45 deg as the overwrap plies. The main spar consists of 2 layers
of IM7 at 0 deg and 2 layers of E-Glass at ±45 degrees. The vertical spar web located at
0.38c has 3 layers of E-Glass at 0/90 degree.
Table 5-1: Rotor Properties
Configuration Articulated
Radius 1.54m
Chord 0.136m
Num of Blades 4
Air Density 1.225 kg/m3
CT 0.0066
RPM 1336
Hinge Location 0.15R
CG (%c) 25.32
SC (%c) 22.8
Airfoil VR7
Figure 5-2: Cross-Sectional Layup for the Baseline Rotor Blade
For testing the rotor blade with active flaps, a piezoelectric actuator is mounted in the
spar region of blade as shown in Figure 5-3. In order to mount the actuator and install the
flap, cutouts are made in the spar region of blade and near the trailing edge. The loss in
stiffness due to these cutouts is balanced by including additional plies in the cutout
154
region. However, the presence of actuator and flap supports leads to additional mass in
the blade which adds to the inertia of the blade. In the aeroelastic analysis performed in
this chapter, the effect of actuator and flap support inertia is included. The reference
actuator used in this study is the X-frame actuator developed at MIT [119] and it is small
enough to fit inside the spar of rotor blade cross section as shown in Figure 5-3. The
details of the first single flap configuration considered in this study are presented in Table
5-2. The structural frequencies of the baseline rotor blade (with the actuator and flap
support inertia) are listed in Table 5-3.
Figure 5-3: Active Flap Mechanism
Table 5-2: Flap and Actuator Dimensions
Flap Dimensions
Flap Length 0.12R
Flap Chord (cf) 0.25c
Flap Center (rf) 0.78R
Actuator Details
Size 3.6" x 0.9" x 0.5"
Actuator Location 0.25c and 0.78R
Actuator+Support Mass 96 gm
155
Table 5-3: Structural Frequencies of the Rotor Blade
Mode Shape Frequency (/rev)

1st Chordwise 0.51
1st Flapwise 1.12
2nd Flapwise 3.57
1st Torsion 4.21
3rd Flapwise 6.86
2nd Chordwise 7.49
5.3 Preliminary Analysis
A “Periodic Analysis” was performed for analyzing the rotor blade with active twist
instead of “Trim Analysis” in the optimization studies performed in the earlier chapter to
reduce the computational time. The trim targets used in the “Trim Analysis” are: C T =
0.0066, no cyclic moments (Mx = 0 and My = 0), and the blade pitch settings are used as
the trim targets (wind tunnel trim). In the case of “Periodic Analysis”, the blade pitch
settings are kept constant and the equations of motion are solved in time domain till the
system response is periodic. It was demonstrated using the results obtained from
optimization in Chapter 3 that a “Periodic Analysis” is sufficient to obtain a design which
maximizes the dynamic twist. Hence, in the case of active flap studies, both “Periodic
Analysis” and “Trim Analysis” are performed in the preliminary analysis to verify if
“Periodic Analysis” is sufficient to capture the amplitude of dynamic twist accurately.
In the first study, the frequency of actuation is varied from 3/rev to 5/rev in hover
conditions for flap deflection of ±4 degrees and the mean value and amplitude of twist at
the blade tip are obtained. The results obtained with “Periodic Analysis” and “Trim
Analysis” is shown in Figure 5-4. The results show that although there is a significant
difference in the mean value of twist at the blade tip, the amplitude of dynamic twist for
156
both the analyses are very close to each other. A higher mean value for the tip twist
implies that the blade is experiencing higher aerodynamic loads in the trim analysis.
Next, the advance ratio was varied from 0.0 to 0.3 and the amplitude of dynamic twist
was determined using “Trim Analysis” and “Periodic Analysis” for 4/rev actuation
frequency. The variation of amplitude of tip twist with advance ratio is shown in Figure
5-5. For low advance ratios, the difference between the amplitude of tip twist predicted
by both the analyses is small; however, it increases with an increase in the forward flight
speed. The blade pitch settings obtained from the “Trim Analysis” (for analysis with and
without the flap motion) and pitch settings used in the “Periodic Analysis” are shown in
Figure 5-6. The results obtained with “Trim Analysis” show that for small values of µ,
only the collective pitch angle is used for obtaining trim while the contribution from
cyclic pitch angles is small. For these cases, the amplitude of dynamic twist obtained
from “Periodic Analysis” and “Trim Analysis” match very well. As the advance ratio
increases, there is a decrease in collective pitch angle and increase in cyclic pitch angles,
which leads to a noticeable in the amplitude of dynamic twist obtained from these two
analyses.
1 6
Trim
Mean Tip Twist (deg)
Periodic
0.8 4 Trim
Periodic
0.6 2
0.4 0
2 3 4 5 6 2 3 4 5 6
Frequency (/rev) Frequency (/rev)
Figure 5-4: Variation of Tip Twist with Actuation Frequency
157
1.2

1
0.8
Trim
0.6 Periodic
0.4
0.2
0
0 0.1 0.2 0.3 0.4
Advance Ratio ()
Figure 5-5: Variation of Amplitude of Tip Twist with Advance Ratio

0 (deg)
0
0 0.05 0.1 0.15 0.2 0.25 0.3

3
1c (deg)
2
1
0
-1 Trim (with Flap)
0 0.05 0.1 0.15 0.2 0.25
Trim (without0.3
Flap)

Periodic Case
1s (deg)
0
-2
-4
0 0.05 0.1 0.15 0.2 0.25 0.3

Figure 5-6: Variation of Trim Variables with Advance Ratio
In order to determine the optimum range of flap deflection required to obtain vibration
reduction, the circle plot technique was used. To obtain a circle plot, the flap was
actuated with 4 degree amplitude at 4/rev frequency and the phase of actuation was
varied from 0 to 360 degrees in intervals of 30 degree. For each of the responses, the
cosine component of the 4/rev load was plotted against the sine component of 4/rev load,
as shown in Figure 5-7. The line drawn from the baseline vibration point joins the point
on the circle plot which corresponds to a phase angle of 0 degree for flap actuation. For
158
each of the vibratory hub load component, the origin (which corresponds to zero
vibration) is enclosed by the circle plot. This implies that the flap deflection of ±4
degrees at 4/rev actuation frequency is sufficient to reduce the 4/rev component of the
vibratory load at hub.
Based on the results obtained in the preliminary analysis, it can be concluded that:
a) The amplitude of dynamic twist obtained from the “Periodic Analysis” is very
close to that obtained from the “Trim Analysis” for different flap actuation
frequencies at small values of µ.
b) The “Trim Analysis” shows that there is approximately 15% variation in the
amplitude of dynamic twist for higher advance ratios. Thus, the amplitude of
dynamic twist obtained in hover condition is representative of the amplitude of
dynamic twist which will be obtained in forward flight conditions.
c) Hence, the amplitude of dynamic twist obtained in hover condition using
“Periodic Analysis” can be used as the objective function for the optimization
studies.
159
50 1000
500
(N)
0
FZ4,sin(N)
Y4,sinsin
sin
FFY4
0
FZ4
-50
-500
-100 -1000
-100 -50 0 50 -500 0 500 1000
FY4 FZ4
FY4,cos (N)
cos FZ4,coscos
(N)
1000 1000
(Nm)
X4,sin(Nm)
500 500
Y4,sinsin
MMX4 sin
MMY4
0 0
-500 -500
-500 0 500 1000 -1000 -500 0 500
MX4 MY4
M cos(Nm)
X4,cos MY4,cos
cos(Nm)

(* = Vibration for the Baseline Case, * = Vibration for Flap Actuated Case, □ = Origin)
5.4 Optimization Problem Definition
In this section, optimization studies are presented in which the cross-sectional layup is
determined for a composite rotor blade with active flap. The aim of the optimization
study is to enhance the effectiveness of active flaps for vibration reduction. In order to
achieve this target, optimization study is performed with the amplitude of dynamic twist
in hover condition as the objective function. The objective function considered in this
study is: Maximize the amplitude of tip twist for 4/rev flap actuation (Max θ4/rev). The
optimization studies are performed for active flap located at three different spanwise
locations, as shown in Figure 5-8.
160
Center of Rotation
Actuator 0.12R
Case 1 Active Flap
rf = 0.78R
0.25c
Case 2
rf = 0.70R
Case 3
rf = 0.85R
Figure 5-8: Spanwise Location for Active Flaps
The cross section for the baseline blade is shown in Figure 5-2. The design variables
used in the optimization are listed in Table 5-4. The design variables used are: the
chordwise location where the spar plies end, chordwise location of the vertical spar web,
the two ballast masses, and the normalized ply thicknesses and ply angles for all the
composite plies used in the cross section. Even though the baseline blade includes only
one leading edge ballast mass, additional ballast mass is used as the design variable since
it is useful in tuning the dynamic properties of rotor blade, as observed in the results
presented in Chapter 3 and 4. In order to make the rotor blade design more realistic, the
location of first ballast mass is fixed near the leading edge at x = 0.02c while the second
ballast mass is located just in front of the vertical spar web. (This is done to ensure that
the ballast mass is added in the region where passive plies can be used to support it and
thus prevent the ballast mass for flying out during the operation). In order to prevent the
mesh generator from crashing, the lower bound on ply thickness is fixed at 0.1 instead of
zero. A value of 0.1 for normalized ply thickness in an optimum design implies that that
161
particular ply is not required in the cross section and should be removed in the next
optimization. In this study, the lower bound for the normalized thickness of Ply 1 is “1,”
since the outermost ply cannot be removed from the cross section. The bounds used for
ply angles depend upon the nature of the prepreg. For the unidirectional plies, the ply
angle varies from -90 to +90 degrees, whereas for the bidirectional plies, the ply angle
varies from 0 to 90 degrees.
Table 5-4: Design Variable used for Optimization Study

1 Spar End (c) 0.4 0.2 0.85
4 Ballast Mass 2 (m2) (kg/m) 0 0 0.25
5 Ply 1 Thickness 2 1 4 E-Glass
6 Ply 2 Thickness 2 0.1 4 IM7
7 Ply 3 Thickness 2 0.1 4 E-Glass
8 Ply 4 Thickness 3 0.1 10 E-Glass
Ply Angle
10 Ply 2 Angle 0 -90 90 IM7
Table 5-5: Constraints used in the Optimization
Constraints Baseline Min Max

SC (%c) 22.8 17 25
CG (%c) 25.3 20 28
M11 (kg/m) 0.309 0.26 0.36
Tor Freq (/rev) 4.22 3 6
Max ε11 (µε) 2605 0 3500
Max ε12 (µε) 4042 0 5000
162
The constraints used in the optimization are shown in Table 5-5. The mass per unit
length for the cross section is constrained to lie within ±15% of the baseline value. In the
optimization studies performed earlier for active twist rotor, the maximum value of
strains in the cross section was not considered as a constraint. The optimum results
obtained for active twist rotor were always stiffer as compared to the baseline cross
section and in general, they had lower cross-sectional strains. However, in the case of
optimization studies with active flap, the tendency of the optimizer is to design a rotor
blades cross section with lower cross-sectional stiffness. Hence, most of the designs
obtained without the constraint on strains had very high cross sectional strains. It should
be noted that the maximum value for allowable strains used in the constraints listed in
Table 5-5 is well below the maximum strain limit for the material. Since a “Periodic
Analysis” is performed here in hover condition, the blade does not experience worst-case
aerodynamic loads. Hence, the strains observed in the cross section are mainly due to the
large centrifugal force acting on the blade.
5.5 Optimization Results
The optimization problem is solved using the Mixed-Variable Optimization
Framework described in Chapter 4. As observed in Chapter 4, the framework is capable
of determining solution with both continuous design variables and with mixed design
variables. In case of optimization with mixed design variables, the ply angles and ply
thicknesses are treated as discrete. In this section, only the best result obtained with
mixed design variables is presented and compared with the results obtained using
continuous design variable. The final results obtained for objective function is shown in
Table 5-6.
163
Table 5-6: Results for Flap centered at rf = 0.78R
Cases Max θ4/rev Max θ4/rev Baseline

Objective Function Cont Mixed
θ4/rev (deg) 1.02 1.01 0.89
Constraints
Tor Freq (/rev) 4.13 3.91 4.21
M11 (kg/m) 0.27 0.28 0.31
SC (%c) 22.53 20.30 22.81
CG (%c) 27.79 24.01 25.32
Max ε11 (με) 1266 1216 2606
Max ε12 (με) 3558 3715 4041
Design Variables
Spar End (c) 0.446 0.456 0.4
Spar Web Loc (c) 0.307 0.289 0.4
m1 (kg/m) 0.050 0.051 0.1
m2 (kg/m) 0.001 0.015 0
Ply 1 (E-Glass) 1.27 1 2
Ply 2 (IM7) 3.72 4 2
Ply 3 (E_Glass) 1.04 1 2
Ply 4 (E-Glass) 7.51 8 3
Ply Angles
Ply 1 (E-Glass) 61.2 64 45
Ply 2 (IM7) 8.2 8 0
Ply 3 (E-Glass) 55 53 45
Ply 4 (E-Glass) 56 57 0
Other Parameters
S44 (Nm2) 97.5 90.0 119.8
2
S55 (Nm ) 532.3 573.4 332.9
nd
2 Flap Freq (/rev) 3.91 3.98 3.57
rd
3 Flap Freq (/rev) 7.89 8.51 6.86
Induced FZ4 (N) 319.08 318.92 272.22
The results obtained show 14% increase in the twist amplitude corresponding to 4/rev
actuation frequency. The percentage increase in the amplitude of twist at the blade tip is
small since the baseline case is close to the optimum design obtained by maximizing the
164
twist amplitude at 4/rev actuation frequency. The most important parameter while
maximizing the dynamic twist amplitude is the first torsion frequency. The optimizer
tries to get the first torsion frequency in the neighborhood of the flap actuation frequency.
Besides this, the mass per unit length for each of the optimized cases is near its lower
bound. The chordwise location of shear center stays close to the value obtained for the
baseline case. The amount of ballast mass used in the cross section has reduced although
there is a redistribution of ballast masses indicating that the optimizer is using the ballast
masses to increase the torsional inertia in order to reduce the torsion frequency. The
torsional stiffness for the cross section is mainly controlled by varying the thickness of
Ply 3 which is the E-Glass ply. For the optimized case, the thickness of unidirectional
IM7 ply (Ply 2) is at the maximum allowable value and its orientation is very close to 0
degree. Thus, the IM7 plies are mainly used to withstand the large centrifugal loads
acting on the rotor blade cross section and thereby reduce the cross-sectional strains. The
effect of increase in the thickness of unidirectional plies can also be seen in the value of
cross-sectional bending stiffness (S55) and second and third flapwise bending frequencies.
The thickness of Ply 1 is at the minimum allowable value. The thickness of Ply 4 is
important to keep the chordwise location of shear center within the bounds specified in
constraints.
165
1000
Max  Case
4/rev
Baseline Case
500
FZ4,sin(N) 0
-500
-1000
-1000 -500 0 500 1000 1500
F (N)
Z4,cos
Figure 5-9: Circle Plot at rf = 0.78R

The circle plot obtained for the optimized case is shown in Figure 5-9. It clearly shows
the increase in control authority for vibration reduction in vertical component of the force
at 4/rev frequency (FZ4) at the hub.
In the next optimization study, the spanwise location of the flap (rf) was fixed at 0.7R.
The design variables and constraints used in the optimization are the same as those listed
in Table 5-4 and Table 5-5, respectively. The final optimized result is shown in Table
5-7, for both the continuous and mixed design variable cases. The trend observed for
design variables is the same as it was observed in previous results. The results show
9.32% increase in the amplitude of dynamic twist. The corresponding circle plot is shown
in Figure 5-10. Here, the results show that when the flap is not actuated, the optimized
case produces higher vibration but the increase in control authority for the optimized case
over compensates for this.
166
Table 5-7: Results at rf = 0.7R

θ4/rev (deg) 0.645 0.628 0.590
Constraints
Tor Freq (/rev) 4.30 4.46 4.39
M11 (kg/m) 0.265 0.265 0.308
SC (%c) 24.9 21.9 22.5
CG (%c) 28.0 25.6 24.2
Max ε11 (με) 976 1552 2556
Max ε12 (με) 2143 4052 4042
Design Variables
Spar End (c) 0.55 0.41 0.4
Spar Web Loc (c) 0.46 0.32 0.4
m1 (kg/m) 0.06 0.04 0.1
m2 (kg/m) 0.00 0.03 0
Ply 1 (E-Glass) 1.01 1 2
Ply 2 (IM7) 4.00 4 2
Ply 3 (E-Glass) 0.28 1 2
Ply 4 (E-Glass) 2.58 6 3
Ply Angles
Ply 1 (E-Glass) 64.30 49 45
Ply 2 (IM7) 4.50 5 0
Ply 3 (E-Glass) 7.90 61 45
Ply 4 (E-Glass) 86.10 15 0
Other Parameters
S44 (Nm2) 88.64 85.77 118.92
S55 (Nm2) 785.27 556.76 332.83
2nd Flap Freq (/rev) 4.03 3.74 3.34
3rd Flap Freq (/rev) 8.85 8.18 6.71
Induced FZ4 (N) 233.12 219.99 204.98
167
600
Max 4/rev Case
400 Baseline Case
200
FZ4,sin(N) 0
-200
-400
-600
-800
-500 0 500 1000
FZ4,cos(N)
Figure 5-10: Circle Plot at 0.7R
Finally, the optimization was performed at rf = 0.85R. The results obtained for this
case is shown in Table 5-8. For this case, the final result obtained with continuous design
variables and mixed design variables and the baseline case are very close to each other.
The increase in the amplitude of dynamic twist for this case is only 4.58%. The circle plot
for the optimized case and the baseline case for flap located at rf = 0.85R is shown in
Figure 5-11. The final results obtained for the three cases considered here are shown in
Figure 5-12. The results show the expected trend where a higher flap deflection is
obtained as the flaps are moved outboard. Also, the percentage increase in the amplitude
of tip twist is different for each of the case. It is highest for the flaps located at 0.78R. It
should be noted that, although the layup for the baseline cases is same in all the cases, the
blade frequencies are different for the three baseline cases due to the inertia effects of
flaps and actuator.
168
Table 5-8: Results for Flap at 0.85R

θ4/rev (deg) 1.14 1.14 1.09
Constraints
st
1 Tor Freq (/rev) 4.26 4.26 4.16
M11 (kg/m) 0.266 0.272 0.308
SC (%c) 23.48 24.38 22.5
CG (%c) 27.39 26.84 24.23
Max ε11 (με) 2010 2004 2658
Max ε12 (με) 3961 3896 4143
Design Variables
Spar End (c) 0.385 0.388 0.4
Spar Web Loc (c) 0.379 0.387 0.4
m1 (kg/m) 0.026 0.03 0.1
m2 (kg/m) 0.016 0.017 0
Ply 1 (E-Glass) 1 1 2
Ply 2 (IM7) 3.04 3 2
Ply 3 (E-Glass) 3.98 4 2
Ply 4 (E-Glass) 6.06 6 3
Ply Angles
Ply 1 (E-Glass) 68 68 45
Ply 2 (IM7) -3 -3 0
Ply 3 (E-Glass) 75 75 45
Ply 4 (E-Glass) 0.9 7 0
Other Parameters
S44 (Nm2) 84.6 86.4 118.9
S55 (Nm2) 473.7 467.4 332.8
nd
2 Flap Freq (/rev) 3.89 3.86 3.62
rd
3 Flap Freq (/rev) 8.4 8.31 6.98
Induced FZ4 (N) 361.8 360.4 350.3
169
Max 4/rev Case
500 Baseline Case
FZ4,sin(N)
0
-500
-1000
-1000 -500 0 500 1000
FZ4,cos(N)
Figure 5-11: Circle Plot for Optimized Result at rf = 0.85R
1.3
Continuous Optimum
1.2 Mixed Design Variable
Baseline
1.1
1
4/rev (deg)
0.9
0.8
0.7
0.6
0.5
0.65 0.7 0.75 0.8 0.85 0.9
Flap Location rf (R)
Figure 5-12: Optimized Results for different Spanwise Locations
The shape of the cross section for the baseline case and the optimized cases is shown
in Figure 5-13. The results show that the spars get thicker as the spanwise location of flap
170
moves outboard. This is expected since the centrifugal force acting on the blade is highest
when the actuator and flap mass is located near the tip region.
Figure 5-13: Optimized Cross Sections
This chapter presented the modified version of the mixed design variable optimization
framework for the design of a composite rotor blade with active flaps. The optimization
framework was used to design composite rotor blade for a Mach-scaled rotor blade which
can be tested in the University of Michigan spin test stand. In this chapter, optimization
study was performed for the flaps located at three different spanwise locations. Here, the
analysis includes the inertia effects of flap and actuator. The stiffness properties are
171
assumed to be uniform over the complete rotor blade (except the root region). The
optimization studies were performed with ply thicknesses, ply angles, chordwise location
of vertical spar web, and ballast masses as the design variables. The constraints are
imposed on the chordwise location of CG and SC, first torsional frequency, mass per unit
length and cross-sectional strains. In all the optimization studies, the amplitude of
dynamic twist at 4/rev actuation frequency is used as the objective function which is
maximized.
The optimization results showed:
a) 14% increase in dynamic twist for rf = 0.78R
b) 9.32% increase in dynamic twist at rf = 0.7R
c) 4.58% increase in dynamic twist at rf = 0.85R
For each of the optimized cases, circle plots were obtained which show higher control
authority for vibration reduction as compared to the baseline case. In all the optimized
cases, there is a decrease in torsional stiffness and an increase in flapwise bending
stiffness. Higher axial stiffness is helpful in reducing strains in the cross section due to
the large centrifugal force. As observed in the case of active twist optimization studies,
the optimizer tries to get the first torsion frequency of the blade closer to the actuation
frequency. The dynamic tuning of the blade was performed by varying the ply angles for
plies in the cross section and by varying the ballast masses used in the cross section. The
optimization results are obtained using both continuous design variables and mixed
design variables. In the case of the mixed variable optimization, ply angles and ply
thicknesses were treated as discrete design variables. For all the optimized results
172
obtained, the results obtained with mixed design variables are close to those obtained
with continuous design variables.
The framework developed here can be extended to analyze and design more
complicated active flap configurations like blade with dual flaps or microflaps. The
framework also allows for including more sections along the span which will result in an
increase in the number of design variables.
173
Chapter 6. Performance Enhancement and Vibration
Reduction in Dynamic Stall Condition using Active Camber
Deformation
Deformable airfoils present a unique way to contribute to vibration reduction and
performance enhancement in rotating wings. Camber deformation of the airfoil section is
seen as an aerodynamically efficient alternative for controlling aerodynamic loads in
order to obtain vibration reduction and performance enhancement. This chapter presents
detailed analysis with cubic and quadratic camber deformation shape function to obtain
these objectives.
In this study, the analysis is performed at µ = 0.33, where the dynamic stall effects
lead to high vibratory loads and poor performance. The aerodynamic analysis performed
in this chapter includes a unified airloads model that allows arbitrary airfoil morphing
with dynamic stall model, as described in [169]. In the next step, a global search over the
parameter space – i.e., camber deformation amplitude, phase, and frequency of actuation
– is conducted to identify the optimum points for the vibration and performance
characteristics of the rotor blades with deformable airfoils. The optimization problem is
solved using a surrogate-based approach as described in Chapter 3. Then starting from
these optimum points of surrogated-based approach as initial points, gradient-based
optimization is conducted using fmincon in MATLAB, to obtain the best results possible.
The objectives of this chapter are:
174
1) To demonstrate the implementation of unified airloads model for modeling the
dynamic stall effects for arbitrary airfoil morphing.
2) To explore the possibility of performance enhancement and vibration reduction in
dynamic stall condition using active camber deformation.
3) To compare the two different modes of camber deformation: camber deformation
with a quadratic shape function and camber deformation with a cubic shape
function.
6.1 Optimization Framework
The mathematical optimization problem can be stated as:
min f(x)
subject to: xl  x  xu
where f is the objective function, which can be the vibratory vertical hub load at 4/rev
frequency (FZ4) or the performance related moment (MZ0), x is the set of design variables
that are bounded between a lower (xl) and an upper (xu) limits. The design variables used
in this optimization problem are the amplitudes and phases of the camber deformation at
different actuation frequencies.
The optimization process involves two different steps as shown in Figure 6-1. In the
first step, initial range of design variables is given as input to create a surrogate model.
Optimum results obtained from the surrogate based optimization (SBO) after multiple
iterations are used as the initial values for the gradient based optimization (GBO)
performed using MATLAB’s fmincon. This process allows the gradient-based
175
optimization to start from different initial feasible points and perform local search for
minima.
Figure 6-1: Two-step Optimization Process
6.1.1 Surrogate-based Optimization (SBO)
The SBO performed in this chapter is similar to that described in the earlier chapters.
In order to form the surrogate, the objective function must first be evaluated over an
initial set of design points. The surrogate is then generated by interpolating the initial
design points. The MATLAB Latin hypercube function lhsdesign was used to generate
the space-filling design of experiments used in this study. The points in the Latin
hypercube represent design points at which aeroelastic simulations are to be conducted.
Each simulation can be run independently of simulations at other design points; therefore
the initial set of sample points is generated using distributed computers.
176
Once an initial set of fitting points have been produced, kriging interpolation [162,
170] is used to create the surrogate for the vibration and performance objective functions.
After the surrogate objective function is created using kriging, the Efficient Global
Optimization (EGO) algorithm [159] is used to determine optimum points. EGO
algorithm accounts for uncertainty in the surrogate and is more efficient.
6.1.2 Gradient-based Optimization
The GBO is performed within MATLAB, using fmincon function from its
optimization toolbox. The fmincon function minimizes a constrained nonlinear
multivariable problem. In each of the iteration, the function solves a quadratic
programming subproblem. Since the objective function is highly nonlinear, and since the
design hyperspace is very complex, it is possible for fmincon to fall into a local extrema,
leading to a sub-optimal solution. Therefore, it is necessary to run the optimization to
completion, starting from different initial points. These initial points are determined from
the optimization performed using the surrogate approach. At the end of the cycle, the
gradient-based optimization provides a better optimum than if only the surrogated was
used.
6.1.3 Aeroelastic Framework (UM/NLABS-A)
In order to analyze morphing-type rotors, one must consider several effects normally
assumed to be unimportant in rotor problems. A quasi-3D geometrically nonlinear model
for the aeroelastic analysis of an airfoil with camber deformation was developed in
UM/NLABS-A (Non-linear Anisotropic Beam Solver - Aeroelastic) [145, 146].
177
Schematic of the various components in the aeroelastic framework is shown in Figure 6-2
and the details are described next.
Figure 6-2: UM/NLABS-A Framework [145]
Structural model
The computational structural dynamics formulation used in the current study has been
presented in [72, 171]. It follows the variational-asymptotic method for the analysis of
composite beams [70]; that is, the equations of motion for a slender anisotropic elastic
three-dimensional solid are approximated by the recursive solution of a linear two-
dimensional problem at each cross section and a one-dimensional geometrically-
nonlinear problem along the reference line. This procedure allows the asymptotic
approximation of the three-dimensional warping field in the beam cross sections, which
are used with the one-dimensional beam solution to recover the three-dimensional
178
displacement field. The present implementation adds an arbitrary expansion of the
displacement field through a set of functions approximating the sectional deformation
field to capture “non-classical” camber deformations, which are referred to as finite-
section modes.
Aerodynamic Model
The aerodynamic analysis requires a unified model that allows for arbitrary airfoil
motion, unsteady freestream, morphing airfoil shape and dynamic stall. The three key
elements of the unified model are: the Peters flexible airfoil theory, the 3D dynamic
inflow model, and the modified ONERA dynamic stall model. The schematic of the
unified airloads model is shown in Figure 6-3.
Figure 6-3: Schematic of Unified Airloads Model
The low-order aerodynamic model uses the two-dimensional finite-state formulation
for deformable airfoils presented in [167]. It is based on a Glauert expansion of the
potential flow equations for a deformable airfoil of infinitesimal thickness. The camber-
wise airfoil deformation is written using the Chebyshev polynomials of the first kind,
which defines the generalized camber-wise displacement amplitudes. The wake-induced
179
velocity is solved using the dynamic inflow theory. It assumes that the velocity normal to
the rotor disk can be expressed in terms of radial and azimuthal expansion functions. The
dynamic inflow model is sufficient to capture effects of Nb/rev vibrations and it can be
used in the design of controllers, however, it cannot be used for modeling blade-vortex
interactions or acoustical phenomena [168]. Also, at μ = 0.33, the effect of blade vortex
interactions are expected to be less dominant as compared to the dynamic stall effects.
Drag and Dynamic Stall
A potential benefit of camber actuation is the ability to alter profile drag and stall
characteristics, which have implications in power and vibration. To include these effects
in the low-order aerodynamic model, the potential flow airload expressions are
augmented with a quasi-static profile drag term as well as a dynamic-stall correction that
is based on the ONERA model. The ONERA model assumes that the dynamic stall states
are governed by a second-order differential equation, and requires static loading
coefficients near and beyond stall. These are determined using the two-dimensional
boundary layer analysis code XFOIL (which is valid to slightly post stall conditions),
along with a simple, empirically derived approximation for deep-stall. The coefficients
are obtained under varying Reynolds number, angle-of-attack and camber deformation. A
detailed account of the method used for determining the coefficients is available in
Thepvongs et al. [145]. The dynamic stall formulation was appended with a first order
model to capture the delay in angle of attack. The steps involved in determining the effect
of dynamic stall on aerodynamic loads are described in Appendix B.3.
180
Coupling with Finite-State Aerodynamics
The finite-state aerodynamics formulation uses Chebyshev polynomials to form a
basis for the camber deformations and associated airloads, while the choice of basis
functions for the finite-section modes are arbitrary. The motion and force variables of the
aerodynamics formulation are related to those of the structural formulation by a simple
linear expression, as derived in Thepvongs et al. [145]. This straightforward connection
between the aerodynamic and structural states allows the same space and time integration
methods to be used for both formulations as well as a simultaneous solution. The
governing structural dynamics equations, aerodynamic load expressions, dynamic stall
equations and wake equations together define the time-domain problem. An explicit
method is used with iterative refinement to achieve the desired convergence. A simple
three-point backwards Euler time-integration scheme is used in accordance with the first-
order form of the structural and potential flow governing equations. A four-point scheme
is used to integrate the second-order dynamic stall equations.
Trim Analysis
The enforcement of vehicle equilibrium adds more variables and constraints to the
aeroelastic problem. The present work assumes a wind-tunnel setup, where the variables
are taken to be the collective, sine and cosine components of the cyclic pitch, and
equilibrium is represented by specifying values for the time-averaged thrust, pitch and
roll moments. These are provided by an autopilot that makes incremental changes to the
control settings at every time step. The “trimmability matrix” can be approximated by
181
numerically computed Jacobian, determined by stepping the controls and examining the
response at an instant one revolution later.
6.2 Camber Deformation Shape
The non-dimensionalized cubic (Ψcubic) and quadratic (Ψquad) camber deformation
shape functions used in this analysis are shown in Figure 6-4. The expressions for these
camber deformation are given by:
1
 quad ( )   2 
3
5 3 
 cubic ( )    3   
4 5 
where  is the airfoil chordwise coordinate non-dimensionalized by its half chord such
that 1    1.
Camber deformation in the current analysis is simulated by assigning an arbitrarily
high cross-sectional stiffness associated with the camber degree of freedom and applying
a conjugate finite-section force in the structural simulation. This method allows the user
to control airfoil deformation without defining the particular actuation mechanism. The
amplitude of camber deformation is obtained by taking the difference between the
maximum camber deformation and the minimum camber deformation along the airfoil
chord. In all the results presented in this study, the camber deformation is shown as a
percentage of the airfoil chord. Airfoil cross section with a 5% camber deformation for
both cubic shape function and quadratic shape function are shown in Figure 6-5.
182
0.5
Cubic
0
-0.5
0 0.2 0.4 0.6 0.8 1
Airfoil Chord
1
Quad
0
-1
0 0.2 0.4 0.6 0.8 1
Airfoil Chord
Figure 6-4: Camber Deformation Shape Function
0.1
0
-0.1
0.2 0.4 0.6 0.8 1
Blade Chord
0.1
0
-0.1
0.2 0.4 0.6 0.8 1
Blade Chord
Figure 6-5: Airfoil Cross Section with 5% Camber Deformation
(Dotted red line: Undeformed NACA 0012 Airfoil, solid black line: Deformed Airfoil)
Note that a 5% camber deformation is shown in Figure 6-5 only for visualization
purpose. Based on the preliminary analysis performed, it was observed that
approximately 0.5%c camber deformation is sufficient for vibration reduction and
performance enhancement.
The NACA 0012 airfoil is used as the baseline airfoil cross section in these studies.
The aerodynamic properties of the airfoil cross section with and without the prescribed
camber deformation are obtained using XFOIL analysis for a range of Mach numbers and
Reynolds numbers. These are used to generate the table lookup for calculating the static
183
stall residual for the ONERA dynamic stall model and for including the profile drag
correction in the aerodynamic forces. Variation of aerodynamic coefficients with the
angle of attack for M = 0.5 and Re = 1.41x106 is shown in Figure 6-6 for 1% camber
deformation. The lift (cl) and the drag (cd) coefficient obtained with cubic and quadratic
camber deformation shape function are close to each other below the stall condition while
the same is not true for coefficient of moment (cm). The difference between their
aerodynamic properties is more apparent in the post-stall regime and hence, they are
expected to provide different results in the dynamic stall condition.
The optimum camber deformation required for performance enhancement and
vibration reduction is obtained using a two-step optimization process that combines
surrogate-based optimization (SBO) with gradient based optimization (GBO) in order to
obtain a more stable optimization process and to reduce the computational time.
Preliminary analysis performed using the aeroelastic analysis code UM/NLABS-A
using both the camber deformation shape functions in dynamic stall condition are
described next.
6.3 Preliminary Numerical Results
The baseline model is a scaled BO105 rotor with four blades, as used in the HART II
experiments. Properties of the baseline rotor are summarized in Table 6-1 and more
detailed information can be found in [145]. All the cases considered here are at an
advance ratio of 0.33 and at a rotor thrust level of 6584 N (CT = 0.008).
184
0.1 1.5
-Quad
1 +Quad
0.05 -Cubic
0.5 +Cubic
Base
cm
0 0
cl
-Quad
+Quad -0.5
-0.05 -Cubic
+Cubic -1
Base
-0.1 -1.5
-15 -10 -5 0 5 10 15 -15 -10 -5 0 5 10 15
 (deg)  (deg)
0.2
-Quad
+Quad
0.15 -Cubic
+Cubic
Base
cd
0.1
0.05
0
-15 -10 -5 0 5 10 15
 (deg)
Figure 6-6: Aerodynamic Properties of Cambered and Baseline Airfoil Section
(Reference Airfoil: NACA 0012)
Table 6-1: Baseline Blade Properties
Property Value
Type Hingeless
Number of blades 4
Radius 2m
Root offset 0.44m
Chord 0.121m
Airfoil Section NACA 0012
Operating RPM 109 rad/s
Advance Ratio 0.33
Shaft angle -5 deg
In order to observe the effect of camber actuation on rotor performance in dynamic
stall condition, a uniform camber actuation force was applied along the blade span at
185
different actuation frequencies and for different phase of actuation. The resulting
distribution of camber deformation along the blade radius for both the camber
deformation shape function is shown in Figure 6-7 which indicates a maximum camber
deformation of 0.07%c at the blade tip. Since, in the current analysis, the same stiffness is
assumed for both the camber deformation modes, the resulting camber distribution along
the blade span is identical. Note that the camber actuation force is only applied from 0.23
< r/R < 1.00.
0.08
Camber Deformation (%c)
0.06
0.04
0.02
0
0 0.5 1 1.5 2
Blade Radius (R)
Figure 6-7: Variation of Camber Deformation along the Blade Span
6.3.1 Effect of Camber Actuation on MZ0
The effect of camber actuation on the mean value of the torque (M Z0) for cubic and
quadratic camber deformation for different actuation frequencies is shown in Figure 6-8.
Results show that the 2/rev actuation frequency is the most effective while the 5/rev
actuation frequency is the least effective frequency for influencing performance of the
rotor blade, for both the camber deformation shape functions considered here.
Preliminary observation show that it is possible to obtain atleast 2% improvement in M Z0
by quadratic camber actuation at 2/rev frequency and phase difference of 240 degree or
186
by cubic camber actuation at 2/rev frequency and phase difference of 60 degree. Even
though the minimum value of MZ0 that can be obtained from camber actuation is similar
for the both the camber deformation shape functions, the adverse effect of camber
actuation in increasing MZ0 is higher for the quadratic camber deformation.
980 980
5/rev
970 4/rev 970
3/rev
960 2/rev 960
Mz (Nm)
Mz (Nm)
Baseline
950 950
940 940
930 930
920 920
0 100 200 300 0 100 200 300
Phase of Actuation (deg) Phase of Actuation (Deg)
(a) With Ψquad (b) With Ψcubic
Figure 6-8: Effect of Camber Actuation on MZ0 at µ = 0.33
6.3.2 Circle Plot for Different Actuation Frequencies
As described in the earlier chapters, circle plots are used to determine if the active
vibration control method has sufficient authority for vibration reduction. In this case,
camber actuation is provided at different actuation frequencies and the phase of actuation
at each frequency is varied from 0 degree to 360 degree in intervals of 30 degree.
The circle plots obtained for the fixed system hub load at 4/rev frequency for Fz and
Mx component are shown in Figure 6-9 and Figure 6-10, respectively. For each of the
camber actuation cases, the camber deformation along the blade span is as shown in
Figure 6-8. Circle plot results show that both 3/rev and 4/rev actuation frequencies are the
most effective for reducing 4/rev vibration in the vertical component of the hub force
187
(Fz4) for both the camber deformation shape functions. Similarly, 3/rev actuation
frequency is the most effective for reducing MX4 for both the camber deformation shape
functions. These results show that the frequency of actuation that is most effective for
reducing 4/rev vibratory hub loads depends upon the hub load component. Hence, in the
analysis performed in this chapter, the actuation signal consists of 2/rev to 5/rev actuation
frequencies.
5/rev
150 150
4/rev
3/rev
100 2/rev 100
FZ4,sin(N)
FZ4,sin(N)
Baseline
50 50
0 0
-50 -50
-200 -150 -100 -50 0 50 -200 -150 -100 -50 0 50
F (N) F (N)
Z4,cos Z4,cos

Figure 6-9: Circle Plot for FZ4
90 90
5/rev
80 4/rev 80
3/rev
70 2/rev 70
MX4,sin(Nm)
MX4,sin(Nm)
Baseline
60 60
50 50
40 40
30 30
-40 -20 0 20 40 60 -40 -20 0 20 40 60
M (Nm) M (Nm)
X4,cos X4,cos

Figure 6-10: Circle Plot for MX4
188
6.3.3 Effect of Amplitude of Actuation
In order to see the effect of amplitude of actuation, the amplitude of actuation for both
the camber deformation shape function was doubled and the analysis was performed for
3/rev actuation frequency. The effect of amplitude on circle plot for F Z4 is shown in
Figure 6-11. The results show an increase in size of the circle plots for both the camber
deformation cases with increase in the amplitude of actuation indicating an increase in
the authority for vibration reduction. However, the results shown in Figure 6-12 for MZ0
show that the effect of camber actuation on performance reduces with an increase in the
amplitude of actuation for 3/rev actuation frequency.
300 300
Amp=A Amp=A
Amp=2A Amp=2A
200 Baseline 200 Baseline
Origin Origin
FZ4,sin(N)
100
FZ4,sin(N)
100
0 0
-100 -100
-200 -200
-300 -200 -100 0 100 -300 -200 -100 0 100
FZ4,cos(N) FZ4,cos(N)
Figure 6-11: Effect of Amplitude on Circle Plot for FZ4
189
990 990
Amp=A Amp=A
980 Amp=2A 980 Amp=2A
970 Baseline 970 Baseline
960 960
M z (Nm)
M z (Nm)
950 950
940 940
930 930
920 920
910 910
0 100 200 300 0 100 200 300
Phase of Actuation (deg) Phase of Actuation (deg)

Figure 6-12: Effect of Amplitude of Actuation on MZ0
6.4 Optimization Problem Definition
The problem of vibration reduction and performance enhancement has been converted
to an optimization problem in this study. The design variables used in the study are the
amplitude and phase of actuation corresponding to actuation frequencies of 2/rev to 5/rev.
Hence, this problem has eight design variables. The objective functions considered in this
study are:
a) Minimize FZ4 (minimize vibration in vertical component of hub force)
b) Minimize FH4 (minimize vibration in all the hub load components)
1
where, FH 4  FX 4 2  FY 4 2  FZ 4 2  M X 42  M Y 42  M Z 42
R
c) Minimize MZ0 (improve performance)
The camber deformation force is assumed to be constant along the blade span
which results in the camber deformation profile as described in Figure 6-7. Without loss
190
of generality, the uniform actuation force required to obtain a camber deformation of
0.07% chord at the blade tip is referred to as 1A. The initial range for camber amplitude
was chosen between no actuation and 2A for each actuation frequency, whereas the phase
of camber actuation varies from 0 to 2π for each actuation frequency, as shown in Table
6-2. Same range was used for the optimization conducted with Ψquad and Ψcubic shape
function.
Table 6-2: Range for Design Variable used in the Optimization Study
Amplitude (A) Phase (radians)

2/rev 3/rev 4/rev 5/rev 2/rev 3/rev 4/rev 5/rev
A2 A3 A4 A5 Φ2 Φ3 Φ4 Φ5
Lower Bound 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Upper Bound 2.00 2.00 2.00 2.00 2π 2π 2π 2π
Thus, a generic camber deformation signal provided to the rotor blade can be
expressed as:
2   
 Blade1  A2 sin(20t  )  A3 sin(30t  3 )  A4 sin(40t  4 )  A5 sin(50t  5 )
2 3 4 5
where, ω0 is the 1/rev frequency in radians.
6.4.1 Optimization Results
The optimization results obtained at the end of two-step optimization process using
Ψquad and Ψcubic shape functions are listed in Table 6-3. Results obtained show a 99.6%
reduction in FZ4 vibratory load with Ψcubic shape function, while 97.6% vibration
reduction can be obtained using Ψquad shape function. Similarly, the percentage
improvement in combined vibratory load (FH4) and performance (MZ0) is higher in the
case of analysis with Ψcubic shape function.
191
Table 6-3: Optimization Results
Objective Percentage Reduction

Function With Ψcubic With Ψquad
FZ4 (%) 99.60 97.60
MZ0 (%) 3.70 3.30
FH4 (%) 51.6 45.20
2.00 2.00
1.75 min FZ4 1.75
min MZ0
1.50 1.50
min FH4
Amplitude (a)
Amplitude (a)
1.25 1.25
1.00 1.00
0.75 0.75
0.50 0.50
0.25 0.25
0.00 0.00
Actuation Frequency Actuation Frequency
Figure 6-13: Amplitude of Actuation for Optimized Cases
The value of design variables for these optimized cases is shown in Figure 6-13. The
trend observed for the amplitude of camber deformation corresponding to different
actuation frequencies for the optimized cases is similar for the case with Ψ quad and Ψcubic
shape functions. In both the cases, the amplitude corresponding to 2/rev actuation
frequency is highest for the “min MZ0 case”. This is consistent with the trend observed in
Figure 6-8 where 2/rev actuation frequency showed maximum influence on MZ0. The
contribution from higher actuation frequencies for minimizing MZ0 is smaller for both the
camber deformation shape functions. For the vibration reduction cases, 2/rev actuation
frequency is the least effective while 3/rev and 4/rev are the most effective frequencies.
In general, the amplitude of camber deformation corresponding to the optimized cases
192
with Ψcubic shape function is higher than that required for the optimized cases with Ψquad
shape function.
2.00 2.00
1.75 1.75 min FZ4
Phase of Actuation ( π rad)
Phase of A ctuation ( π rad)

1.50 1.50 min MZ0
1.25 1.25 min FH4
1.00 1.00
0.75 0.75
0.50 0.50
0.25 0.25
0.00 0.00
Actuation Frequency Actuation Frequency
Figure 6-14: Phase of Actuation for Optimized Cases
The variation for phase of actuation corresponding to different actuation frequencies
for the optimized cases is shown Figure 6-14 .
6.4.2 Advantage of Two-step Optimization Process
The two-step optimization process used in this paper provides robust solution and is
very well suited for exploring a large design space where multiple solutions exist. As
discussed in the earlier section, initially the non-gradient optimization is performed using
surrogate method and the results obtained from the surrogate based optimization (SBO)
are used as the starting point for gradient based optimization (GBO).
The variation of objective function with iteration number for the optimizations
performed using SBO is show in Figure 6-15. Note that an additional iteration was
193
performed for the Ψcubic case in order to obtain the pareto front described in Section 6.4.5.
Results indicate that design points very close to the optimum value are obtained in first 2-
3 iterations.
% Reduction in M Z0
% Reduction in F Z4
100 4
3.5
90
Cubic 3
Quad
80 2.5
0 2 4 6 8 0 2 4 6 8
Figure 6-15: Variation of Objective Function with Iteration Number for SBO
2
Amplitude (A)
1
2/rev
3/rev
0 4/rev7
1 2 3 4 5 6
SBO Iteration 5/rev
400
 (deg)
200
0
1 2 3 4 5 6 7
SBO Iteration
Figure 6-16: Variation of Design Variables with Iteration Number for SBO for Min
MZ0 case for analysis with Ψcubic
The variation of design variables with SBO iteration for Min Mz0 case for analysis
performed using Ψcubic is shown in Figure 6-16. The trend observed highlight that 2 and
3/rev actuation frequencies are the most effective for performance enhancement. For
these frequencies, the variation in amplitude and phase of actuation is small. These
194
results also show that the 5/rev actuation frequency is least effective for improving the
performance of rotor blade, as observed earlier in the parametric studies.
For each of the objective functions considered, three best points obtained from the
surrogate optimization are selected as the starting points for gradient based optimization
as shown in Figure 6-17. Here, only the results for the “min FZ4” case corresponding to
Ψcubic shape function are shown; but a similar trend is observed for all the objective
functions considered. In Figure 6-17, the bottom part of the column (in blue) represents
the vibration reduction obtained just from the SBO, while the top part (in red) represents
the additional improvement in the objective function due to the GBO. Results presented
here show that, although the initial starting points obtained from SBO provide different
level of vibration reduction, the final vibration reduction obtained at the end of GBO
process for each of the three cases considered here are close to each other.
Surrogate Result fmincon Result

100.0
% Vibration Reduction in FZ4
80.0
60.0
40.0
20.0
0.0
Sol 1 Sol 2 Sol 3
Optimized Solution
Figure 6-17: Percentage Reduction in FZ4 using Two-Step Optimization Process
195
The variation of design variables for these three cases is shown in Table 6-4. For each
of the cases, the final solution obtained at the end of GBO is close to the corresponding
starting point obtained from SBO. Also, the final value of design variables obtained for
the three cases are very far from each other, even though the value of objective function
for these cases are close to each other. This highlights the fact that multiple local minima
exist in the design space which can provide a similar level of vibration reduction.
Table 6-4: Optimum Design Variables for min FZ4 Case
Amplitude (A) Phase (rad) FZ4

2/rev 3/rev 4/rev 5/rev 2/rev 3/rev 4/rev 5/rev % Change
SBO Sol 1 0.11 1.76 0.90 0.82 2.93 1.56 3.64 4.62 -96.5
GBO Sol 0.15 1.76 0.88 0.84 3.00 1.53 3.58 4.63 -99.6
SBO Sol 2 1.34 0.16 0.59 1.14 1.60 5.11 0.07 6.23 -93.0
GBO Sol 1.36 0.17 0.61 1.11 1.57 5.14 0.07 6.17 -99.4
SBO Sol 3 1.30 1.35 0.12 0.00 1.74 1.80 2.86 3.90 -89.6
GBO Sol 1.32 1.31 0.12 0.06 1.74 1.74 2.96 3.90 -98.4
6.4.3 Effect of Optimization on other Hub Load Components
In this section, the effect of vibration reduction and performance enhancement on the
other hub loads is examined. The mean value of torque and the dynamic amplitude of
various forces at the hub in fixed frame corresponding to 4/rev frequency and 8/rev
frequency for the baseline case are shown in Table 6-5(a).
Table 6-5: Variation in Hub Loads for the Optimized Cases

(a) Absolute Value of the Hub Loads for the Baseline Case
MZ0 FX4 FY4 FZ4 MX4 MY4 MZ4 FZ8 FH4 FH8
(Nm) (N) (N) (N) (Nm) (Nm) (Nm) (N) (N) (N)
Baseline -939.9 312.8 160.7 91.7 58.4 34.2 6.8 51.7 397.4 62.8
196
(b) Percentage Changes in Hub Loads with Ψcubic Shape Function
% Changes MZ0 FX4 FY4 FZ4 MX4 MY4 MZ4 FZ8 FH4 FH8
min FZ4 Case 1.1 -13.0 -50.4 -99.6 -30.7 92.7 -71.5 -7.6 -18.9 -7.4
min MZ0 Case -3.7 1.6 51.9 34.9 18.2 -34.7 34.0 -75.5 14.7 -58.6
min FH4 Case 0.2 -57.5 -39.2 -62.7 -78.3 33.8 -42.6 1.7 -51.6 -5.8
(c) Percentage Changes in Hub Loads with Ψquad Shape Function
% Changes MZ0 FX4 FY4 FZ4 MX4 MY4 MZ4 FZ8 FH4 FH8
min FZ4 2.2 -19.7 -49.1 -97.6 -22.7 83.5 -76.4 6.5 -23.8 2.4
min MZ0 -3.3 7.6 57.3 -3.8 14.4 -37.3 11.7 -84.8 17.1 -66.8
min FH4 1.0 -32.3 -77.0 -38.7 -64.3 105.7 -55.4 -27.4 -45.2 -26.2
The percentage changes in different component of hub loads for the optimized cases
obtained with Ψquad and Ψcubic shape function is shown in Table 6-5(b) and Table 6-5(c),
respectively. The result obtained for the “min FZ4” case in Table 6-5(b) shows that the
vibration reduction in FZ4 is accompanied by a 1.1% increase in torque. This case also
results in 18.9% reduction for FH4, however it is smaller than the optimum reduction that
can be obtained when FH4 is minimized. For the “min FZ4” case, minimizing the
amplitude at 4/rev frequency also results in small decrease in vibratory loads at 8/rev
frequency. Similarly, the 3.7% improvement in performance for the “min MZ0” case is
accompanied by a 34.9% increase in FZ4 and 14.7% increase in FH4. Surprisingly,
minimizing MZ0 also results in significant vibration reduction for 8/rev frequency loads.
In the “min FH4” case, the optimizer tries to reduce the amplitude of vibration at 4/rev
frequency for all the hub load components. Even though a small increase in vibration is
observed for MY4, the absolute values listed in Table 6-5(a) shows that the magnitude for
this component is small as compared to other hub load components. A similar trend is
observed in the results obtained with Ψquad camber deformation shape function, as shown
in Table 6-5(c).
197
6.4.4 Analysis of Optimized Cases
In this section, the optimization results are analyzed further in order to understand the
mechanism through which vibration reduction and improvement in performance is
obtained using camber deformation. Here, only the optimized results obtained from Ψcubic
shape function are shown.
Figure 6-18: Variation of Angle of Attack for the Baseline Case (Units: Deg)
The variation of angle of attack over the rotor disk for the baseline case (µ = 0.33, C T
= 0.008, no camber actuation) is shown in Figure 6-18. The distribution of angle of attack
shown here highlights small angle of attack observed on the advancing side (almost
negative near the tip) and large angle of attack on the retreating side which result in
dynamic stall. (Note that on the retreating side, near the root region, there is a region of
reverse flow and a region with very large negative angle of attack. The angle of attack in
these two region is well below -5deg, however, the lower limit for the “colorbar” used in
this figure has been set to -5 to focus more on the region with large positive angle of
attack.)
198
min FZ4 min MZ0 min FH4
3
1 1 1
2
0.5 0.5 0.5
1
0 0 0 0
-1
-0.5 -0.5 -0.5
-2
-1 -1 -1
-3
-0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5
Figure 6-19: Difference in Angle of Attack for the Optimized Cases (Unit: Deg)
The angle of attack variation obtained for the optimized cases is subtracted from that
observed for the baseline case and the results obtained are shown in Figure 6-19. Note
that the reverse flow region and region with large negative angle of attack is removed
from the figure to highlight the dynamic stall region. The highest variation in angle of
attack from the baseline case is observed for the “min MZ0” case, where the optimizer is
trying to reduce large angle of attack encountered on the retreating side (shown by dark
blue region). For the “min FZ4” and “min FH4” cases, the decrease in vibration is obtained
by manipulating the phase of the additional loads obtained from camber deformation.
Thus, in these cases, the variation in angle of attack from the baseline case is not
significant as it was observed in the “min MZ0” case. The same trend can also be seen in
Figure 6-20 which shows the angle of attack variation with azimuth angle at r = 0.74R.
199
At r = 0.74R
15
min FZ4
min MZ0
min FH4
Angle of Attack (deg)

10 Baseline
0
0 100 200 300
Azimuth Angle (Deg)
Figure 6-20: Variation of Angle of Attack at r = 0.74R
The camber deformation profile for the optimized cases over the rotor disk is shown in
Figure 6-21 while Figure 6-22 shows the camber deformation at the blade tip. Results
show that the camber deformation required is highest for the “min FH4” case. For all the
cases, a maximum camber deformation of 0.35%c is sufficient to obtain the objective
function listed in Table 6-3.
min FZ4 min MZ0 min FH4
1 0.3 1 1
0.5 0.2 0.5 0.5

0.1
0 0 0
0
-0.5 -0.5 -0.5
-0.1
-1 -0.2 -1 -1
-0.5 0 0.5 -0.5 0 0.5 -0.5 0 0.5
Figure 6-21: Variation of Camber Deformation for the Optimized Cases (Unit: %c)
200
Camber Deformation at Blade Tip
0.4
min F
Z4
0.3 min MZ0
min FH4
0.2
Camber Deformation (%c)

0.1
-0.1
-0.2
-0.3
-0.4
0 50 100 150 200 250 300 350
Azimuth Angle (Deg)
Figure 6-22: Camber Deformation at the Blade Tip
6.4.5 Pareto Optimization
In this study, a multi-objective optimization was conducted using the Ψcubic shape
function where vibration reduction (min FH4) and performance enhancement (min MZ0)
were considered as the objective functions. Optimization was performed using the
GODLIKE toolbox developed for unconstrained optimization in MATLAB. The Pareto
front obtained for these two objective functions is shown in Figure 6-23. The results
presented here show that it is possible to obtain simultaneous reduction in vibration and
improvement in performance with camber deformation.
201
1
Pareto Points
min FH4 case
0 min MZ0 case
% Change in MZ0
-1
-2
-3
-4
-60 -40 -20 0 20 40
% Change in FH4
Figure 6-23: Pareto Front for Vibration Reduction and Performance Enhancement
In this chapter, the use of quadratic and cubic camber deformation shape functions for
vibration reduction and performance enhancement in dynamic stall condition was
studied. A modified version of the ONERA dynamic stall model which can account for
morphing airfoil section was included in the aeroelastic analysis performed using
UM/NLABS-A. Preliminary parametric studies at different actuation frequencies showed
the capability of camber deformation in influencing both vibratory loads and performance
of the rotor blade.
In the next step, a novel two-step optimization process was used to obtain reduction in
vibratory loads and hub torque (as a metric of rotor performance). In the first step of the
process, a global search is performed using surrogate modeling to provide a good feasible
initial design for the second step in the process: gradient-based optimization. The use of
202
gradient-based optimization allows the objective function to converge to a minimum,
from the initial designs provided by the surrogate approach. Thus, a more stable
optimization process is achieved, thereby reducing the overall required computational
time.
In the optimization studies, the amplitude and phase of camber actuation at 2/rev,
3/rev, 4/rev and 5/rev actuation frequencies were used as the design variables and the
analysis was performed with both cubic and quadratic camber deformation shape
function. The optimization results showed:
a) 99.6% reduction in vertical component of 4/rev vibratory load at the hub (FZ4)
b) 3.7% improvement in performance (MZ0)
c) 51.6% reduction in combined 4/rev vibratory load the hub (FH4)
with cubic camber deformation shape function. The results obtained for these objective
functions with quadratic shape function were slightly less as compared to those obtained
with the cubic shape function. Post-processing of the optimized results obtained showed
that the performance improvement is obtained by reducing the angle of attack in the
dynamic stall region while vibration reduction is obtained by adjusting the phase of
camber actuation in such a way that the vibratory loads at 4/rev frequency at the hub are
minimized. Finally, a multi-objective optimization study was performed where vibration
reduction and performance enhancement was obtained simultaneously and a Pareto front
was determined for these objective functions.
203
Chapter 7. Conclusions and Future Work
This chapter provides a summary of the work presented in this thesis and outlines the
key results and contribution made. And finally, few recommendations are made for the
future work.
7.1 Summary
The aim of the work presented in this thesis was to develop a multidisciplinary
analysis and design framework and exercise it to explore various approaches available for
vibration reduction.
As a first step, an aeroelastic analysis/design environment was developed which can
be used to design composite rotor blades with vibration reduction or performance
enhancement as the objective function. The design environment included several well-
established analysis codes from different sources: IXGEN for mesh generation,
UM/VABS for cross sectional analysis, RCAS for rotorcraft simulation and ModelCenter
for optimization and parametric studies. This design environment enables
conceptual/early preliminary multidisciplinary rotor blade design, allowing rapid design
trade studies early in the design process with realistic structural properties for modern
composite rotor blades. The design environment was successfully used to perform
detailed parametric and optimization studies on a full scale model of a UH-60 composite
rotor blade.
204
For design of rotor blade with active twist, a new design strategy and framework was
developed where dynamic twist was maximized instead of maximizing the static twist.
The optimization framework included the aeroelastic design environment described
earlier along with surrogate based optimization technique. The surrogate based
optimization is performed in combination with Efficient Global Optimization algorithm
which is better suited for aeroelastic problems where the runtime for each iteration is very
high and there are significant amount of failed cases due to convergence issues within the
analysis cycle. Results showed that the amplitude of dynamic twist is a true indicator of
control authority of active twist rotor for vibration reduction. Optimization was
performed at different actuation frequencies to maximize the control authority for
vibration reduction at a range of frequencies. For the optimization studies presented here,
the NASA/Army/MIT active twist rotor, which had been tested in TDT, was selected as
the baseline rotor blade.
In the next step, the optimization framework was extended to include discrete design
variables in the optimization and the solution for mixed design variable problems was
obtained using different techniques. In this extended framework, the genetic optimization
algorithm was combined with the gradient based optimization to obtain an optimum
design with continuous design variables and an optimum design with mixed design
variables in an efficient manner. The results obtained highlighted the effect of
discretizing design variables and helped in obtaining a realistic composite rotor blade
design.
Although active flaps have been around for last two decades, very few studies in the
literature have focused on the detailed design of a composite rotor blade with active flaps.
205
In this research, composite cross sections along the blade span for a Mach-scaled rotor
blade were designed using the mixed-design variable optimization framework described
above.
This thesis also includes (in Appendix) work related to design and fabrication of a
composite rotor blade with dual active flaps which can be tested in a Mach-scaled spin
test stand. The work done highlights the steps involved in the design process and
discusses difficulties and issues encountered during the testing phase. At the end, possible
corrections for the issues are presented and modifications which can be made for future
tests are listed.
Finally, within the same framework introduced here but with a different analysis tool,
the use of quadratic and cubic camber deformation shape function for vibration reduction
and performance enhancement in dynamic stall region was studied. A modified version
of the ONERA dynamic stall model which can account for morphing airfoil section was
included in the aeroelastic analysis performed using UM/NLABS-A. Optimization results
obtained shows 50% reduction in 4/rev vibratory loads at the hub and more than 3.5%
improvement in the performance using camber actuation at advance ratio of 0.33.
7.2 Main Results
This section summarizes the main results obtained in the thesis. These conclusions
support the contributions made in this thesis which are listed in Section 7.3.
The new rotor blade design environment was used to design a composite cross section
for a full scale model of a UH-60 rotor blade. The optimization results showed:
206
c) 52% vibration reduction in FZ4 (Objective function: min FZ4);
d) 28% vibration reduction in FH4 (Objective function: min FH4);
where FZ4 is the amplitude of the 4/rev vibratory vertical force at the hub and FH4 is the
amplitude of the combined 4/rev vibratory load at the hub. The results obtained indicated
that the reduction is FZ4 is obtained by reducing the coupling between the structural
modes while a decrease in FH4 can be obtained by increasing the torsional stiffness of the
cross section.
With an optimization framework/strategy developed for the design of active composite
rotor blades, the new framework was set to use the amplitude of dynamic twist as the
objective function. Using the NASA/Army/MIT Active Twist Rotor design as the
baseline rotor blade, the optimized designs showed:
a) 18.5% increase in static twist per unit length;
b) 63.6% increase in tip twist amplitude for 4/rev twist actuation;
c) 52.6% increase in tip twist amplitude for 3/rev twist actuation;
d) 101% increase in tip twist amplitude for 5/rev twist actuation;
e) 71% increase in twist amplitude for actuation at a range of frequencies (3, 4 and
5/rev).
without varying ply thicknesses and ply angles.
Further it was shown that the blade designs obtained by maximizing the amplitude of
dynamic twist have higher authority for vibration reduction as compared to the blade
design obtained by maximizing the static twist amplitude. Based on the optimization
studies conducted, important factors identified for maximizing the dynamic twist are: a)
207
first torsional frequency of the rotor blade, b) active moment generated by active
material, and c) aerodynamic loads acting on the rotor blade.
An augmented version of this optimization framework was used to design active twist
rotor blades with both continuous and discrete design variables. In this analysis, ply
thicknesses and ply angles were treated as discrete design variables. The optimization
studies with ply thicknesses showed:
a) A thick prepreg layer is required near the leading edge (Ply 2a) to obtain higher
active twisting moment;
b) Increasing the chordwise coverage of active plies is better than increasing the
thickness of active plies, in order to get higher static and dynamic twist;
c) A boxed-shaped spar design, in which the chordwise location where the spar plies
end and the chordwise location of vertical spar web are close to each other, is
suitable for maximizing the dynamic twist amplitude;
d) The two ballast masses are very useful for tuning the dynamic properties of the
rotor blade, which eventually results in an increase in the dynamic twist
amplitude.
Besides this, the framework was useful for quantifying the effects of discretizing
design variables for different objective functions considered.
The same analysis framework was also used to design composite blade with active
flaps after modifying the aeroelastic analysis performed by RCAS. In this analysis,
optimum rotor blade cross sections along the blade span were determined for different
spanwise locations of the flap. The analysis included the effect of actuator and flap
inertia. For these studies, the amplitude of dynamic twist at 4/rev flap actuation frequency
208
was used as the objective function. A 5-ft radius Mach-scaled composite rotor blade that
can be tested in the University of Michigan spin test stand was selected as the baseline
rotor blade. The optimization results showed:
d) 14% increase in dynamic twist for a 12% flap located at 0.78R;
e) 9.32% increase in dynamic twist for a 12% flap located at 0.7R;
f) 4.58% increase in dynamic twist for a 12% flap located at 0.85R.
All the optimized blade designs showed a decrease in the torsional stiffness and an
increase in the axial stiffness.
The optimization studies conducted showed that a torsionally stiff blade is desired in
order to obtain higher active twist whereas a torsionally soft blade is desired in order to
obtain higher control authority from flap actuation.
For the analysis of a rotor blade with camber actuation in forward flight condition, a
modified version of the ONERA dynamic stall model was included in UM/NLABS-A. In
the optimization studies with camber actuation, the amplitude and phase of camber
actuation at 2/rev, 3/rev, 4/rev and 5/rev actuation frequencies were used as the design
variables and the analysis was performed with both cubic and quadratic camber
deformation shape function. The optimization results showed:
d) 99.6% reduction in vertical component of 4/rev vibratory load at the hub (FZ4);
e) 3.7% improvement in performance (MZ0);
f) 51.6% reduction in combined 4/rev vibratory load the hub (FH4);
with cubic camber deformation shape function. Post-processing of the optimized results
obtained showed that the performance improvement is obtained by reducing the angle of
attack in the dynamic stall region while vibration reduction is obtaining by adjusting the
209
phase of camber actuation in such a way that the vibratory loads at 4/rev frequency at the
hub are minimized.
7.3 Key Contribution
The key contributions made in the thesis are:
 Created a new framework for the design and analysis of composite rotor blades
with and without on-blade devices for vibration reduction. This new framework
enables the designer to optimally size (at the ply level) realistic composite rotor
blades. Among the active ones, this dissertation studied: active twist rotors, active
camber scheduling, and multiple flaps for lower vibration, higher performance
solutions.
 Introduced dynamic twist as the objective function to be used when designing
active twist rotor blades. Through examples, it was shown that the dynamic twist
is a true indicator of control authority for vibration reduction and not the static
twist as done in the past.
a. Extended the dynamic twist beyond a single actuation frequency and
introduced an objective function to capture a range of frequencies when
designing a composite rotor blade.
b. Established that the optimization studies can be performed in hover
condition (instead of multiple advance ratios) using periodic analysis
(instead of full trim analysis) within a design cycle in order to reduce the
computation time.
210
 Developed an optimization strategy for the design of composite rotor blade with
mixed design variables. The framework provides three different techniques to
ensure that a global optimum solution is obtained.
 Extended the optimization strategy and framework to also include the design of
composite rotor blades with multiple flaps.
 Modified the aeromechanics solution of the new framework so to use
UM/NLABS-A for the analysis of composite rotor blades with active camber
deformation.
a. Introduced the ONERA dynamic stall model in UM/NLABS-A so to
capture the performance and vibration effects associated with dynamic
stall.
b. Showed effectiveness of camber deformation as a mean to improve
performance and reduce vibration. The numerical studies conducted in
dynamic stall conditions at μ=0.33 showed that the cubic camber
deformation is more effective than the quadratic one.
 Designed, fabricated and tested the first composite rotor blade with dual flaps. It
was a 1/6th Mach-scaled CH-47D blade for testing in the UM spin stand.
7.4 Future Work
Based on the research conducted in this thesis, the following areas have been
identified for further studies:
a) Failure Analysis: The blade failure approach used in this research was based on
the loads observed in hover condition (and hence lower values were used for
maximum allowable strain in the optimization studies to correct for that). This
211
could be modified by directly including the worst-case loadings determined for a
given design and different advance ratios within an optimization loop. The
framework can also be extended to include fatigue analysis based on the dynamic
loads acting on the blade.
b) Improvement in Surrogate Modeling: The surrogate modeling techniques used
needs to be improved further, especially when a large number of design variables
are involved to reduce the error in the prediction. Improved accuracy in the
prediction of the response function will also reduce the number of iterations
required to obtain the optimum solution.
c) Including Closed Loop Controller: In this thesis, the design of composite cross
section to enhance the vibration reduction capabilities of active rotor blade at
different operating conditions was performed. For a direct evaluation of vibration
performance under actuation, a control strategy needs to be introduced. The
framework is, in principle, capable of evaluating a given controller. But the
design of such may require further reduced order modeling coming from the
current approach. This is a rich area of research that should be pursued in the
future.
d) Effect of Advance Ratio: All the optimization studies for the design of composite
rotor blade were performed in hover condition. Although it was shown that the
optimum design obtained in hover condition is close to the optimum in forward
flight, this should be formally demonstrated and its limitations established. Also,
propulsive trim needs to be considered instead of the wind tunnel trim used in this
thesis for forward flight conditions.
212
e) Effect of Actuator-flap Dynamics: In the optimization studies performed in this
thesis, only the inertia effect of actuator and flap was included in the aeroelastic
analysis. However, the dynamics of the combined flap-actuation mechanism also
should be modeled to completely capture its effects when assessing the
effectiveness of the flap system for vibration reduction. The dynamic properties of
the flap actuation mechanism can be determined by performing experimental
analysis on the bench as described in Appendix C.
f) Non-Harmonic Camber Actuation: To obtain performance enhancement, it was
observed that camber actuation is required only in the retreating side of the blade.
Thus, non-harmonic periodic actuation (instead of harmonic) should be
investigated for performance enhancement in dynamic stall conditions.
213
APPENDICES
214
Appendix A. Surrogate Based Optimization
This section provides mathematical expressions related to the development of
surrogate models and Efficient Global Optimization (EGO) algorithm used in the thesis.
Further details about these concepts can be found in [96, 172].
A.1 Construction of Surrogate
The purpose of creating a surrogate model is to map a function y=y(x) to a black-box
function which can be evaluated very quickly. The generic solution method is to collect
output values y(1), y(2), …., y(n) that result from a set of inputs x(1), x(2), ….. x(n) and find a
best guess yˆ ( x) for the mapping, based on these known observations. The set of points x
are selected from the chosen range of design variables using Latin Hypercube Sampling
(LHS) technique. In this method, the points are selected in such a way that the distance
between these points in the design space is maximized.
Kriging is based on the fundamental assumption that the errors involved in the
prediction yˆ ( x) are correlated. This implies that error obtained at two close points
together will be close. In kriging method, the unknown function yˆ ( x) is assumed to be of
the form:
yˆ ( x) = f(x) + Z(x)
where, f(x) is an assumed function (usually a polynomial) and Z(x) is a realization of a
stochastic process. The function f(x) can be thought of as a global approximation of y(x),
215
while Z(x) accounts for local deviation which ensures that the kriging model interpolates
the data points exactly. The function Z(x) is assumed to follow a distribution (Gaussian or
normal distribution) with zero mean value and variance of σ2var . The covariance matrix
of Z(x), which is a measure of how strongly correlated two points are, is given by:
Cov  Z  x (i)  , Z  x ( j)    2 var R krg
where each element of the correlation matrix Rkrg is given by:
R  krg ij  R krg  x (i) , x ( j) 
and R krg  x (i) , x ( j)  is a correlation function which accounts for the effect of each
interpolation point on every other interpolation point.
Correlation Models
The DACE Toolbox [163] used for developing the surrogate models in this thesis
provides following options for correlation function:
1) Gaussian Function
R krg  x (i) , x ( j) ,   exp( | x (i) - x ( j) |2 )
2) Exponential Function
R krg  x (i) , x ( j) ,   exp( | x (i) - x ( j) |)
3) Spline Function
1  15 2  30 3 for 0    0.2



R krg  x (i) , x ( j) ,    1.25(1   )3 for 0.2    0.1

 0 for   1

where,    | x (i) - x ( j) | .
216
4) Linear Function
R krg  x (i) , x ( j) ,   max(0,1   | x (i) - x ( j) |)
The variation of correlation function for different values of parameter θ is given in Figure
A-1.
Exp Gauss
1 1
Rkrg
Rkrg
0.5 0.5
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
(i) (j) (i) (j)
|x -x | |x -x |
Lin Spline
1 1
 = 0.25
=1
=5
Rkrg
Rkrg
0.5 0.5
0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
(i) (j) (i) (j)
|x -x | |x -x |
Figure A-1: Correlation Functions
The fitting parameter θ is an unknown correlation parameter which needs to be
determined. The value of θ in turn depends on the form of f(x) chosen for the surrogate
model.
Regression Models
In order to predict the value of f(x), a regression model is used which is a linear
combination of p chosen functions fi . Thus, f(x) can be written as:
f  x   1 f1  x   2 f 2  x ..   p f p  x  = Fβ
where, F=[ f1(x(1)) f2(x(1)) …… fp(x(1)) ]n x p.
217
The coefficients β are called regression parameters. The toolbox provides regression
models with polynomials of the order 0, 1 and 2. The value of fi (x) for each of these
cases is given by:
a) Order 0, p = 1:
f1 ( x)  1
b) Order 1, p = n+1:
f1 ( x)  1, f 2 ( x)  x1 , ......., f n1 ( x)  xn
1
c) Order 2, p  (n  1)(n  2)
2
f1 ( x)  1
f 2 ( x)  x1 , f3 ( x)  x2 , ......., f n1 ( x)  xn
f n2 ( x)  x12 , f n3 ( x)  x1x2 , ......., f n1 ( x)  x1xn
………………………………… f p ( x)  xn xn
For each response, the correlation function and the regression model which gave the
minimum error in the prediction of response function was selected for developing the
surrogate model.
Estimating Kriging Parameter θ
The value of kriging parameter θ is determined by using the likelihood (L) estimates.
The likelihood function is a measure of probability of the sample data being drawn from
a probability density function associated with a Gaussian process. The maximum
likelihood estimate of θ represents the “best guess” for fitting parameter. Although, any
value of parameter θ would result in a surrogate that interpolates the sample point
exactly, the “best” kriging surrogate is obtained by maximizing the likelihood function.
218
The log of likelihood function (also known as concentrated ln-likelihood function in
literature) is given by:
 n ln(ˆ 2var )  ln | Rkrg |

ln( L)  
2
where , ̂ 2var is the generalized least square estimate of  2 var and is given by :
( y  F ˆ )T ( Rkrg ) 1 ( y  F ˆ )
ˆ 2var 
n
and ˆ is given by:
ˆ  ( F T ( Rkrg )1 F )1 F T ( Rkrg )1 y
The auxiliary optimization process of determining optimum θ can result in significant
fitting time depending upon the size of the system. During this optimization process,
scaling of the design space from 0 to 1 is very useful to ensure that the value of θ does
not vary significantly for different design variables. Hence, kriging is only appropriate
when the time needed to generate the interpolation points is much larger than the time to
interpolate the data, which is true for all the aeroelastic analyses performed in this thesis.
This auxiliary optimization for determining the parameter θ is performed using the
MATLAB function “UMDIRECT” described in [173].
When all the parameters are known, the kriging approximation to a function y(x), for
any order regression function, can be written as:
yˆ krg  F ˆ  rkrg ( x)T ( Rkrg )1 ( y  F ˆ )
where,
rkrg  [ Rkrg ( x, x(1) ), Rkrg ( x, x(2) ),........., Rkrg ( x, x( n ) )]T
219
The column vector rkrg of length n is the correlation vector between an arbitrary point x
and the interpolation points, x(1), ……, x(n) .
Error Estimate
The mean square error (MSE), at any point in the design space, of the kriging
predictor can be written as:
 1
(1  1T Rkrg rkrg )2 
s ( x)  ˆ
2 2
var
1
1  rkrg Rkrg rkrg 
T
1 
 1T Rkrg 1 
The error described above is a measure of uncertainty in the prediction.
A.2 EGO Algorithm
After the surrogate models are obtained for objective function and constraints,
optimization analysis can be performed directly on these surrogate models using gradient
based or non-gradient based techniques like genetic algorithm. In this case, the result
obtained at the end depends on the accuracy of the surrogate models and the final result
may be a poor design. In order to obtain accurate surrogate over the complete design
space, large number of function evaluations are required which can be very time
consuming. The alternative to this “one-shot” approach is to account for the uncertainty
in the surrogate model. This can be achieved using the Efficient Global Optimization
(EGO) algorithm which accounts for uncertainty in the surrogate and is more efficient.
The optimization performed using EGO algorithm is iterative in nature, as
described in Chapter 3. In EGO, a small number of initial design points are used to fit a
kriging approximation in the first iteration, instead of starting with a large number of
fitting points to obtain an accurate surrogate model. In the next step, the objective
220
function to be minimized or maximized is replaced by the Expected Improvement
Function (EIF) which is maximized during the optimization process. The optimized set of
points obtained by maximizing EIF are referred to as “Infill Samples” and are chosen to
be in the region where there is a high probability of producing a superior design over the
current best design and/or where the predictions of the surrogate are unreliable due to the
high amount of uncertainty. Thus, these infill samples represent a balance between the
local consideration of finding an optimal design based on the information in the
surrogate, and the global consideration of sampling in the design space where there is
much uncertainty in the surrogate’s predictions. The objective function and constraints
are determined again at these “Infill Sample” points and the surrogate models are updated
using the old and new set of fitting points. This process is repeated multiple times till the
stopping criterion is satisfied.
Therefore, the EGO algorithm is able to adapt to potential errors in the
approximate objective function by sampling at points at which there is much uncertainty
in the surrogate’s predictions.
Expected Improvement Function
The expected improvement function can be written as:
EIF ( x)  1   2 if s > 0
=0 if s = 0
where,
 ymin  yˆ krg 
1  ( ymin  yˆ krg ) dist  
 s 
221
and
 ymin  yˆ krg 
 2  sden  .
 s 
The functions  dist (*) and den (*) represent the standard normal distribution function and
the standard normal density function respectively. The first term in the EIF, 1 , is the
difference between the current best objective function value and the response at an
arbitrary design, x, multiplied by the probability that y(x) is better than ymin. This term is
large when yˆ krg is likely to be better than ymin. The second term,  2 , is large when the
error metric s(x) is large which signifies significant uncertainty in the surrogate’s
prediction. The design point with the highest EIF value represents the balance between
finding a better point and finding regions of high uncertainty. In MATLAB, EIF(x) can
be evaluated using the following expression:
1 1  ymin  yˆ krg  1  ( ymin  yˆ krg )2 

EIF ( x)  ( ymin  yˆ krg )   erf    s exp  
2  2s 2
2 2  s 2   
where, erf(*) is the error function defined in MATLAB.
222
Appendix B. UM/NLABS-A Aerodynamics Model
This section provides detailed description of the unified aerodynamic model used in
UM/NLABS-A. The unified airloads model includes three key elements: Peters flexible
airloads theory, the 3D dynamic inflow model and the modified ONERA stall model. The
schematic of the unified airloads model is shown in Figure B-1. The unified airloads
model accounts for arbitrary airfoil motion, morphing airfoil shape, and the dynamic stall
effects. The description provided in this section is based on the detailed analysis given in
[145, 168, 169].
Figure B-1: Unified Airloads Model
B.1 Peters Flexible Airfoil Theory
The 2D aerodynamic analysis is based on 2-D finite state formulation for a flexible
airfoil, as originally presented in [167]. Consider a thin airfoil of arbitrary shape moving
through the thin air as shown in Figure B-2. As shown in the figure, the co-ordinate
system is centered at the mid-chord and b is the semichord. With respect to the frame, the
223
fluid moves with horizontal velocity u0, vertical velocity v0 and rotation v1. The
deformation of the airfoil is given by the distribution h(y,t), which is defined positive
down. It is assumed that the deformation within the reference frame is small, such that,
, and . Trailing edge vorticity is assumed to be emitted in
the direction of +Y axis.
Figure B-2: General Airfoil Coordinate System
For this airfoil configuration, the non-penetration boundary condition can be written
as:
h h y
w  v    u0   v0  v1
y t b
where, w is the total induced flow, λ is induced flow from the trailed circulation and v is
the induced flow from bound circulation. Expressing v in terms of bound circulation per
unit length  b :
b
1  b ( , t )
2 b y  
v d
The vorticity equation gives the loading due to circulation as:
224
 b
y
P   u0 b    d where, (b  y  b)

b
t
The spatial gradient of the induced flow due to shed wake is related to the temporal
gradient of the induced flow:
  1 d
 u0 
t y 2 (b  y)dt
Above equations define the flexible airfoil theory, which must be expressed in terms
of frame motion, and blade deformation. To carry out this transformation, all the
variables are expressed with respect to the Glauert variable, φ. The change of variables is
given by:
y  b cos 
where, b  y  b, and 0     .
After the substitution, the expansions are as follows:
   cos   
b  2  s  0    n sin(n ) 
 sin  sin  n 1 
   cos   
P  2   s  0   n sin(n ) 
 sin  sin  n 1 
Similarly, the blade deformation, velocity and induced flow may be expressed as
expansions in the Glauert variable:

h   hn cos(n )
n 0

w   wn cos(n )
n 0

   n cos(n )
n 0
225
The cos(n ) terms in the equation above correspond to the Chebyshev polynomial of the
first kind along the nondimensional chordwise direction. Thus, there is a physical
meaning for each term in the expansion. The first three terms correspond to plunge, pitch
and camber respectively.
The airloads can be expressed in terms of the airfoil motion wn and the uniform
component of the induced flow, λ0, by use of the vorticity equation:
u0 (w0  0 )   0
1
b( w0  w2 )  u0 w1  1
2
b
( wn1  wn1 )  u0 wn   n , for n > 2
2n
The generalized loads are determined by substitution into the pressure distribution and
integrating over the airfoil, that is:

b 
Ln    P cos(n )dx    bP cos(n )sin d
b 0
The final results for the generalized loads are:
1
L0  2bfu0 ( w0  0 )   bu0 w1   b2 ( w0  w2 )
2
1 1
L1  bu0 (w0  0 )  bu0 w2  b 2 (w1  w3 )
2 8
1 1 1 1
L2  bu0 ( w1  w3 )  b2 (w0  w2 )  b 2 (w2  w4 )
2 2 2 12
1 1 1 1
Ln   bu0 ( wn1  wn3 )  b2 ( wn2  wn )  b2 ( wn  wn2 )
2 4(n  1) 2 4(n  1)
226
The first two generalized loads, which correspond to lift and moment about the
midchord, are completely defined by the first few terms of the velocity expansion. The
lift and pitching moment are completely defined by the plunge, pitch and camber of the
airfoil. The load L0 is uniform force acting in the negative Z direction, that is, the
negative of the conventional definition of lift. The load L1 is a linear force distribution, so
the quantity is the conventional nose up pitching moment about the midchord.
Writing the velocities in terms of the frame motion and blade deformation:

nhn
w0  v0  h0  u0 
n 1,3,5 b

nhn
w1  v1  h1  2u0 
n  2,4,6 b

nhn
wm  hm  2u0 
n  m 1, m  3 b
, m≥2
Using these expressions, the final load equation can be written in matrix form.
The total circulation is found to be:
1 1
  2 bf ( w0  0 )  w1  1
2 2
The chordwise loading includes the induced drag and the leading edge suction force. It
can be written as:


h
D   b(P) sin  d  2bf ( w0  0 ) 2
0
x
B.2 Dynamic Inflow Model
The dynamic inflow theory proposes the following solution to the velocity normal to
the rotor disk in terms of radial and azimuthal expansion function:
227
 
w( s, , t )     jr ( s)  rj (t ) cos(r )   jr (t )sin(r ) 
r  0 j  r 1, r  3
where, s and t are the non-dimensionalized radius and time and ψ is the azimuth angle.
The inflow states,  rj and  rj correspond to coefficients of the coupled terms containing
the azimuthal harmonics and radial expansion functions,  . Using the circulatory part of
the lift obtained from flexible airfoil theory, wake skew angle, and the freestream
velocity, the coefficients corresponding to the inflow states can be determined. Detailed
description is provided in [168]. The zero-order inflow coefficient needed for the airloads
expression is obtained from:
 
 rb 
0    J 0    jr ( s)  rj (t ) cos(r q )   jr (t )sin(r q ) 
r 0 j  r 1, r 3  s 
The expression contains the Bessel function of the first kind, J 0, which can be
approximated by taking first few terms of the Taylor series expansion, under the
assumption that b/s is small for blades of typical aspect ratio, at sufficient distance from
the rotor center.
B.3 Dynamic Stall Model
Dynamic stall occurs when some of the sections along the blade span oscillate in and
out of the stall regime, as the blade rotates around the azimuth, resulting in hysteresis
behavior for lift, moment and drag coefficeints. The static loss of lift acts as the forcing
function to drive the ONERA differential equation for dynamic stall. When dynamic stall
occurs, airloads display a time delay and an overshoot due to the passing of shed
vorticity. In order to allow for this phenomenon, a second order transfer function is
introduced.
228
Different steps involved in determining the loads generated due to dynamic stall
effects are as follows:
a) Calculate 2D Aerodynamic Loads
As described in earlier section, Peters flexible airfoil theory is used to determine
the aerodynamic forces. At the end of the analysis, load vector Ln and induced
drag force component D are obtained.
b) Calculate the Angle of attack
It should be noted that in the Peters flexible airfoil theory, the angle of attack is
not calculated explicitly. However, the angle of attack needs to be determined in
order to calculate the static stall residuals and the coefficient of drag for profile
drag correction term. In the current analysis, the angle of attack is defined as:
 vt 
  tan 1  
 u0 
where, u0 is the incoming freestream velocity and vt is defined as:
vt  hn (0)  vn (0)  0 (0)
c) Determine Critical Angle of attack
In the current analysis, critical angle of attack (in degree) is defined as
cr  13(1  M 2 )
where, M is the local mach number at the airfoil cross sections.
d) Calculate Delayed Angle of Attack
The time delay equation for determining the delayed angle of attack is given by:
 d  d d  d
229
where,  d is the delayed angle of attack and  d  U
 d b . In the current analysis,
τd = 12, however, its correct value needs to be determined using experimental
results.
e) Calculate Static Stall Residual
As described in the earlier section, static stall residuals are the forcing
parameters for the second order differential equation. Static stall residual
represent the difference between the thin airfoil values for the airloads with
appropriate static correction – and the experimental observation. In the current
analysis, the experimental observations are substituted by the results from X-
FOIL analysis. Static stall residual for lift and drag coefficient are shown in
Figure B-3. For this particular case, Δcl is positive while Δcd is negative. Static
stall residuals for different aerodynamic coefficient are defined below:
For lift coefficient:
CL  CL sin( )  CL,Table for  d   cr
CL  0 for  d   cr
For moment coefficient:
a
CM  CM  sin( ) cos( )  CD ,Table sin( )  CM ,Table for  d   cr
2
CM  0 for  d   cr
For Drag coefficient:
CD  CD 0  CD,Table for  d   cr
CL  0 for  d   cr
230
Figure B-3: Static Stall Residual
f) Calculating Loads due to Dynamic Stall Effect
The loads generated due to the dynamic stall effects, for each of the
generalized coordinate, are computed using the following differential equation:
b2 b  d Cn b 
2
 n  ˆ  n  ˆ 2 n  buT ˆ 2  Cn  eˆ 
uT uT  dt uT 
The parameters , and are assumed to be of the functional form:
ˆ  0  2  CL 
2
ˆ  0  2  CL 
2
eˆ  e0  e2  CL 
2
where, the parameters e0 ,e2, ω0, ω2, η1 and η2 are determined by parameter
identification.
231
g) Profile Drag
The effect of profile drag is included here in the quasi-steady sense using the
table-lookup generated using XFOIL.
h) Combining all the loads
Since the stall model is based on flow-based reference system, it is assumed
that the stall corrections to lift and drag are perpendicular to and parallel to the
local freestream velocity, as shown in Figure B-4. The total life and drag forces in
the large angle reference frame obtained by including the effect of dynamic stall
loads and profile drag effects are given by:
LT 0  L0  u00  bcd uT vL
DT  D  vL0  bcd uT u0
Figure B-4: Direction of Aerodynamic Forces
232
Appendix C. Design of Active Flap
Active flaps have proven to be very effective in reducing vibratory loads at the hub,
minimizing noise and in some cases improving performance. Among active flaps, dual
flaps have shown promising results in reducing vibrations and improving performance on
a rotor as compared to a single flap. A number of experiments have been conducted to
show the potential of active flaps in influencing hub loads. However, the performance
penalty associated with oscillating flaps is yet to be quantified using experimental data.
This performance penalty is critical for the implementation of active flaps on a rotor. The
data obtained would be useful for validating the results obtained from CFD based
simulations and ROM (reduced order models) based on CFD. Thus, the aim of the
experimental analysis is to:
1) Test the effectiveness of dual flaps in influencing vibratory loads at the hub on a
rotor in hover conditions.
2) Experimentally determine the performance penalty associated with oscillating
dual flaps on a rotor blade in hover conditions.
For this purpose, an active 5ft long Mach-scaled composite rotor blade with dual flaps
was designed and fabricated which could be tested in a hover spin test stand.
In this chapter, the design and fabrication of a flap-actuation mechanism is
presented, which can be installed on a Mach-scaled rotor blade for testing the effects of
dual flaps. The analysis performed includes detailed characterization of the two X-frame
233
piezoelectric actuators for determining their stiffnesses. In the next step, different
components in the flap-actuation mechanism were sized to ensure sufficient flap
deflection and strength of different components. The sizing was performed by
determining the stiffness of the actuation system and the load path and using the
impedance matching criteria to ensure maximum energy transfer. Finally, the flap
supports were fabricated and tested on a bench set up and the amplitude of flap deflection
and the output force obtained were measured.
The active flaps mounted on a rotor blade require high frequency of actuation for
vibration reduction which is difficult to obtain using a servo-valve/hydraulic actuator
[38]. The typical requirement for actuation frequency ranges from 2/rev to 5/rev
frequency, depending upon the number of rotor blades in the helicopter. This corresponds
to a frequency range of 20Hz to 50Hz for a rotor blade rotating at 600RPM. Besides the
high bandwidth of actuation, piezoelectric actuators offer the advantages of: direct
conversion of electrical energy into linear motion, less number of parts like pipes which
are required for the hydraulic actuators, and smaller weight penalty. Piezoelectric
material are capable of providing a large force, however the stroke provided by the piezos
is very small and is limited by the inherent 0.1% cap on the free induced strain. Thus, for
the practical implementation of piezoelectric material for active flap application, some
form of amplification mechanism is required.
The requirements for an actuator [119, 128], which can be used for actuating flaps on
a rotor blade, are based on the fact that they provide sufficient mechanical output of force
and displacement without incurring any penalty on the structural and aerodynamic
properties of the blade. The actuator must provide sufficient force to act against the
234
aerodynamic hinge moment and the stiffness of the flap-hinge mechanism and
demonstrate flap deflection of ±4 deg at nominal RPM in different dynamic operating
conditions. The actuator should be light in weight and the increase in mass of the blade
due to the actuation system should be less than 20%. The installation of the actuator near
the leading edge of the airfoil is beneficial from the aeroelastic stability point of view.
The actuator must be capable of oscillating flaps at high frequencies (for an N bladed
rotor, the actuator should provide sufficient amplitude of flap deflection up to (N+2)/rev
frequency of actuation). To avoid any undesirable aerodynamic effects due to the
actuation system, the actuator must be small enough such that it can fit in the blade spar.
Actuators which extend over the large chordwise span create an issue of mass imbalance
and are difficult to incorporate in the rotor blade. The actuator designed should have
sufficient fatigue life and it should be able to perform in the presence of large vibration,
unsteady aerodynamic loads and thermal environment. In most of the piezoelectric
actuators, piezoelectric material is in the form of piezostacks which are connected in
parallel to maximize the output displacement. In order to maintain the integrity of
piezostacks and prevent them from discharging, a constant prestress is required. Thus, the
actuator/actuation mechanism should be capable of providing prestress to the piezostacks.
The control system operating the active flap should provide resolution and position
sensing accuracy of 3% or less of the full range.
Among the current programs testing active flap, the SMART rotor program at
Boeing [115] is using a double X-frame actuator developed by Hall et al. [174], while at
the Eurocopter’s ADASYS (Adaptive dynamic systems) rotor system [46], an amplified
piezoelectric actuator developed by Cedrat Technologies in France [175] is being used.
235
More recently, the on-blade electro-mechanical actuator (EMA) developed by Hamilton
Sundstrand Claverham and UTRC [116] was used by Sikorsky in their whirl and wind
tunnel testing. Since the design of an actuator was not the main aim of this thesis, an off-
the shelf actuator had to be obtained. Based on the literature survey conducted for
different kinds of actuators available, the X-frame actuator manufactured by Axis
Engineering Technologies, Boston and the Amplified Piezoelectric Actuator (APA)
manufactured by Cedrat Technologies, France were short listed.
Table C-1: Comparison between X-frame Actuator and APA Actuators
Features X-frame APA 200M APA 400 M APA 900 M

Physical
Length (mm) 80.9 55 48.4 49
Width (mm) 18.99 17 13 11.5
Height (mm) 10.72 9 11.5 10
Mass (gm) 40 15.7 19 19
Mechanical
Displacement(μm) 1404 230 400 900
Blocked Force (N) 95 73 38 16
Stiffness (N/μm) 0.06 0.32 0.1 0.02
Resonance Freq NA 900 495 248
Electrical
Max Voltage (V) 1000 150 150 150
Capacitance (μF) 0.113 3.2 3.15 3.15
Calculated
Lever Arm (s) (mm) 2.88 1.03 1.85 4.13
Flap deflection(δ) (deg) 19.42 8.63 8.38 8.44
Physical size of the actuator and the mechanical displacement and force provided
by the actuator were the main driving factor for selection of the actuator. The actuators
shown above were shortlisted based on their size and the size of supports required to hold
them inside the blade. Among all the actuators listed in Table C-1, the X-frame actuator
provides the maximum mechanical displacement and blocked force. During earlier
experiments with active flaps [112], it was observed that the friction in rotating condition
236
due to high RPM can lead to a reduction in the amplitude of flap oscillations. Hence, it is
desired that the actuator with maximum energy output is selected to ensure sufficient flap
deflection at nominal RPM.
C.1 Quasi-static Tests
Quasi-static test of the actuator is performed to study the performance of the
actuator and to determine its stiffness. The actuator performance is monitored by
observing the actuator deflection as a function of input voltage and externally applied
elastic load.
Figure C-1: Schematic of Quasi-Static Test
Figure C-2: Experimental Setup for Quasi-Static Tests
237
Figure C-3: Cage Region for holding the Actuator
The schematic of the experimental setup which was used for the quasi-static tests
is shown in Figure C-1. The actual components of the setup are shown in Figure C-2 and
Figure C-3. As shown in Figure C-2, the initial setup was done on a normal table but the
initial tests results showed large unexpected vibrations. As a result, the experimental
setup had to be transferred to an optical table for final testing. The setup consists of a
steel cage in between which the actuator is held as shown in Figure C-3. The output end
of the X-frame actuator is attached to a steel wire which runs across the table. At the
other end of the steel wire, a constant mass of 19lb is attached through a pulley to provide
a constant pre-stress to the actuator. The stiffness of the elastic load acting on the actuator
is varied by changing the length of the wire (Lv) between the actuator and the table vise.
This is obtained by clamping the wire at different locations along the length of the table
during the tests. The length Lv was varied between 16” to 61” during the quasi-static
tests. The actuator pre-stress was measured using a single axis load cell mounted behind
the X-frame actuator. The displacement produced by the actuator was measured using a
laser extensometer. The laser extensometer used for the tests has an accuracy of 0.1 mils
(The expected value of displacement for these tests was around 20 mils). Both the load
238
cell and the laser extensometer were calibrated prior to the tests. The diameter of the
metal wire was selected in such a way that the load stiffness due to the wire is in the
range of expected aerodynamic hinge moment stiffness. Table C-2 shows the value of
load stiffness provided by the metal wire for different lengths. Based on the approximate
aerodynamic hinge moment loads calculated, a steel wire with diameter of 0.018 inch
was used for these tests.
Table C-2: Variation of Load stiffness with the Length of Wire
Sr. No Lv Stiffness
(in) (lbf/in)
1 16 481.0
2 20 384.8
3 25 307.8
4 45 171.0
5 61 126.2
The tests were performed at peak-to-peak voltages of 500V, 600V, 700V and 800V at
1Hz frequency to simulate the quasi-static conditions. In all these cases, the DC offset for
the input voltage was adjusted such that the minimum value of the voltage applied was
0V. This is required to ensure that no negative voltage is applied to the actuator, which is
detrimental to the health of piezo-stacks used in the actuator. FFT analysis was performed
on the data obtained from the load cell and laser extensometer using MATLAB and the
amplitude corresponding to 1Hz frequency was obtained.
Figure C-4 shows the results obtained from the quasi-static tests on both the X-frame
actuators for characterization. The results obtained are compared to results obtained in
[112] for 800Vp-p actuation. It can be seen that the value of displacements and loads
obtained for both the actuators are close to those obtained for the reference case. The
239
reference case corresponds to the data obtained for an X-frame actuator of similar size
tested at MIT [112]. The actuator stiffnesses obtained from Figure C-4 are listed in Table
C-3. It can be seen that the stiffness obtained from the current set of experiments is
higher than that obtained in [112]. This was expected, since a small modification was
made in the new X-frame design to improve the performance of X-frame actuator.
10
X1
9 X2
X(Ref)
8 800V
Load (peak-to-peak) (lbf)
700V
7
6 600V
500V
5
1
5 10 15 20 25 30
Displacement(peak-to-peak) (mil)
Figure C-4: Characterization of X-frame Actuators
Table C-3: Actuator Stiffness obtained from Quasi-Static Tests (Units: lbf/in)
Voltage X1 X2 X (Ref)[112]
500V 592.5 396.0 397.0
600V 527 465.2 393.2
700V 617.3 424.7 357.1
800V 541.2 446.3 399.0
Average 569.5 433.0 386.6
C.2 Dual-flap Design
Based on the requirements for the tests, following parameters were fixed for the dual-
flap design as shown in Figure C-5:
240
1) Chord length of the flap was fixed to 0.25c and the span-wise length of each flap
was fixed to 0.06R
2) The first flap of the dual flaps extends from 0.72R to 0.78R while the second flap
extends from 0.79R to 0.85R. Flaps could not be moved further towards the tip
since there is a decrease in the thickness of the airfoil cross section beyond 0.85R.
Both the flaps were kept close to each other so that when both the flaps operate
with zero degree phase difference in the actuation voltage supply, it is
approximately equivalent to a single 12% flap.
3) The actuators were mounted in the blade spar such that they are centered at 0.25c.
As a result of this, the maximum thickness is available in the airfoil cross-section
to mount the actuator and no additional ballast mass is required to balance the
actuator weight.
Figure C-5: Location of Flaps and Actuators
241
C.2.1 Aerodynamic Hinge Moment
The aerodynamic hinge moment had to be determined prior to designing the
components for flap hinge mechanism. In order to minimize the aerodynamic hinge
moment against which the active flaps needs to operate, an optimum position for the flap
axis needs to be determined. For the purpose of these tests, a flap chord of 0.25c was used
as discussed above.
The geometry of the VR7 airfoil with a 25% plain flap is defined in Figure C-6. A
hinge gap of 1%c is present between the airfoil and the flap, as illustrated in Figure C-6.
In order to determine the optimal flap hinge location so as to minimize the flap actuation
power requirement, simulations using the CFD++ code were conducted for this two-
dimensional airfoil/flap configuration. The CFD++ code is a compressible Reynolds-
averaged Navier-Stokes flow solver which uses a finite volume formulation. Airfoil and
flap grids were generated using ICEM-CFD, and an overset mesh approach is employed
where a separate body-fitted mesh for the flap is generated in addition to the airfoil mesh,
as illustrated in Figure C-7.
The simulations are conducted for the flow condition of M=0.538 and Re=1.79x106,
which corresponds to the flow at the 0.85R spanwise location of the Mach-scaled model
rotor. The Spalart-Allmaras (S-A) turbulence model is used and a fully turbulent
boundary layer is assumed.
242
Figure C-6: VR7 airfoil with a Plain Flap
Figure C-7: Grids for the Airfoil with Flap
Figure C-8: Pressure Contour for α=4° and M=0.538

A sample result of pressure contour from the CFD simulation is shown in Figure C-8,
for the case of =4°. To determine the optimal flap hinge location, hinge moment curve
slope CHδ is calculated as a function of various hinge locations, defined by the distance ch
243
from the leading edge of the flap, as illustrated in Figure C-9. The hinge location ch is
given as a percentage of flap chord cf. Figure C-9 shows the variation of CHδ versus ch, at
various airfoil angles of attack. A negative hinge moment curve slope implies an unstable
configuration. From this figure, a flap location of 0.365cf appears to be a good tradeoff
between low flap actuation power and stability, for this flap configuration. The value of
hinge moment coefficients for flap-hinge located at 0.365cf are: CHδ = 0.948x10-4 and
CH0f = -0.0016.
Figure C-9: Hinge Moment Curve Slope (CHδ)
In order to make sure that the flap effectiveness was not affected by moving back the
hinge axis location, the lift generated by flap deflection was determined for different
hinge axis location at angle of attack of 0 deg and 8 deg, as shown in Figure C-10. The
results obtained from CFD show a very small difference in the lift coefficient curves
corresponding to different flap axis locations. Very small decrease in the CLδ is observed
244
due to flap hinge location as shown in Figure C-11. Thus the flap effectiveness was not
compromised by moving back the flap hinge location.
Figure C-10: Variation of Lift Coefficient due to Flap Deflection
Figure C-11: Variation of CLδ with Hinge Location
C.2.2 Sizing of the Parts for Flap Hinge Mechanism
Different components of the flap-hinge mechanism had to be sized properly and
analyzed to ensure:
245
1) sufficient flap deflection,
2) sufficient strength and fatigue life,
3) small size of the parts so that they can be installed inside the blade easily,
4) small mass so that there is no mass penalty, and
5) low friction during operation at full RPM
Based on the above criteria and design used in [112], an approximate design for the
flap hinge mechanism was developed. Many improvements were made in the new design
to reduce friction and compliance in the system. As shown in Figure C-12, the airfoil
cross section includes a cutout in the spar to hold the actuator. The supports required for
holding the actuator and the flaps are integrated inside the blade during the fabrication
process. Figure C-13 shows the detailed view for one of the flaps. The flap hinge-
mechanism used for the both the flaps was identical.
Figure C-12: Schematic of the Dual Flap Section of Rotor Blade
246
Figure C-13: Detailed View of the Flap Supports
Important components involved in the flap actuation system are:
1. Control rod: It is used to transfer actuation from the actuator to flaps (to clevis
which is linked to flap horn on the active flap) as shown in Figure C-12. Its
dimension was fixed by the size of the 0-80 threading used in the actuator and
clevis pin. The minimum diameter corresponding to 0-80 threading is 0.056”
(1.42 mm). It is expected to carry the load due to the prestress (~ 19lb) and
actuation (max of 12lb). This corresponds to maximum load of 31lb on the control
rod. For this load, the axial stress obtained in control rod is 85MPa.
2. Prestress (PS) wire (Flap axis): It acts as the rotational axis for the flap rotation.
It has a torsional pre-twist which helps to keep the control rod in tension and thus,
the piezo-stacks in compression. In this design, a steel rod with diameter of dps =
0.0465” (~1.18mm) was used as the prestress wire. This corresponds to a
247
torsional stiffness of 1.5862 in-lbf/rad for a 6% flap. The reasons of using this
particular diameter are :
a) its torsional stiffness is of the order of expected aerodynamic stiffness,
b) it is easily available,
c) it matches the inner diameter of the ball bearing that was used at outboard
end, and
d) twisting this steel rod by 60deg was sufficient to obtain required prestress
without yielding the material.
3. Inboard end of prestress wire: The inboard end of the pre-stress wire is welded to
a wire flange which in turn goes inside the reaction rib. The inboard end of the
prestress wire should have very small compliance and the welded region should
be able to carry the shear force due to pretwist and actuation. The diameter of the
inboard end of the prestress wire used was 0.086”.
4. Wire flange: It is welded to the inboard end of the prestress wire inside the flap. It
includes two 0-80 threaded holes, which are used for holding the flaps during the
operation. The wire flange is welded at an angle of 60 degree to the horizontal to
provide prestress to the actuator.
5. Flap Horn: It is used to convert the linear motion of the actuator in to rotational
motion for the flaps. Flap horn includes two holes, one for the prestress wire
inside the flap and the other for clevis. The vertical distance between these two
holes is the moment arm for converting the linear motion to rotational motion.
The flap horn is fixed inside the inboard end of the flap during the fabrication of
composite flap.
248
6. Clevis: Clevis is the link between the control rod and the flap horn. It includes a
0-80 threaded hole at one end which holds the control rod. At the other end, it is
shaped like a fork and holds the flap horn in between using a steel pin.
C.2.3 Calculation of Compliance in the System
The effective stiffness of the actuator reduces mainly due to the compliance of
following components in the flap actuation system: the axial strain in control rod, the
bending of servo-flap horn, the torsion of flap skin and the bending of inboard end of the
prestress wire. The effective compliance of the actuator is obtained by adding the
compliance for all the components in the actuation load path. The compliance of different
components in the actuation path is described below:
a) Compliance of actuator: The actuator 1 has a stiffness of 569.5 lb/in while the
actuator 2 has stiffness of 433.01 lb/in (as shown in Table C-3). This corresponds
to a compliance of Cact1 = 0.0018 in/lb for actuator 1 and a compliance of Cact2 =
0.0023 in/lb for actuator 2.
b) Compliance of control rod: The control rod has a diameter of 0.056” at its ends so
that it can be fixed to moving frame of the actuator at one end and clevis at other
end. In the middle, it has diameter of 0.125” to avoid bending of the control rod.
This corresponds to a stiffness of 5942 lbf/in or compliance of Ccr = 1.682x10-5
in/lb.
c) Torsion of flap skin: The flap skin consisted of 2 layers of E-glass 120 oriented at
±45 deg and a layer of unidirectional IM7 ply added to front 55% of the flap. A
finite element mesh for the cross section of the flap was developed in order to
249
determine its torsional stiffness. From the UM/VABS output, the torsional
stiffness (GJ) of the cross section was obtained and it was equal to 2.02 Nm 2. The
torsional stiffness for the flap was obtained by using the expression TFS
=(GJ/L)flap (= 21.99 Nm/rad). Assuming a moment arm of 0.12”, this is equivalent
to stiffness of 2.36x106 N/m. Thus, the compliance due to torsion of flap is CFS =
7.395x10-5 in/lbf (=4.223x 10-7 m/N).
d) Inboard end of the pre-stress wire: The flexing of the inboard end of the pre-stress
wire adds compliance to the system. The diameter of this wire is dips = 0.086” and
its length is Lips = 0.2”. The compliance of this section is approximately given by
Cips = Lips3/3EI, where I =πdips4/64. Substituting these values, the compliance due
to the inboard end of the pre-stress wire was obtained and it was equal to Cips =
3.26x10-5 in/lbf (= 1.862x10-7 m/N)
e) Bending/Shearing of flap horn: The bending stiffness of flap horn was
approximated using Timoshenko beam analysis and it was equal to 1.343x106
lbf/in. This corresponds to a compliance of Cfh = 7.44x10-7 in/lbf.
Net compliance of actuation path is given by:
Ca = Cact + Ccr + CFS + Cips + CFH
Final stiffness of the actuation system is given by:
Ka = 1/Ca
The final stiffness obtained for the actuation path after all the calculation was equal to
527.3 lbf/in for actuator 1 and 408.2 lbf/in for actuator 2.
250
C.2.4 Stiffness of the Load Path
The load stiffness consists of torsional stiffness due to the aerodynamic hinge-
moment, torsional load due to stiffness of the pre-stress wire and stiffness of cross-
flexures of the centrifugal flexure in the X-frame actuator. An approximate analysis to
predict each of these components is given below.
a) Aerodynamic stiffness: The aerodynamic torsional stiffness for a flap mounted on
a rotor blade with inboard end at Ri and outboard end located at Ro is given by :
1
K aero  2cs 2 ( Ro3  Ri 3 )CH 
6
where CHδ is the flap hinge moment obtained from aerodynamic analysis and cs is
the chord of the airfoil section. Substituting values of the different variables (ρ =
1.226 kg/m3, Ω = 140rad/sec (1336 RPM), cs = 5.388 in, Ro = 0.78R and Ri =
0.72R for inner flap and Ro = 0.85R and Ri = 0.79R for outer flap and CHδ=
0.948x10-4 /deg) in the above expression, aerodynamic stiffness values for two flaps
are obtained and they are given by :
Kaero,1 = 1.331 in-lb/rad and Kaero,2 = 1.590 in-lb/rad
(Note: Subscript ‘1’ corresponds to inboard flap while subscript ‘2’ corresponds to
outboard flap)
b) Torsional stiffness of prestress-wire: The prestress wire helps in providing
prestress for the X-frame actuator and it acts as the hinge for the flap rotation.
Torsional stiffness of prestress wire is given by :
where
251
The torsional stiffness of prestress wire obtained after substituting all the values
is: 1.586 in-lb/rad.
c) Stiffness of flexure: Stiffness of the flexure used in Ref [112] was 0.152 in-lb/rad.
Assuming 10% increase in the stiffness (as mentioned by manufacturer), stiffness
of flexure in our case can be approximately given by Kf = 0.167 in-lb/rad.
Thus, the total load stiffness is given by:
KL = Kaero + Kps + Kf
Summing up all the stiffness, load stiffness obtained for the two flaps are:
KL1 = 3.084 in-lbf/rad and KL2 = 3.344 in-lbf/rad
C.2.5 Impedance Matching
Moment arm required to obtain the impedance match condition is:
KL
s
Ka
This gives, moment arm of s1 = 0.074 inch for the inboard flap and moment arm of s2 =
0.088 inch for the outboard flap.
252
15
X1-Actuator
X2-Actuator
Flap 1
10 Flap 2
Force (lbf)
5
0
0 5 10 15 20 25
Deflection (mil)
Figure C-14: Ideal operating point obtained using impedance matching
Combining flap 1 with actuator 1 (set1) and flap 2 with actuator 2 (set 2), we get peak-
to-peak displacements of 33.6 mil and 30.74 mil for set 1 and set 2 respectively, and
peak-to-peak forces of 13.94lbf and 16.64 lbf for set 1 and set 2 respectively.
Due to manufacturing constraints, it was difficult to fabricate parts with very small
moment arms. Hence, the moment arm used for both the flaps was fixed to 0.12 in. As a
result, there were small changes in the results obtained earlier as shown in Table C-4 and
Figure C-15. The results obtained with modified moment arm show higher flap deflection
but reduced forces. Based on above analysis, different parts were designed and fabricated
as shown in Figure C-18.
253
15
X1-Actuator
X2-Actuator
Flap1 (Ideal)
Flap2 (Ideal)
Flap1 (Act)
Flap2 (Act)
10
Force (lbf)
0
0 5 10 15 20 25
Deflection (mil)
Figure C-15: Effect of Moment Arm Modification on Operating Point
Table C-4: Difference between Ideal and Actual Operating Condition
Ideal Case Actual Case

Set1 Set2 Set1 Set2
Mom Arm (mil) 75.6 89.8 120 120
Actuator Disp (mil) 33.6 30.74 44.9 42.27
Actuator Force (lbf) 13.94 16.6 10.44 9.05
Flap Deflection (deg) 12.85 9.91 10.59 9.98
Prestress (lb) 22.56 18.89 13.84 13.84
C.3 Bench Test for Active Flaps
Bench test was performed to validate the flap hinge mechanism designed for
oscillating the flaps on a rotor blade. The parts which were designed for the bench test
were such that they can be easily incorporated on the active blade with very minor
changes. The bench test conducted with flap proved to be very useful in improving the
flap-hinge mechanism design to get maximum flap deflection output from the actuator.
For example, the initial design used for clevis was a curved one as shown in Figure C-12
254
and Figure C-17. This design was chosen so that a shorter moment arm, as required by
the impedance matching condition, can be obtained. However, during the early bench
tests, it was noticed that a curved clevis lead to bending of the control rod during its
motion which increased the compliance of the actuation mechanism. As a result, very
small flap deflections were obtained when the flap was actuated. This was corrected by
using a flat clevis as shown in Figure C-16. Small modifications were also made in the
inboard flap support in order to reduce the compliance. The CAD models for final parts
used in the flap hinge mechanism are shown in Figure C-16. Since these parts had to be
installed inside the rotor blade, they follow the airfoil contour, wherever possible. As a
result of this complicated profile, some of the parts had to be cut using water-jet cutting
technique in order to obtain an accurate profile. The final machined parts obtained are
shown in Figure C-17. All these components, along with active the flap, were installed on
a base plate for bench tests as shown in Figure C-18.
Figure C-16: CAD Model of the Flap Parts
255
Figure C-17: Actual Fabricated Parts for Flap Hinge Mechanism
Figure C-18: Setup for Bench Test of Active Flap
Experimental results obtained for flap deflection and peak-to-peak force at different
actuation frequencies are shown in Table C-5. Results show that the variation in
amplitude of flap deflection and output force is small at low actuation frequencies. At
high actuation frequencies, increase in flap deflection was observed at the cost of small
decrease in the prestress force. In all the cases, the actuation voltage was kept constant at
256
approximately 760Vp-p with an offset of 400V. The hysteretic behavior observed in the
flap-actuation mechanism is shown in Figure C-19. The presence of hysteresis is
common in piezoelectric material and several attempts have been made to capture this
effect in vibration reduction studies using active flaps [176].
Table C-5: Experimental Results for Bench Tests
Test Freq Peak-to-peak Offset Pre-Stress (lbf) Mean load Deflection (deg)
(Hz) (V) (V) (peak-to-peak) (lbf) (peak-to-peak)
5 1 563.52 293.33 10.61 19.31 4.86
6 1 751.24 392.77 12.54 20.67 6.74
7 10 732.92 393.14 10.92 20.65 6.41
8 22 748 393.1 11.86 20.72 7.83
9 44 748 393.1 11.18 20.4 7.24
10 66 756.68 392.94 8.68 20.14 9.56
11 88 757.04 392.86 11.55 18.56 9.49
12 110 765.28 392.99 14.1 17.69 10.45
4
1Hz
10hz
3 20Hz
30Hz
40Hz
2 50Hz
Flap delfection (deg)
-1
-2
-3
-4
0 100 200 300 400 500 600 700 800
Applied Voltage (V)
Figure C-19: Hysteresis in Flap Actuation
257
C.4 Conclusion
This chapter presented the design and experimental analysis of a flap-actuation
mechanism that can be installed in a Mach-scaled rotor blade for testing the effect of dual
active flaps. The analysis performed in this chapter is based on the X-frame actuator
which was developed at MIT in 2000. As a first step, quasi-static tests were performed on
the X-frame actuator to determine its stiffness and load-deflection relationship. This
property was used in the design of supports for active flap using the impedance matching
criteria. All the parts required for flap-actuation mechanism were designed and fabricated
in order to maximize the dynamic flap deflection amplitude while ensuring sufficient
output force to act against the prestress in flap and the aerodynamic loads. The final flap
actuation mechanism was bench-tested and sufficient flap deflection and output force was
observed. The experimental results showed flap deflection amplitude of more than 6deg
at high frequencies of actuation (till 5/rev actuation frequency which corresponds to 110
Hz, for nominal operating condition at 1336 RPM).
The flap actuation mechanism designed in this chapter was installed on a 10 ft
diameter Mach-scaled rotor blade and tested in the spin-test stand at the University of
Michigan. Further details related to the design and fabrication of rotor blade and
experimental tests are provided in following appendices.
258
Appendix D. Active Blade Design
The composite rotor blade to be used for testing the dual active flaps had to be
designed to meet the requirements for strength and sufficient fatigue life. Besides this, the
active blade includes cut-out in the spar to make space for mounting the X-frame actuator
and cut in the fairing to hold the flaps. Before fabricating the final blade, sample sections
of the blade were fabricated and tested in a tensile testing machine to check the blade
strength. The active blade also includes instrumentation like strain gages, accelerometers
to measure the blade deformation and hall effect sensors to measure flap deflection
angles during the tests. The shape of the blade was fixed by the mold being used for the
fabrication of rotor blade. In the current study, a 1/6th scaled version of the CH-47D blade
was used which has a radius of 5ft and chord of 5.38 in. Although, the outer mold line of
the blade (and hence the geometry) was already fixed, the composite layup required to
meet the design requirements had to be determined.
D.1 Geometry of the Baseline Blade
A 1/6th mach-scaled Chinook CH-47D blade is used as the baseline for integral blade
design. The spin test stand used for this test was designed with this particular blade in
view. The fact that the same blade was used in earlier tests on a similar test stand was
useful for obtaining preliminary data for design and validation purpose [58, 112]. Basic
dimensions of the blade are given in Table D-1 and the planform view of the passive
blade (without the active flaps) is shown in Figure D-1. The spin test stand hub is
259
articulated with the flap hinge axis located at 0.028R and the lead-lag hinge located at
0.15R. Blade pitch is fixed at a particular angle depending upon the required collective
setting.
Table D-1: Blade Properties
Property Value
Geometric Scaling 1/6th
Radius 60.619 in (1.539 m)
Chord 5.388 in (0.1368 m)
Number of blades 2
Rotor Type Articulated
Flap Hinge Location 0.028R
Lag Hinge Location 0.15R
Rotor Speed 1336 RPM
Figure D-1: Blade plan-form View [58]
The rotor blade has a non-uniform chord and thickness variation near the root region.
From 0.27R to 0.85R, rotor blade has a uniform cross section of VR7 airfoil. The blade
tapers from VR7 airfoil at 0.85R to VR8 airfoil at the tip. Both the airfoil sections are
shown in Figure D-2. The VR7 airfoil is a 12% thick airfoil while the VR8 airfoil is 8%
thick. Linear interpolation is used to obtain cross section shape for 0.85 R < r < 1.0R. The
built-in twist for the rotor blade is shown in Figure D-3.
260
1
VR7 AIRFOIL
VR8 AIRFOIL
Thickness (in)
0.5
-0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Airfoil Chord (in)
Figure D-2: Blade Cross Section
Figure D-3: Blade Twist Distribution
Cross-sectional details for the composite rotor blade are shown in Figure D-4. It
consists of prepreg plies wrapped around the foam core. In the first cure, the front D spar
of the blade is cured, while the fairing is attached to the front spar in second cure. Nose
weights are added near the leading edge while fabricating the spar to get the CG of the
cross section closer to the quarter chord. The number of plies for spar, web region and
fairing and their ply angles had to be determined prior to fabrication.
261
Figure D-4: Cross Section of the Rotor Blade
D.2 Blade-Layup Design Process
This section provides details about the blade structural analysis and the design process
used for obtaining the final composite blade design. The spanwise view of the rotor blade
with dual active flaps is shown in Figure D-5.
Figure D-5: CH47D Rotor Blade with Dual Active Flaps
Blade Analysis
The complete 3D analysis of the rotor blade is broken down into two steps, as
discussed in Section 2.2. In the first step, a linear cross-sectional analysis of the
composite rotor blade is performed to determine the cross-sectional properties of the
blade at different spanwise locations. And in the second step, aeroelastic analysis of the
rotor blade is performed on the 1D beam model. This breakdown of the analysis is valid
for slender structures like aircraft wing and rotor blades. For the cross-sectional analysis,
the blade shape and the layup information is required. Depending upon the complexity in
rotor blade geometry and variation in the layup, the blade is divided into several spanwise
sections. Blade properties are assumed to be constant in each section. In the current
262
analysis, the blade is divided into ten spanwise sections as described Table D-2. Since the
blade’s geometry and layup, both vary significantly in the root region, the number of
sections is more in the root region.
Table D-2: Spanwise Regions for Cross-sectional Analysis
Sections Span
Section 1F 0.33R to 0.728R
Section 1R 0.33R to 0.246R
Section 6R 0.151R to Root
Cross-sectional shape of different sections was determined based on the description
provided for the mold geometry by the mold manufacturer. The layup used for the root
section of the blade (from 0.15R to 0.33R) was same similar to that used in [58]. Once
the layup information and the cross section profile was available, a MATLAB based
mesh generator was used to generate the finite element mesh for UM/VABS. The finite
element mesh generated for different root sections is shown in Figure D-6. Reference axis
for the full airfoil region (Section 1R to Section 4F) is at the quarter chord, while for the
root region (section 6R and 5R); it is located at the mid chord. For sections 2R to 4R, the
reference line varies between the midchord and quarter chord. The reference point for
each section was determined such that the reference points for all the sections are
collinear points, as in the actual blade. Figure D-7 shows the finite element mesh
generated for the main blade (without the foam). In order to obtain the baseline values for
263
strains and aeroelastic loads, the blade with a single flap manufactured in [112] was also
modeled. The cross-sectional analysis was done using UM/VABS to obtain the inertia
and stiffness properties for 1-D aeroelastic analysis and strain influence coefficients for
the failure analysis.
Figure D-6: Finite Element Mesh for Root Section
264
Figure D-7: Finite Element Mesh for Main Blade Sections
The 1-D beam model for the whole blade was developed in both AVINOR and RCAS
for aeroelastic analysis. The model designed takes into account the position of hinges as
described in Figure D-8.
Figure D-8: Location of Hinges in Spin Test Stand
Design Process
Figure D-9 shows the steps followed in the design process used for the design of
active blade. Different components of the design process are described below.
265
Initial starting point: The initial starting point for the layup is similar to that used in
[112]. For the purpose of analysis, blade is divided into 10 span-wise sections as
discussed in the previous section.
Stiffness and Inertia Properties: The cross-sectional layup and geometry is assumed to be
constant within each span-wise section. Cross-sectional inertia and stiffness matrix for
each of the cross section are obtained using UM/VABS [70]. These properties are used as
input for the aeroelastic analysis of the rotor blade using the AVINOR code [97].
Worst Case Loading: In the analysis performed using AVINOR, the flaps are actuated at
different frequencies from 2/rev to 5/rev and the blade loading is extracted for each of the
cases. From these cases, maximum value of the load for each of the six components is
determined at each station. Also, the maximum amplitude of oscillatory component of the
load is determined for the fatigue analysis at each section.
Figure D-9: Steps used in Blade Design
266
Failure Analysis: Worst case loading obtained from the AVINOR analysis is used as the
input for the Failure Analysis. Strain influence coefficients obtained from UM/VABS
analysis are used for determining all the six strain components for each element of the
cross-sectional mesh. The maximum strain criterion is used to determine the failure point
for the blade. Maximum allowable value of the strain for each of the six components is
determined from a similar analysis of the baseline blade (blade with a single flap
described in [112]). Similarly, the maximum dynamic strain for each of the component is
determined using the amplitude of the oscillatory loads determined using AVINOR in the
previous step.
Design constraints: Design constraints used in the analysis include upper and lower
bounds on the location of the shear center, cross-sectional center of gravity and blade
dynamic frequencies.
Flap control authority: Flap control authority depends upon the dynamic properties of the
blade. It is a measure of oscillatory load generated by per unit deflection of the flap. In
the current set of experiments, measurement of the unsteady aerodynamic drag was the
main aim of the experiment. Thus, it was desirable that the blade torsional stiffness be
very high to ensure that the contribution of aeroelastic loads to the hub loads is minimal.
Blade Analysis Tools
Detailed description of UM/VABS and RCAS is given in earlier chapters. This section
provides details about the aeroelastic code AVINOR and the MATLAB based mesh
generator.
267
AVINOR
The AVINOR (Active Vibration and Noise reduction) code [97] has been developed
over the years at UCLA and University of Michigan. It performs aeroelastic rotorcraft
analysis with emphasis on computational efficiency while retaining sufficient fidelity.
The AVINOR aerodynamic model consists of four main components – (1) an attached
2D time domain unsteady aerodynamic model that accounts for compressibility and time-
varying free stream Mach numbers, (2) a semi-empirical dynamic stall model for
separated flow regime at high advance ratios, (3) a free-wake model which calculated
non-uniform inflow distribution, and (4) a reverse flow model. The structural model is
based on 1D finite element method that accounts for moderately large deflections. The
structural dynamic model used in the code can use cross-sectional properties provided by
UM/VABS for modeling composite rotor blade. The simulation code has been primarily
used to investigate active and passive approaches to improve rotor blade design. For
active control, code allows for single or multiple actively controlled flaps along the blade
span. The optimal flap deflections for various combinations of vibration reduction, noise
reduction and performance enhancement are determined by a variant of the higher
harmonic control.
Mesh Generator
The finite element mesh required for 2D cross-sectional analysis was generated using
a MATLAB-based mesh generator specially developed for UM/VABS. To create a
general airfoil wetted surface, pairs of co-ordinate points defining the contour of the
airfoil must be supplied. From the wetted surface, layers of given material are defined in
order to create the stacking sequence needed for internal structural configuration.
268
Material properties for each material are defined using table lookup. The inertial effects
associated with the ballast masses are added directly to the inertia matrix generated by
UM/VABS.
Final Blade Design
The initial layup used for the blade cross section was based on tests conducted at MIT
in 2000 [112]. As shown in Table D-3, the front spar consisted of 4 layers of fiberglass
and 1 layer of IM7 graphite, vertical web consisted of 3 layers of E-Glass while the blade
fairing consisted of one layer of E-Glass fiber.
Table D-3: Initial and Final Blade Design
Baseline
Spar Fairing Web
1. E-Glass 0 deg 1. E-Glass ±45 deg 1. E-Glass ±45 deg
2. IM7 0 deg 2. E-Glass ±45 deg
3. S Glass +45 deg 3. E-Glass ±45 deg
4. S Glass -45 deg
5. S Glass 0 deg
Final Blade Design
Spar Fairing Web
1. E-Glass 0 deg 1. IM7 + 45 deg 1. IM7 + 45 deg
2. IM7 0 deg 2. IM7 - 45 deg 2. IM7 - 45 deg
3. IM7 + 45 deg 3. E-Glass ± 45 deg
4. IM7 - 45 deg 4. E-Glass ± 45 deg
5. IM7 0 deg 5. E-Glass ± 45 deg
6. E-Glass 0 deg
During the design process, various ply configurations were tried. Final composite
layup configuration obtained which satisfied failure and design criteria are shown Table
D-3. Since the blade with dual active flaps includes two actuators, it experiences higher
centrifugal force. Hence, the number of plies in the blade cross section had to be
269
increased. Also, one of the design criteria was to obtain high torsional stiffness such that
the contribution to hub loads from aeroelastic effects is minimized. Thus, IM7 graphite
plies oriented at ±45 deg were used in the fairing and the blade web.
Table D-4: Cross-sectional properties for Section 1F and Section 2F
Section 1F Baseline Final Section 2F Baseline Final

E11(x106) (N) 4.97 11.9 E11(x106) (N) 4.58 9.54
E44(x102) (N/m2) 1.23 4.27 E44(x102) (N/m2) 0.50 2.25
E55(x102) (N/m2) 2.03 3.40 E55(x102) (N/m2) 1.76 2.25
m (x10-1) kg/m 3.02 3.62 m (x10-1) kg/m 2.41 3.76
SC(%c) 29.8 32.3 SC(%c) 39.85 40.66
Table D-4 shows the cross-sectional properties of Section 1F and Section 2F for the
baseline case and for the final design obtained. Table D-5 shows the structural dynamic
frequencies of the baseline blade and the final design in vacuum at 1336 RPM (100%
RPM). All the frequencies for the final design are higher than that for the baseline blade.
The torsional frequency of the final design is 33% higher than that of the baseline blade.
Table D-5: Dynamic Blade Frequencies
Type Baseline Final

st
1 Flap Freq (/rev) 1.02 1.019
2nd Flap Freq (/rev) 3.05 3.43
3rd Flap Freq (/rev) 5.38 5.84
1st Lag Freq (/rev) 0.51 0.502

2nd Lag Freq (/rev) 5.84 9.18
1st Torsion Freq (/rev) 4.4 5.88
D.3 Failure Analysis
Failure analysis performed for the composite rotor blade is based on the maximum
strain criteria. All the six components of strain for each element of the cross section at
270
different stations along the blade span were determined. The maximum value of strain for
each component was compared to the corresponding value for the baseline case. Table
D-6 shows the maximum strain for each of the 10 stations in compression and tension in
both longitudinal and transverse direction. Results indicate that the blade section 5, 6 and
7 experience the maximum strains. These sections correspond to the transition region
between the blade root and rest of the blade. Since the blade root has additional plies and
thus higher stiffness, strains in the root region are smaller even though it experiences
larger internal force. Table D-7 shows the maximum strain for the final design while
Table D-8 shows the percentage difference in the maximum strains experienced by the
baseline case and the final design case. As observed in the baseline case, the new blade
designed also experiences maximum strains in the transition region. The percentage
difference obtained for the maximum strains indicate that the new blade design has
smaller strains as compared to the baseline case.
Table D-6: Maximum Strains for the Baseline Case (Units: με)
Section ε11 ε22 ε11 ε22 Section ε12 ε13

1 1563 1643 -1077 -1053 1 3557 3630
2 1175 1199 -751 -704 2 2472 1465
3 2614 3695 -2455 -1402 3 5153 5640
4 2907 3810 -2602 -1597 4 6588 5971
5 3549 3599 -2371 -2278 5 6699 5491
6 3684 3510 -2494 -2722 6 4689 5669
7 3553 4829 -3681 -2579 7 4654 7591
8 2497 2936 -2252 -1797 8 3193 4422
9 1345 2034 -1617 -864 9 2168 3118
10 1065 1497 -1253 -786 10 1547 2471
max 3684 4829 -3681 -2722 max 6699 7591
271
Table D-7: Maximum Blade Strain for the Final Design (Units: με)

1 901 958 -662 -662 1 4372 3022
2 955 962 -694 -692 2 2118 942
3 1904 2458 -1500 -1268 3 4463 6517
4 1912 2588 -1647 -1289 4 5280 7361
5 1990 2908 -1495 -1360 5 5550 7119
6 1758 2331 -1688 -1241 6 3330 6729
7 2950 4133 -1482 -2185 7 5417 4307
8 2315 1660 -743 -1639 8 5525 4824
9 1339 1641 -601 -975 9 2425 1726
10 763 1320 -567 -612 10 1389 1379
max 2950 4133 -1688 -2185 max 5550 7361
Table D-8: Percentage Variation in Cross-Sectional Strains

1 -42.4 -41.7 -38.5 -37.1 1 22.9 -16.7
2 -18.7 -19.7 -7.6 -1.7 2 -14.3 -35.7
3 -27.2 -33.5 -38.9 -9.6 3 -13.4 15.6
4 -34.2 -32.1 -36.7 -19.3 4 -19.9 23.3
5 -43.9 -19.2 -37.0 -40.3 5 -17.2 29.7
6 -52.3 -33.6 -32.3 -54.4 6 -29.0 18.7
7 -17.0 -14.4 -59.7 -15.3 7 16.4 -43.3
8 -7.3 -43.5 -67.0 -8.8 8 73.0 9.1
9 -0.5 -19.4 -62.8 12.8 9 11.9 -44.6
10 -28.3 -11.8 -54.7 -22.2 10 -10.2 -44.2
max -19.9 -14.4 -54.2 -19.7 max -17.2 -3.0
Similar analysis was also carried out for the dynamic strains. The maximum amplitude
of dynamic strain for each component of the strain was obtained and compared with the
corresponding strain for the baseline case. The final result obtained is shown in Table
D-9, Table D-10 and Table D-11.
272
Table D-9: Alternating Strains for the Baseline Case (Units: με)
section ε11 ε22 ε12 ε13

1 271.1 279.5 813.8 624.9
2 226.3 229.2 511.5 324.9
3 434.2 425.4 518.5 876.2
4 447.2 448.1 615.7 950.6
5 406.4 415.6 593.4 857.7
6 382.4 394.9 586.4 798.8
7 402.1 406.6 490.3 766
8 638.1 593.3 432 814.4
9 283.4 299.6 248.4 537.8
10 267.6 274 189.8 511.5
Max 638.1 593.3 813.8 950.6
Table D-10: Alternating Strains for the Final Design (Units: με)
Section ε11 ε22 ε12 ε13

1 128.7 133.1 468.1 324.5
2 98.3 100.9 236.9 149.5
3 135.8 257.4 209.5 457.8
4 140.3 261.4 239.9 489.7
5 121.1 265.5 212.6 440.5
6 117.4 236.8 217.1 401
7 83.1 263.5 232.3 267.8
8 164.9 183.4 195.5 681.1
9 92.2 115.7 169.7 107.4
10 73.3 84 137.1 87
Max 164.9 265.5 468.1 681.1
273
Table D-11: Percentage Difference for the Alternating Strains
Section ε11 ε22 ε12 ε13

1 -52.5 -52.4 -42.5 -48.1
2 -56.6 -56.0 -53.7 -54.0
3 -68.7 -39.5 -59.6 -47.8
4 -68.6 -41.7 -61.0 -48.5
5 -70.2 -36.1 -64.2 -48.6
6 -69.3 -40.0 -63.0 -49.8
7 -79.3 -35.2 -52.6 -65.0
8 -74.2 -69.1 -54.8 -16.4
9 -67.5 -61.4 -31.7 -80.0
10 -72.6 -69.3 -27.8 -83.0
Max -74.2 -55.3 -42.5 -28.4
D.4 Strength Test
Pull (tensile) test was conducted on sample section to test the strength for the layup
designed in the previous section. Sample sections were fabricated specifically for the pull
test with metal inserts (see Figure D-10) at the end such that it can be easily held in the
MTS testing machine. During the test, a pure tensile load was applied to simulate the
centrifugal force which is the most dominant force. Based on the computational analysis,
two critical areas were identified for structural testing, namely, the root section and the
cutout section for holding the actuators. Since, most of the structural strength of the blade
is due to the front spar, only spar part of the airfoil cross section was used for testing.
Only a small addition to strength and stiffness is expected by including fairing to the
airfoil section. However, its main purpose is to provide a smooth aerodynamic surface.
274
Figure D-10: CAD Model of the Metal insert used for Pull Tests
Pull Test for Root Section
For testing the root section, a 0.5R long blade section was fabricated with blade root at
one end and the metal insert at the other end. Metal insert was fixed inside the blade
section during the fabrication process to ensure that the joint has sufficient strength.
Sample section was then mounted on the tensile testing machine as shown in the Figure
D-11. In order to monitor blade strains during the tension test, 2 pairs of strain gages
were mounted on the top and bottom surface of the blade at 0.2R and 0.34R. As shown in
Figure D-12, this corresponds to section 4R and 1F in the blade design.
The tension test was performed in a quasi-static manner. The displacement at the
moving end was increased at the rate of 0.001 inch per sec. The load was allowed to
increase till the failure point. At tensile load of 4800 lbf, damage was observed in the
region where the metal insert was attached to the blade as shown in Figure D-13. (It
should be noted that, in this case, the hole used for attaching the metal insert to testing
machine was drilled in the metal insert after the blade was fabricated. As a result of
drilling this hole, some of the joint-strength between the rotor blade section and metal
275
insert was lost. And hence the damage occurred at the joint. This issue was avoided in the
next pull test by drilling the hole in the metal insert prior to fabrication.) During the test,
no damage was observed in rest of the blade. Thus the root section of the blade was able
to withstand tensile load of atleast 4800 lbf without any damage which is 20% higher
than the maximum load expected at the root at 100% RPM (13336 RPM) determined
from numerical analysis.
Figure D-11: Set up for pull test of blade root section
Figure D-12: Location of Strain Gages used for the Pull Test of Root Section
276
Figure D-13: Damaged Section after the Pull Test for Root Section
The variation of tensile load and blade strains with time is shown in Figure D-14 and
Figure D-15 respectively. The output produced by strain gage on top surface of section
4R shows some unexpected variation from the general trend which may be due to loose
electrical wiring. The strain observed for section 4R is consistently less than that
observed for section 1F as expected, since the root section has more plies. Also, for
section 1F, the strain gage on the top surface indicates higher strain as compared to the
strain gage on the bottom surface. This may be due to some asymmetry in the applied
load which can arise from the cambered profile of airfoil section. It should be noted that
the blade cross section consists of VR7 airfoil which is not a symmetric airfoil. In
general, all the strain components vary linearly with the applied load indicating that no
damage occurred to the blade section during the tension test. Thus, the blade root section
was able to withstand loads more than 20% of the maximum expected load without any
damage.
277
Load variation with time
5000
4000
3000
Load (lbf)
2000
1000
0
0 50 100 150
Time (sec)
Figure D-14: Blade Loading Profile used for the Pull Test of Root Section
Experimental Strain
4000
Sec4R Top
3500 Sec4R Bot
Sec1F Top
3000 Sec1F Bot
2500
Strain (  )
2000
1500
1000
500
0
0 1000 2000 3000 4000 5000
Load (lbf)
Figure D-15: Strain Recorded by different Strain Gages during the Blade Loading
Pull Test for Cut-out Section
A sample section was fabricated for pull test of cutout region with metal inserts at both
ends of the specimen as shown in Figure D-16. The sample section mounted on the
tensile testing machine is shown in Figure D-17. In this case, the metal inserts used had
drilled holes to avoid drilling holes after the blade fabrication. In order to monitor the
278
blade strain at critical locations, strain gages were mounted on the blade as shown in
Figure D-18.
Figure D-16: Test Section Fabricated for Pull Test of Cutout Section
Figure D-17: Setup used for the Pull Test for the Cutout Region
279
Figure D-18: Location of Strain Gages used for the Pull Test of Cutout region
In this test, the loading cycle was modified to observe the hysteresis effect in the
stress-strain curve. The applied axial load was increased to 1000lbf and then decreased by
500lbf and then increased by 1000lbf again till the maximum expected load of 3500lbf
was reached. After that, the blade load was allowed to increase steadily till 6500 lbf as
shown in Figure D-19. The blade strain given by strain gages is shown in Figure D-20.
All the strains observed show a linear relationship with the applied load and the effect of
alternating the load cycle is minimal. This indicates that no significant damage occurred
to the blade during the tensile test. As expected, higher strains are recorded by strain
gages 8, 3 and 4 which are mounted in the web region of actuator bay. In this region,
additional plies were included to reduce the effect of cutout in the cross section. The
maximum strain observed in this region is well below the maximum allowable strain
limit for the fiberglass material. The strain value observed by strain gages 7, 1 and 2 are
of similar magnitude. Also, as observed in the earlier pull test, there is some difference in
the strains recorded by the top and bottom strain gages at the same spanwise location.
280
Load variation with time
7000
6000
5000
4000
Load (lbf)
3000
2000
1000
-1000
0 200 400 600 800 1000
Time (sec)
Figure D-19: Loading Cycle used for Pull Test of Cutout Region
Axial Strain in Section 2F Axial Strain (5 & 6) in Section 2F in front region
9000 4000
Gage 7 Top(IN)
8000 Gage 8 3500 Bot(IN)
7000 3000
6000
2500
Strain (  )
Strain (  )
5000
2000
4000
1500
3000
2000 1000
1000 500
0 0
0 1000 2000 3000 4000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 7000
Load (lbf) Load (N)
Axial Strain (1 & 2)in Section 1F Axial Strain(3 & 4)in Section 2F in web region
5000 8000
Top(IN) Top (IN)
Bot(IN) 7000 Bot (IN)
4000
6000
5000
Strain (  )
Strain (  )
3000
4000
2000 3000
2000
1000
1000
0 0
0 1000 2000 3000 4000 5000 6000 7000 0 1000 2000 3000 4000 5000 6000 7000
Load (lbf) Load (lbf)
Figure D-20: Strain Gage Output
281
D.5 Blade Instrumentation
The active blade was instrumented with a variety of sensors to monitor the blade
response in real time. Instrumentation used in the active blade included strain gages to
determine blade strains, hall effects sensors to measure flap deflections and
accelerometers at the blade spanwise tip to measure tip twist and acceleration. For all the
sensors used in this blade, a 36AWG wire was used for making the wiring connections.
The wires were run along the blade spar and pulled out near the root. Detailed description
of all the sensors used on the active blade is given Table D-12 and Figure D-21.
Table D-12: Details of the Sensors used on the Active Blade
Strain Gages in the spar

Sensor Sensor Description Type Location (%R)
1 Chordwise Bending Strain Half 0.18
2 Flapwise Bending Strain Half 0.19
3 Torsional Strain Full 0.2
4 Flapwise Bending Strain Full 0.39
5 Flapwise Bending Strain Full 0.59
6 Chordwise Bending Strain Full 0.52
11 Axial Strain in Actuator bay 1 Quarter 0.75
12 Axial Strain in Actuator bay 2 Quarter 0.82
Accelerometers
Gage No Sensor Description Location (%c)
1 Near Leading edge 6%
2 At quarter chord 25%
3 Near Leading edge 44%
Hall Effect Sensor
No Sensor Description Location
(at 0.75c) (%R)
1 For actuator 1 72%
2 For actuator 2 79%
282
Active Blade Instrumentation (Part A)
Active Blade Instrumentation (Part B)

Figure D-21: Instrumentation used in the Active Blade
Besides the instrumentation wires, high voltage wires were also run along the blade
spar which were used to power the piezoelectric actuators. The high voltage wires were
shielded by aluminum foil.
D.6 Passive Blade design
The passive blade used in spin test was designed to have similar dynamic properties as
the active blade. Thus the passive blade also included the cutout region and similar
283
composite layup. In place of the actuator and flaps, ballast masses were used in the
passive blade to obtain similar inertia properties. As compared to the active blade profile,
only difference for the passive blade was in the flap region where it did not include any
cutout.
Detailed description of the fabrication process used for manufacturing active and
passive blade is provided in the Appendix E.
284
Appendix E. Blade Manufacturing
The manufacturing steps involved in the fabrication of the composite rotor blade are
similar to that highlighted in [58, 112]. In order to account for the presence of dual flaps,
some modifications were made, which will be discussed in detail in the following
sections.
The basic cross section of the rotor blade is shown in Figure E-1 and Figure E-2. It
consists of prepreg layers wrapped around the foam core with tungsten ballast mass at the
leading edge. The blade cross section is cured in two stages: namely, the spar cure and
the fairing cure. In the spar cure, the front spar of the blade is cured which also includes
the root section. In order to create space for mounting the actuators, spar includes cutouts
near the actuator location. Most of the instrumentation for active blade is included in the
spar region, thus it also houses the wires for transferring the sensor output to hub.
Similarly, the fairing includes cutouts for mounting flaps.
Figure E-1: Exploded View of the Airfoil Cross Section
285
Figure E-2: Assembled View of the Airfoil Cross Section
E.1 Fabrication of Foam Core
Foam core is required in the fabrication process to provide sufficient back pressure for
prepreg plies during the curing process inside the mold. The presence of foam core has
very little effect on the stiffness properties for the cross section; however, its effect is
more apparent on the inertial properties. In the cross-sectional analysis performed in this
chapter for the active-flap blade design, the effect of foam is included for all the blade
cross sections along the span. The shape of the foam section is determined using the
shape of outer mold line (OML), the number of prepreg layers used in the cross section
and the backpressure required for curing. To ensure sufficient backpressure, the foam
core used in spar section (71IG) was oversized by 5 mils while the foam core used in the
fairing section (31 IG) was oversized by 20 mils. These values were determined by
fabricating sample sections and observing the quality of cured parts.
286
Figure E-3: Shape of the Foam Core for Spar and Fairing Section
Initially, the CNC method was explored to fabricate the foam core sections. However,
it did not work well due to the cutting time required for getting a smooth finish and the
flexibility of foam section. As a result, a different method was used. Here, plexiglass
profiles were fabricated using a laser cutting machine. Laser cutting method provides
very high accuracy which is required for the foam core fabrication. These profiles are
attached on the either ends of a 6” long foam section and the foam was sanded using a
sanding machine and hand files. The root section of the blade has a non-uniform profile
which varies along the span. In order to accurately capture the non-uniformity, four
different sections were selected along the root part of the blade and the cross-sectional
shape was determined using the mold geometry. A tolerance of 5 mils in the thickness
was used for the fabrication of foam core. All the 6” foam pieces were joined together by
5min epoxy. The joined pieces of foam core in the root region are shown in Figure E-4.
Before joining the foam pieces, it was verified that the five-minute epoxy does not lead to
gassing of the foam section when heated inside autoclave at 250 deg F.
287
Figure E-4: Joined pieces of Foam Core
The foam core used for the spar section included cutouts as shown in the Figure E-5.
Similarly the foam core used for the fairing section included cutout in the flap region.
These cutouts were made with a sharp knife after joining all the foam pieces together.
Figure E-5: Cutout made in the Spar Region for Actuators
E.2 Instrumentation
The active blade included strain gages, HET sensors and accelerometers as the sensors
for measuring the blade deformation and flap deflection. Details and specifications of the
sensors and wires used in the blade are given in Table E-1. Wires for these sensors were
run along the trough made in the blade spar (shown in Figure E-7) and they exit near the
root as shown in Figure E-6. For installing the strain gages, precured E-Glass tabs were
used and strain gages were glued on them. These tabs with strain gages were glued upside
288
down on the foam core such that the strain gage records strain for the innermost layer of
the prepreg. Wiring diagram for the flap-wise bending strain gages and the torsional
strain gages are shown in Figure E-8 and Figure E-9, respectively. The front spar also
included two accelerometers near the blade tip as shown in Figure E-11. Wires running
along the blade spar also included high voltage wires for providing power supply to the
actuators in the cutout region as shown in Figure E-10.
Figure E-6: Instrumented Spar Section
Table E-1: Instrumentation used in the Spar Section
Manufacturer Part Number

Strain gages
Flapwise Vishay Micro measurements
Chordwise Vishay Micro measurements
Torsion Vishay Micro measurements
Accelerometer Analog Devices ADXL 193
HET Micronas Hal 815
Wires Vishay Micro measurement 36 gage
289
Figure E-7: Trough made in the Blade Spar for Instrumentation Wires
Figure E-8: Wiring Diagram for Full Bridge Flap wise Bending Strain Gage
Figure E-9: Wiring diagram for Full Bridge Torsion Strain Gage
290
Figure E-10: High Voltage Wires in the Cutout Region
Figure E-11: Accelerometers mounted on the Blade Tip in the Spar Region
Calibration for the accelerometer was verified using the guidelines provided in the
datasheet. Figure E-12 shows different configuration which can be used to calibrate the
accelerometer. According to the datasheet, ADXL193 has a sensitivity of 8mV/g. Table
E-2 gives the output voltage measured for 2 accelerometers. The obtained results are
close to the sensitivity given in the datasheet. The Hall Effect transducer was calibrated
after the fabrication of the rotor blade.
291
Figure E-12: Different Configurations used for Calibration of Accelerometer
Table E-2: Measured Voltage from Accelerometers
Acc 1 Diff Acc 2 Diff

(V) (mV) (V) (mV)
Config 1 2.4977 -8.4 2.4979 -8.1
Config 2 2.5143 8.2 2.514 8
Config 3 2.5061 0 2.506 0
E.3 Spar mandrel Design
The blade includes two cutouts in the spar region for mounting actuators after the
blade fabrication. In order to create space for the actuators, two spar mandrels are used.
During the manufacturing process, inboard actuator support and outboard actuator
support are installed in the blade spar. Actual fabricated parts used during the blade
manufacturing are shown in Figure E-13. In order to fabricate these parts, detailed CAD
models for each of the component were prepared as shown in Figure E-14. Final parts
were fabricated in the Machine Shop (by Terry Larrow) in Department of Aerospace
Engineering. In order to ease the process of removing spar mandrel from the cured blade,
292
two ¼” x 20 threaded holes were made in the middle spar mandrel and one in the inboard
spacer as shown in Figure E-13.
Figure E-13: Spar Mandrel Parts
Figure E-14: CAD Model Developed for the Parts of Spar Mandrel
293
E.4 Spar Manufacture
Once the instrumented foam core was ready and the two spar mandrels were
fabricated, the layup process for the blade spar was started. Based on the cross-sectional
design finalized in Appendix D, all the plies were cut to the desired shape and size. The
root section of the blade included additional plies to withstand the large centrifugal force.
The root plies used in the cross section consisted of 0.52” wide IM7 ply strips which
wrap around the root pin. For each layer (which consisted of 4 strips), two of these layers
wrap around the root pin and cover the top surface of spar, while the other two layers
wrap around the root pin and cover the bottom surface of blade spar. Unlike the root
plies, the mail spar plies wrap (except Spar Ply 2) around the leading edge of airfoil
section. Dimensions of all the plies used in the layup are shown in Table E-3. The main
plies used in the cross section had to be cut on the top surface in actuator region. The
cutout made in the plies is shown in Figure E-15. Ply 2 in the spar plies is a 0 deg IM7
ply to provide additional axial stiffness against the centrifugal loads. Similar to the
process followed for the root plies, Spar 2 plies are split into 4 strips of 0.52” width and
they wrap around the root region. In the cutout region, a diversion is made in these plies,
which will be shown later. Based on the sizes given in Table E-3, all the plies are cut to
the exact size as shown in Figure E-16.
Table E-3: Dimension of the Plies cut prior to Blade Fabrication
Root Plies
Sr. No Name Ply Length Width Angle
1 SPD 10b 1 IM7 2.4 0.52 0
2 SPD 10b 2 IM7 2.4 0.52 0
3 SPD 10b 3 IM7 2.4 0.52 0
4 SPD 10b 4 IM7 2.4 0.52 0
294
5 SPD 10a 1 IM7 2.4 0.52 0
6 SPD 10a 2 IM7 2.4 0.52 0
7 SPD 10a 3 IM7 2.4 0.52 0
8 SPD 10a 4 IM7 2.4 0.52 0
9 SPD 9a 1 IM7 3 0.52 0

10 SPD 9a 2 IM7 3 0.52 0
11 SPD 9a 3 IM7 3 0.52 0
12 SPD 9a 4 IM7 3 0.52 0
13 SPD 9b 1 IM7 3 0.52 0

14 SPD 9b 2 IM7 3 0.52 0
15 SPD 9b 3 IM7 3 0.52 0
16 SPD 9b 4 IM7 3 0.52 0
17 SPD 8 1 IM7 3.5 0.52 0

18 SPD 8 2 IM7 3.5 0.52 0
19 SPD 8 3 IM7 3.5 0.52 0
20 SPD 8 4 IM7 3.5 0.52 0
21 SPD7 1 IM7 4.1 0.52 0

22 SPD7 2 IM7 4.1 0.52 0
23 SPD7 3 IM7 4.1 0.52 0
24 SPD7 4 IM7 4.1 0.52 0
25 SPD6 1 IM7 4.6 0.52 0

26 SPD6 2 IM7 4.6 0.52 0
27 SPD6 3 IM7 4.6 0.52 0
28 SPD6 4 IM7 4.6 0.52 0
29 SPD5 1 IM7 5.2 0.52 0

30 SPD5 2 IM7 5.2 0.52 0
31 SPD5 3 IM7 5.2 0.52 0
32 SPD5 4 IM7 5.2 0.52 0
33 SPD4 1 IM7 7.9 0.52 0

34 SPD4 2 IM7 7.9 0.52 0
35 SPD4 3 IM7 7.9 0.52 0
36 SPD4 4 IM7 7.9 0.52 0
37 SPD3 1 IM7 10.7 0.52 0

38 SPD3 2 IM7 10.7 0.52 0
39 SPD3 3 IM7 10.7 0.52 0
40 SPD3 4 IM7 10.7 0.52 0
41 SPD2 1 IM7 13.5 0.52 0
295
42 SPD2 2 IM7 13.5 0.52 0
43 SPD2 3 IM7 13.5 0.52 0
44 SPD2 4 IM7 13.5 0.52 0
45 SPD1 1 IM7 24.5 0.52 0

46 SPD1 2 IM7 24.5 0.52 0
47 SPD1 3 IM7 24.5 0.52 0
48 SPD1 4 IM7 24.5 0.52 0
SPD: Spar Double Ply
Web Plies
49 Web 1 IM7 51.5 1.51 -45
50 Web 2 IM7 51.5 1.52 45
51 Web 3 E120 51.5 1.53 45
52 Web 4 E120 51.5 1.54 45
53 Web 5 E120 51.5 1.55 45
54 Web D1 E120 8 1.5 45
55 Web D2 E120 8 1.5 45
Main Plies
56 Spar Ply 1 E120 51.5 4.3 0
57 Spar Ply 2_1 IM7 105.5 0.52 0
58 Spar Ply 2_2 IM7 105.5 0.52 0
59 Spar Ply 2_3 IM7 105.5 0.52 0
60 Spar Ply 2_4 IM7 105.5 0.52 0
61 Spar Ply 3 IM7 51.5 4.3 45
62 Spar Ply 4 IM7 51.5 4.3 -45
63 Spar Ply 5 IM7 51.5 4.3 0
64 Spar Ply 6 E120 51.5 4.3 0
Root Pin
65 Root pin IM7 50 0.85 0
296
38.45
34.25 38.2 42.4
1.170.92
0.25
Figure E-15: Cutout made in Spar ply 1, 3, 4, 5 and 6.
Figure E-16: Cut Prepreg Plies prior to Layup
Prior to the layup process, the foam core is heated in an oven for 30 min at 150F
temperature to remove all the moisture. In the next step, the adhesive film (pink colored
AF163-2U) is wrapped around the foam core. Adhesive film facilitates the bonding of
prepreg to the foam core. It has a thickness of 3 mils and has the same curing temperature
(250F) as the prepreg used in the cross section. Adhesive film wrapped around the root
region and the cutout region is shown in Figure E-17 and Figure E-19. The cutout region
also includes additional chordwise and spanwise plies to provide strength in the cutout
region. These plies are added to the foam core prior to complete layup as shown in Figure
E-18 and Figure E-19.
297
Figure E-17: Adhesive Film wrapped around Foam Core
Figure E-18: Additional Plies in the Cutout Region
Figure E-19: Cutout Region with Ribs
In the next step, leading edge weight is added to the spar layup. This weight is added
near the leading edge of the cross section such that the cross-sectional center of gravity
lies near the quarter chord of airfoil section. The amount of ballast mass required is
298
determined from the cross-sectional analysis performed using UM/VABS. The ballast
mass used in this research consists of tungsten rods with cross-sectional diameter of 0.04
inch and 0.125 inch. The leading edge weight (LEW) required and the number of
tungsten rods needed to obtain that weight is shown in Table E-4. These tungsten rods are
cut into 1 inch pieces (so that they do not result in additional cross-sectional stiffness)
and rolled inside the IM7 ply as shown in Figure E-20. Since the flap region (section 2F)
includes flap supports and flaps in the trailing edge region, higher leading weights are
required in this region. Leading edge weights are attached to the spar section after the
root layup is completed as shown in Figure E-23 and Figure E-24.
Table E-4: Ballast Mass used in each Cross Section
Section LEW Reqd 0.04 Dia 0.125 Dia LEW calc

(gm/m) (#) (#) (gm/m)
3R 0.014 5 0.111
2R 0.09 4 0.096
1R 0.143 7 0.148
1F 0.125 6 0.126
2F 0.35 2 2 0.355
3F 0.11 5 0.111
4F 0.108 5 0.111
In the next step, a root pin (required for alignment and for creating a 0.5” diameter
hole for mounting the blade in the spin test stand) is added to the spar foam with adhesive
film on it. The unidirectional IM7 ply is wrapped around the root pin such that the
mounting hole has sufficient stiffness and strength. Root pin with wrapped IM7 plies and
root plies added near the root section are shown in Figure E-21. As discussed earlier, root
plies consist of 0.52” wide strips and they wrap around the root pin. Figure E-22 shows
the root section when all the root plies are added to the root section.
299
Figure E-20: Leading Edge Ballast Mass
Figure E-21: Wrapping Root plies around the Root Pin
Figure E-22: All Root Plies wrapped around the Root Pin
300
After the root plies, leading edge weight and web plies are added to the spar as shown
in Figure E-23. The cutout section of the blade includes larger ballast mass (Figure E-24)
to account for the additional mass due to flap and flap supports in the trailing edge
region.
Figure E-23: Leading Edge Weights and Spar web plies
Figure E-24: Cutout Region with Spar Mandrel and Additional Leading Edge
Weights
Next, the main spar plies are added. As discussed earlier, the main spar plies include a
cut in the top part to account for the cutout. Figure E-25 and Figure E-26 show the first
main spar ply (E-Glass at 0 deg) in the cutout region and the root region, respectively. A
small modification had to be made in the main spar plies near the root region so that they
conform better to the tapered and non-uniform root section. Figure E-27 shows the fourth
main spar ply (IM7 ply at 45 deg) on the spar section. The unidirectional IM7 ply is
301
wrapped around the root pin like root plies. On the top surface, these plies are steadily
moved around the cutout region as shown in Figure E-28.
Figure E-25: First main Spar ply Wrapped Around Spar in the Cutout Region
Figure E-26: First Main Spar Ply near the Root Region
Figure E-27: 4th Spar Ply in the Cutout Region
302
Figure E-28: Unidirectional Plies moved around the Cutout Region
The cutout section of blade with all the main spar plies is shown in Figure E-29. It also
shows the spar mandrel in the cutout region. In order to ease the process of spar mandrel
removal from the cured spar, it is coated with the releasing agent (Frekote 700NC) and
then taped with Teflon. This also prevents the residual epoxy from the prepreg from
pouring into the threaded holes which are required for the removal of spar mandrel. The
lower surface of the inboard and outboard actuator support are cleaned and covered with
adhesive film to facilitate the attachment of supports to the lower surface of the cutout
region in blade spar. In the next step, peel ply is added on the top and bottom surface of
spar near the web region to create space for the overlapping fairing plies as shown in
Figure E-29. Similarly, a spacer is used on the top of cutout region to create space for the
spar cover which covers the actuators on blade. The bottom surface of the spar is shown
in Figure E-31. Besides the peel ply, it also includes holes to allow the alignment pins
from spar mandrel to pass through.
Figure E-29: Cutout Region with All Spar Plies
303
Figure E-30: cutout region with spacer and peel plies
Figure E-31: Bottom part of the cutout region
All the instrumentation wires coming out from the root region are passed through a
shrink tube as shown in Figure E-32. This prevents the prepreg epoxy from getting in
contact with the wires which can make them brittle. Wires are also covered with the flash
tape to prevent any kind of damage from rubbing which might happen while closing the
molds. The instrumentation wires exit from the mold as shown in Figure E-33 through
the cavity machined earlier in the mold.
304
Figure E-32: Root region before spar cure
Figure E-33: Instrumentation wires exiting the mold
The bottom and the top molds are closed with the heavy duty steel clamps as shown in
Figure E-34. The steel clamps are tightened such that the space between the molds is less
than 8 mils along the entire perimeter. The blade section inside the closed mold is cured
in a 6ft long autoclave at 250F for 90 min. In this research, an autoclave was used instead
305
of the traditional mold heater to provide a more uniform heating over the entire blade
span and achieve a better control over the temperature profile. A very small pressure of
10 psi was used during the cure such that it is sufficient to ensure uniform heat transfer
inside the autoclave without affecting the foam core in the blade cross section.
Figure E-34: Blade Molds Closed with Heavy Duty Steel Clamps
The cured blade spar is shown in Figure E-35 and Figure E-36. The peel ply and the
spacers after the cure are shown in Figure E-36. It also shows the Teflon-taped spar
mandrel in the cutout region which had to be removed. The spar mandrel was removed
using the threaded hole in the middle spar as shown in Figure E-37. During the removal
of spar mandrel, care is taken to make sure that the inboard and outboard actuators
supports are not affected and they remain fixed in the blade spar. And finally, the high
voltage wires in the cutout region are soldered to the solder taps on the wall of the cutout
region.
Figure E-35: Active Blade Spar after Cure
306
Figure E-36: Cutout Region after Cure
Figure E-37: Spar Region after removing Spar Mandrel
E.5 Fairing Manufacture
The process used in the faring cure is similar to that followed during the spar cure. In
this case, the cure is complicated by the fact that the accurately aligned flap supports need
to be installed in the flap region for holding the flaps. In order to help in the alignment,
holes and notches are machined in the lower blade mold as shown in the Figure E-38.
The position of these holes was fixed relative to the location of alignment holes used for
the spar mandrel. Before the fairing cure, the flap mandrel had to be designed and
fabricated which is used to create space for actual flaps. The CAD model of the flap
307
mandrel is shown in Figure E-39. It also includes extensions at the ends to hold the flap
supports during the fairing cure.
Figure E-38: Holes machined in the bottom mold for alignment of flaps during the
cure
Figure E-39: CAD model of the flap mandrel used during fairing cure
The flap mandrel was machined out of aluminum. The CAD model of the inboard and
outboard flap support that had to be installed on the blade during the fairing cure is
shown in Figure E-40. The flap supports were fabricated using water jet cutting to get the
precise shape. As in the case of spar foam, the fairing foam was also fabricated in 6 inch
308
long pieces and then joined together using the five-minute epoxy. The fairing foam had to
be tapered near the root region to follow the mold profile as shown in Figure E-41. The
fairing section includes wires from HET that are used to measure the flap deflection. Hall
Effect transducers (HETs) are mounted after the fairing cure to protect them from high
temperature during the curing process.
Figure E-40: Location of Flap Supports mounted in the Flap Region during Fairing
Cure
309
Figure E-41: Root Region with Fairing Foam Core
In order to attach the flap supports to blade section, super plies and ribs are used.
Figure E-42 shows superplies wrapped around the inboard and outboard flap supports.
This figure also shows the flap mandrel, the cut made in fairing foam core for the
mandrel and the wires for connecting HET to the flap support. The ribs that are used to
transfer the loads generated by flap supports to the blade spar are shown in Figure E-43.
This figure also shows the TE stiffener (0.3 inch wide unidirectional IM7 ply) which runs
along the blade span to provide longitudinal stiffness in the fairing region.
310
Figure E-42: Instrumented Flap Region for Fairing Cure with Flap Mandrel
Figure E-43: Flap Region with Additional Plies for holding Flap Supports
Fairing plies (IM7 plies at +45 deg and -45 deg) were added on the top as shown in
Figure E-44. As it can be seen in the figure, they overlap the cured spar region over a
width of 0.3 inch. Similar rib plies and trailing edge stiffener were added on the bottom
surface of the fairing foam core as shown in Figure E-45.
311
Figure E-44: Flap Region with Fairing Plies
Figure E-45: Bottom Part of the Fairing Region
Before closing the mold for final cure, a spacer was used to create space for the spar
cover in the cutout region and critical areas were covered with the flash tape to prevent
extra epoxy from seeping into the parts as shown in Figure E-46. Final fairing layup near
the root section is shown in Figure E-47. As in the case of spar cure, the instrumentation
wires were run along the machined cuts in the mold to prevent them from damage during
the curing process.
312
Figure E-46: Fairing region before Final Cure
Figure E-47: Root part of the Blade before Fairing Cure
The final cured blade obtained after the fairing cure is shown in Figure E-48. It shows
the flap supports attached to the fairing of blade and the flap mandrel which was used as
spacer. The flap mandrels were carefully removed so as not to damage the HET wires and
fixed flap supports as shown in Figure E-49.
313
Figure E-48: Flap part of the Blade after Cure
Figure E-49: Cured Active Blade
The second blade used for testing on the spin test stand did not include active flaps
and is referred as “passive blade”. The passive blade used for testing was designed to
have similar dynamic properties as the active blade to avoid undesirable loads due to the
blade dissimilarities. Thus, the passive blade had cutout as in the case of active blade and
314
it included ballast masses in the spar region and in the flap region to account for the
actuators and flaps, respectively. The passive blade also included one flapwise bending
strain gage, one torsional strain gage and one chordwise strain gage near the root region.
The final passive blade that was fabricated is shown in Figure E-50. The ballast masses
used in the spar region in place of the actuator are shown in Figure E-51.
Figure E-50: Cured Passive Blade
315
Figure E-51: Ballast mass added in passive blade instead of actuator
E.6 Fabrication of Active Flap
The fabrication process used for manufacturing the active flaps is similar to that used
for fabricating the active blade. It consisted of prepreg plies wrapped around the foam
core. It also included supports at the end to help in installation of the flaps on the active
blade. The cross-sectional shape of the active flap is shown in Figure E-52. The final
fabricated flap had a chord of 1.34 inch (~ 0.25c) and a span of 3.85 inch (~0.06R). As
discussed in Section C.2.1, in order to minimize the hinge moment generated by
aerodynamic forces, the active flap was designed to have an overhang. Based on the CFD
analysis carried out for the airfoil-flap, the location of flap hinge axis was fixed at
0.365cf.
Figure E-52: Cross-sectional Shape of the Flap
During the CAD assembly of the flap and flap hinge mechanism, it was noticed that
some part of the flap was interfering with the control rod which is used to transfer the
316
actuation from X-frame actuator to active flap. Thus, a small notch was made near the
leading edge of the flap as shown in Figure E-53 to avoid the interference.
Figure E-53: CAD Model of the Active Flap
The foam core for active flaps was prepared in a similar manner as the foam core in
blade spar and fairing were fabricated. Plexiglass template was designed and laser cut
based on the number of plies and oversizing required to get sufficient back pressure for
the mold cure. The flap section before the cure is shown in Figure E-54. The spacer used
to create space near the leading edge of the flap and flap horn used in the flap can be seen
in the picture. For curing the flaps, a new aluminum mold was designed and machined. It
included small cutouts on the sides to help in the alignment of flap supports during the
fabrication process as shown in Figure E-55. The active flap was also cured in the
autoclave as shown in Figure E-56. The final fabricated flaps are shown in Figure E-57.
317
Figure E-54: Flap section before cure
Figure E-55: Flap inside the Mold before Cure
Figure E-56: Flap inside the Autoclave for Cure
318
Figure E-57: Cured Flap Sections
E.7 Blade Section for Pull Test
For pull test, sample sections of the blade were fabricated with metal insert at the end.
The CAD model of the metal insert used in the pull test is shown in Figure E-58. As
shown in the figure, half of the metal insert resembles the shape of blade spar such that it
can be easily attached to the blade, while the other half is a flat rectangular extension,
where a hole is drilled, such that it can be easily attached to the tensile testing machine.
The drilling of the hole or any other machining required for the metal insert should be
performed prior to attaching the metal insert to the blade spar. In the next step, the metal
insert is attached to the end of foam core as shown in Figure E-59. Thus, the metal insert
becomes a part of the blade during the layup process.
319
Figure E-58: CAD model for the Metal Insert
Figure E-59: Metal Insert Attached to the Foam Core
The final parts fabricated for pull test included strain gages to measure strains during
testing. To allow the wires to pass through, a small grove was made in the metal insert
along the spar thickness. Sample section for tensile testing with metal inserts attached at
either ends is shown in Figure E-60.
320
Figure E-60: Fabricated Section for Pull Test
321
Appendix F. Results from the Dual Flap Experiments
This appendix provides a summary of the experimental results obtained from the tests
conducted on spin-test stand with dual active flaps. The results obtained are compared
with numerical analysis performed using RCAS code.
F.1 Introduction
The main aim of the experimental analysis was to measure the unsteady drag produced
by active flaps in rotating conditions. With this objective in mind, a composite rotor blade
with dual active flaps was designed and fabricated as described in Appendix C, Appendix
D and Appendix E. The active flaps on the rotor blade were actuated by a couple of X-
frame actuators developed at MIT. The characterization of the X-frame actuator and the
development and testing of flap-actuation mechanism is described in Appendix C of the
thesis. Once the blade was fabricated, it was tested on the spin test stand. Besides
determining the unsteady drag produced by active flaps, other objectives of the
experiment are to: a) test the effectiveness of dual active flaps in influencing vibratory
loads at the rotor hub and b) generate experimental database for comparison with
numerical analysis.
Summary of the Spin Test Stand
The UM/MIT spin stand facility was developed for testing a Mach–scaled two-bladed
rotor with a diameter of 10ft. As shown in Figure F-1, the test stand consists of a steel
322
frame in a pyramid configuration which houses an electric motor with a direct coupled
shaft which passes through a slip ring assembly. The base of the stand is isolated on
rubber cushions to attenuate transmissions of floor vibration. The main characteristics of
the spin-test stand are given in Table F-1.
Table F-1: Spin-test Stand Characteristics
Property Value
Hover speed (for Mach scaling) 1336 RPM
Max rotor power 150 hp
Lowest stand elastic mode > 150Hz
Flap articulation 0.0286R
Lag articulation 0.15R
Feathering degree of freedom clamped at 0.0673R
Number of slipring channel for sensor signal 138
Number of slipring channel for high voltage signal 28
Number of blades 2
Radius 60.619 in (5ft)
Blade Chord 5.388 in
Figure F-1: Spin-test Stand
323
The primary sensor used for the measurement of unsteady aerodynamic loads due to
flap oscillation is a six-axis JR3 load cell. It measures all the three forces and three
moments at the rotor hub in the rotating frame. The maximum load carrying capacity for
various load components of the load cell is given in Table F-2. Axes orientation for the
load cell is shown in Figure F-2.
Table F-2: Maximum loads
Load Component Max Load

Fx ± 300 lbf
Fy ± 300 lbf
Fz 2000 /-500 lbf
Mx ± 250 ft-lbf
My ±250 ft-lbf
Mz 200 / -800 ft-lbf
Note: 1. Data for Table F-2 is provided by the manufacturer

2. Except Fz and Mz, all other loads generated by the blades subtract at the hub
Figure F-2: Axes Convention for the Load Cell
During previous experiments conducted with the spin-test stand at the old MIT
location [112], basic accuracy and resolution of the load cell under rotating conditions
were characterized. Results are reproduced in Table F-3. Ideally, these calibration tests
need to be conducted again for the new facility at the University of Michigan. However,
it is expected that due to the improved flow conditions in the new facility, the flow
fluctuations will be smaller than that detected earlier and the results shown in Table F-3
are expected to improve.
324
Table F-3: Preliminary Accuracy and Resolution of Various Load Components
Load Component Accuracy Resolution

Fx 1.5 lbf 0.08 lbf
Fy 1.5 lbf 0.08 lbf
Fz 6.2 lbf 0.25 lbf
Mx 1.2 ft-lbf 0.06 ft-lbf
My 1.2 ft-lbf 0.06 ft-lbf
Mz 2.5 ft-lbf 0.05 ft-lbf
List of working Sensors
During the process of active blade fabrication and instrumentation, some of the
sensors were damaged. The list of all working sensors is given in Table F-4.
Table F-4: List of working Sensors

Strain gages in blade spar
Sensor Sensor Description Type Location (R) Status
1 Chordwise Bending Strain Half 0.18 Quarter bridge working
2 Flapwise Bending Strain Half 0.19 Working
3 Torsional Strain Full 0.2 Working
4 Flapwise Bending Strain Full 0.39 Half Bridge working
5 Flapwise Bending Strain Full 0.59 Working
6 Chordwise Bending Strain Full 0.52 Quarter bridge working
11 Axial Strain in Act Bay 1 Quar 0.75 Failed
12 Axial Strain in Act Bay 2 Quar 0.82 Failed
Note: Chordwise strain gages to measure axial strain in actuator bay were mounted again
during fabrication of blade fairing
Accelerometers (at blade tip) in spar

Sensor Sensor Description Location Status
(c)
13 Near Leading edge 0.06 Failed
14 At quarter chord 0.25 Yes
325
Strain gages (in fairing)
Gage No Sensor Description Type Location (R) Status
15 Chordwise Bending Strain Quarter 0.76 Working
Accelerometer (in spar)

Gage No Sensor Description Location (c) Status
18 Near Leading edge 0.44 Yes
Note: HET sensors mounted in the sample blade can get damaged due to the high
temperature and lack of sufficient space. Hence, HET wires were installed for HET
which come out near flap`s outboard support. Actual HET are soldered near the flap
supports after the blade is manufactured.
Data Acquisition and Flap Actuation
The data acquisition setup used for acquiring the data from spin test and power supply
setup used to power the two X-frame actuators is shown in Figure F-3. The data collected
from all the sensors on active and passive blades and the data from the load cell in
rotating frame are transferred to the fixed frame through slip-rings in the spin test-stand
hub. Data from the fixed frame on the spin-test stand is transferred through long
intertwined wires to the National Instruments Data Acquisition Box in control room, as
shown in the Figure F-3. The data acquisition, visualization and elementary post-
processing analysis were performed using the LABVIEW software. It allows the user to
observe the data in real time and store the acquired data at a desired sampling frequency
in a text file. The output obtained in the text file can be used for further post processing
using MATLAB. The LABVIEW was also programmed to include a warning system in
case the output from any of the sensors exceeds a limiting value. For example, if the
value of strain from one of the strain gages exceeds a critical value, the LABVIEW
software issues a warning.
326
Figure F-3: Setup for Data Acqisition and Power Supply
The spin test stand room included four cameras at various locations in the room which
can be used to record videos during the operation. Since the cameras had poor resolution,
they could not be used for image processing or further analysis. One of the cameras was
used for tracking purpose. Further details related to tracking and balancing is provided in
the following section. The videos from all the four cameras can be seen by the user in
real-time and it can also be recorded. One wireless camera was installed on the blade hub
and it was made to point towards the tip of active blade from the hub. This helped in
capturing the blade tip motion.
The two X-frame actuators used on the active blade were powered independently so
that the motion of both the flaps can be controlled independently. The power supply to
each actuator was provided by a set of function generator and amplifier. The function
generator used for both the actuators can be synced, in case their phase difference or
frequency needs to be controlled simultaneously. The 5V DC signal for the HET is
provided through a separate power supply as shown in the Figure F-3.
327
Static Balancing and Tracking
Static balancing is performed in order to make sure that the total mass on the two
blades used in the spin test stand are balanced. An aluminum fixture, where both the
active and the passive blades used in the testing can be mounted, was designed and
fabricated. The size of the fixture was the same as that of the spin-test stand hub. The
setup used for static balancing is shown in Figure F-4 and Figure F-5.
Additional Mass
Spirit Level
(Collar Sleeve)
Passive Blade Active Blade
Aluminum Fixture
Figure F-4: Aluminum Fixture for Static Balancing
Figure F-5: Static Balancing of Blades
The active blade used for static balancing includes both the flaps and spar cap, as
shown in Figure F-5. The additional ballast mass in the form of collar sleeve is used to
328
balance both the blades. The size of the collar sleeve is such that it can be easily fitted on
the pitch shaft assembly. The advantage of using a collar sleeve is that it allows balancing
of the two blades without adding any mass to the blade itself.
After the completion of static balancing, the collar sleeve was added to the pitch shaft.
In the next step, blade tracking (or dynamic balancing) needs to be performed in order to
make sure that the two blades are producing the same amount of lift. If the blades are not
tracked, the pitch angle for one of the blades is adjusted till the balancing criterion is met.
The procedure used for tracking is the same as that described in [112], where a
combination of laser and camera were used. The schematic of the blade tracking process
used in the current experiments is shown in Figure F-6. In this case, a high intensity laser
was used to point on the blades. Whenever one of the blades hit the laser, a dot is
produced at the point of contact and it is recorded by the camera. In case the blades are
not balanced (as in the case of blades shown in Figure F-6), there is a shift in the point of
contact, which can be easily seen on the camera output. If this occurs, then the blade
pitch angle for one of the blades is adjusted till the dots almost coincide.
329
Figure F-6: Schematic of Blade Tracking
F.2 Testing Process
As discussed above, the main aim of the tests was to determine the unsteady
aerodynamic loads produced by the rotor blades. In order to determine the vibratory
loads, following steps were followed:
Step 1: Obtain Baseline Data
In this case, the rotor was spun at 900RPM while the flap position was fixed at zero
degree. (For Mach scaling, the required rotor speed is 1336RPM. However, it was
observed that at very high speeds, it was not possible to obtain sufficient flap deflections.
Hence, the operating speed was reduced to 900 RPM). For this operating condition, the
hub loads generated were recorded. For all the tests conducted in this thesis, a sampling
frequency of 1000 samples per second was used. The data from hub load cell and other
sensors was recorded once the steady state was obtained. In the steady state condition, the
data was recorded for a period of 10 seconds that corresponds to 150 revolutions at 900
330
RPM. It should be noted that the default position of flaps (when flaps are installed on the
blade) is not zero (usually it is close to +3 to +4 degree depending on the amount of
prestress). Thus in order to bring this non-zero flap deflection to zero, a DC voltage
supply was given to the actuators through the waveform generator and amplifier. Thus,
the flaps were being controlled during the baseline tests. For the results presented in this
section, the blades had a collective pitch setting of 6.5 degree which resulted in C T =
0.0033 (CT/σ = 0.0584).
Step 2: Obtain data with flap oscillations
In this step, the flaps on active blade were actuated at a desired frequency and phase
difference. A sample test matrix for analysis at different actuation frequencies and phase
angles of actuation is shown in Table F-5. In the sample results shown later in this
Appendix, the data collected for Case 1 and Case 5 is used. As in the baseline case, the
rotor was spun to the nominal speed of 900RPM and the data was collected in steady
state condition for a period of 10 seconds.
Table F-5: Test Matrix for Analysis at Different Frequency
Freq\Phase 00 900 1800 2700

2/rev 1 2 3 4
3/rev 5 6 7 8
4/rev 9 10 11 12
5/rev 13 14 15 16
Step 3: Extraction of Vibratory Loads due to Flap
The data obtained from the spin test stand hub load cell was very noisy, as shown in
Figure F-7 (a) for the Fz component. The spin test stand also included a RPM counter in
the hub and the output from the counter was also recorded. This was used to plot the data
331
as a function of number of revolutions, as compared to plotting the data as a function of
time. For the analysis of periodic systems, it was observed that this technique is more
meaningful and eases the process of filtering noise. Next, the data collected for 150
revolutions was split into 10 sets where each set had the data for 15 revolutions. The data
observed for all the 10 sets is shown in Figure F-7 (b). And finally, the data obtained for
all the ten sets was averaged to get a single signal for 15 revolutions as shown in Figure
F-7 (c). The results show that this process helped in getting a cleaner signal which was
used for obtaining the mean value of the signal and frequency content using the FFT
option in MATLAB, as shown in Figure F-8. In this figure, top part (in red) shows FFT
of the raw data (shown in Figure F-7(a)), while the bottom part (in blue) shows FFT of
the time averaged data (shown in Figure F-7 (c)).
Figure F-7: Time Averaging of Data
332
5
Fz(lbf)
0
0 1 2 3 4 5 6 7 8 9 10
Freq (/rev)
5
Fz(lbf)
0
0 1 2 3 4 5 6 7 8 9 10
Freq (/rev)
Figure F-8: Effect of Averaging on FFT
5
Fz(lbf)
Step 4: System Identification

0
0 1 2 3 4 5 6 7 8 9 10
Freq (/rev)
In step 3 of the analysis, the relation between an input variable and output response
was obtained only at specific points in the design space. In order to obtain a transfer
function between the input variable and the response which can be used later for controls
related studies, a frequency sweep analysis is required. The process of obtaining an
empirical transfer function estimate (ETFE) using the frequency sweep analysis is
described below.
An empirical transfer function is the ratio of the FFT of the output signal to the FFT of
input signal. As observed earlier, the response obtained from the test stand includes large
noise and hence the ratio was also dominated by noise and thus, not very accurate. In a
typical frequency sweep test, the frequency of the actuation voltage provided to the
actuator was varied from 20Hz (1.33/rev) to 105Hz (7/rev) in a period of 10 seconds (one
chirp) at constant voltage amplitude (800Vp-p with an offset of 400V). After a gap of 2
seconds, the sine sweep signal was repeated again and 3 sets of data were obtained.
333
(Ideally, 10 sets of data would be more suitable to get a cleaner response, however in the
current tests it was observed that the flap deflection was starting to reduce after the third
repetition.) The collection of chirps was averaged in the frequency domain using the
cross and auto spectrum to eliminate some of the noise.
If the Fourier transform of the output and input signal are given by and ,
respectively, then the averaged cross spectrum is given by:
1 N
ˆyu   Yi ( f )Ui* ( f )
N i 1
where, is the complex conjugate of the control signal, and N is the total number of
chirps (N= 3 in this case). Similarly, the auto spectrum of the input is given by:
1 N
ûu  Ui ( f )Ui* ( f )
N i 1
Using these two spectrums, the average transfer function can be written as:
ˆyu
G1 ( f )  .
ûu
F.3 Sample Results
Here some sample results obtained from the spin test are shown for the purpose of
demonstration. After the blades were installed on the spin test stand, a number of tests
were carried out during the course of one year in trying to remove the noise in the signal,
improve the flap deflection at higher RPM, to obtain repetitive data, etc. The results
presented here highlight the issues faced during the testing of blade with dual active
flaps. In this section, the results are presented for the baseline case (Base-1 and Base-2
conducted at two different times) and Case 1 and Case 5 listed in Table F-5. As described
334
earlier, in Case 1, both the flaps were actuated at 2/rev frequency while in Case 5, flaps
were actuated at 3/rev frequency. In both the cases, the phase difference is zero degree. It
should be noted that all these tests were conducted on the same day.
Table F-6: Mean value of Flap Deflection
δ2,mean (deg)
Base-1 -0.72
Base-2 -0.67
Case 1 0.95
Case 5 1.09
The mean values of flap deflection for the baseline cases and for the cases with flap
oscillation are shown in Table F-6. For the baseline cases, the mean value of flap
deflection was set to 0 deg (approximately) at 100RPM by varying the DC voltage given
to the actuator and observing the realtime output shown by the data acquisition system.
Next the RPM was slowly increased to 900 and the flap deflection was recorded in the
steady state condition. For the cases with flap oscillation, a fixed input voltage of 800Vp-
p with an offset of 400V was provided to maximize the amplitude of flap deflection. As a
result of this, there is a difference in the mean value of flap deflection for the baseline
cases and the cases with flap oscillation. The FFT of the flap deflection response for both
the flaps for Case 1 and Case 5 is shown in Figure F-9. For each of the cases, the FFT
shows peak at the actuation frequency, as expected. The amplitude of flap deflection is
higher for Flap 2 for the results shown here. In general, this trend is not consistent and it
depends on the actuation frequency and the initial prestress in the flaps during the
installation.
335
Case 1 Case 5
3 3
 (deg)
1(deg)
2 2
1 1 1
0 0
0 2 4 6 8 0 2 4 6 8
Freq(/rev) Freq(/rev)
3 3
2(deg)
 (deg)
2 2
1 1
2
0 0
0 2 4 6 8 0 2 4 6 8
Figure F-9: FFT of the Flap Deflection

For the purpose of smoothing the data and to remove noise from the signal, the data
collected for 150 revolutions was split into 10 sets and averaged. Figure F-10 shows the
variation of mean value of the loads for each of these 10 sets for the baseline case. For all
the components, the results show noticeable variation in the mean value, which is
unexpected. The periodic variation in the result indicates the presence of some lower
order harmonics in the system.
-6 -13.5
Fx(lbf)
Fy(lbf)
-8 -14
-10 -14.5
0 2 4 6 8 10 0 2 4 6 8 10
Data Set Data Set
145 40
Mx(in-lbf)
Fz(lbf)
140 30
135 20
0 2 4 6 8 10 0 2 4 6 8 10
Data Set Data Set
0 -750
My(in-lbf)
Mz(in-lbf)
-10 -760
-20 -770
0 2 4 6 8 10 0 2 4 6 8 10
Data Set Data Set
Figure F-10: Variation of Mean Loads for Base1
The variation in the mean value of the loads for the two baseline cases and the cases
with flap oscillation is shown in Figure F-11. The error bars for these results were
336
obtained by using the data shown in Figure F-10. For the results shown here, no trend can
be seen between the mean loads obtained for the baseline cases and the cases with flap
oscillations. For some of the components, the variation between the baseline cases is
higher than that between the baseline and flap actuated cases.
-6 -13
-8
Fx(lbf)
Fy(lbf)
-14
-10
-12 -15
Base-1 Base-2 Case1 Case5 Base-1 Base-2 Case1 Case5
150 50
Mx(in-lbf)
145 40
Fz(lbf)
140 30
135 20
10 -720
My(in-lbf)
Mz(in-lbf)
0 -740
-10 -760
-20 -780
Figure F-11: Variation of Mean Loads for Different Cases
Next FFT is performed on the steady state response obtained for the baseline cases.
Since the tests were performed in hover condition for similar rotor blades (although, only
one of the blades has flaps, they are designed to be structurally and aerodynamically
similar when the flaps are not oscillating), ideally there should not be any vibratory loads
in Fz and Mz component. However, due to closed boundaries around the spin test stand
and ground effects, baseline cases show significant vibration. Thus, these vibratory loads
observed for the baseline cases represent noise in the system. (If the numerical
simulations include the effect of walls and boundaries around the spin test stand, it is
possible to capture these vibratory loads, as it was demonstrated by the research
performed at Georgia Tech). Vibratory loads at different frequency observed for the Fz
337
and Mz components for both the baseline cases are shown
Time Average in Figure F-12 and Figure
Data
F-13, respectively.
Base 1
5
Fz(lbf)
Time Average Data
0
0 2 4 6 8 10
Freq (/rev)
Base 2
5
Fz(lbf)
0
0 2 4 6 8 10
Freq (/rev)
Figure F-12: FFT of Fz for the Baseline Cases
Results for Fz show most dominant amplitude at 4/rev frequency. In case of “Base 1”
results, the contribution to vibratory loads from other harmonics is also significant. For
these two cases, the average amplitude of vibratory loads is approximately 3.2% of the
mean loads. Also, the hub loads show very small contribution from 1/rev frequency,
which implies that the blades are dynamically balanced.
The results obtained for Mz component shows large vibratory loads at 1/rev frequency,
which are unexpected since the blades are balanced. Besides the dominant 1/rev
component, there are small contributions from 2/rev and 6/rev component for both the
baseline cases. The amplitude of vibratory load for Mz component is approximately 4.1%
of the mean Mz load.
338
Time Average Data
Base 1
40
Mz(in-lb)
20 Time Average Data
0
0 2 4 6 8 10
Base
Freq 2
(/rev)
40
Mz(in-lb)
20
0
0 2 4 6 8 10
Freq (/rev)
Figure F-13: FFT of Mz for the Baseline Cases
In order to capture the effects of flap actuation, the amplitude of vibratory loads
obtained for the flap actuated cases is subtracted from the amplitude of vibratory loads
obtained for the baseline cases. For the results presented here, the difference is taken with
both the baseline cases shown in previous results. The difference in amplitude obtained
for the Fz component is shown in Figure F-14. The results obtained vary depending upon
the baseline case which was subtracted. When the result obtained for the Base 1 case was
subtracted, the results showed increase in vibration for 3/rev frequency loads and
decrease in vibration for 4/rev frequency loads, for both Case 1 and Case 5. Similarly,
when the amplitude corresponding to Base 2 case was subtracted, results showed large
increase in vibration at 3/rev frequency, for both Case 1 and case 5. Thus, the results
show that the trend observed for the difference of amplitude from the FFT analysis is
independent of the actuation frequency and the amplitude of vibration observed is within
the noise level obtained for the system.
339
4 FFT Case1-FFT Base1 4 FFT Case5-FFT Base1
2 2
Fz(lbf)
Fz(lbf)
0 0
-2 -2
0 5 10 0 5 10
4 FFT Case1-FFT Base2 4 FFT Case5-FFT Base2

2 2
Fz(lbf)
Fz(lbf)
0 0
-2 -2
0 5 10 0 5 10
Figure F-14: Difference in FFT for Fz Component

Figure F-15 shows the difference in vibration level between the baseline cases and the
flap actuated cases for Mz component. In all the cases, there is an increase in the
amplitude of Mz loads at 1/rev frequency. Other than that, small increase is observed for
2/rev frequency and 3/rev frequency loads. However, the difference in amplitudes for
these frequencies is much smaller than that observed for the 1/rev frequency. Thus, the
experimental results show that flap actuation leads to an increase in Mz loads at 1/rev
frequency independent of the actuation frequency.
30 30
FFT Case1-FFT Base1 FFT Case5-FFT Base1
20 20
Mz(in-lbf)
Mz(in-lbf)
10 10
0 0
-10 -10
0 5 10 0 5 10
30 30
20 20
Mz(in-lbf)
Mz(in-lbf)
10 10
0 0
-10 -10
0 5 10 0 5 10
Figure F-15: Difference in FFT for Mz component
340
F.4 Comparison with RCAS
For numerical analysis, a model of the blade with dual active flaps was implemented
in RCAS. As in the experimental setup, RCAS model included active flaps on only one
of the two blades. The structural analysis was performed using the non-linear beam
model while the aerodynamic analysis was performed using the Peters flexible airfoil
theory. The analysis also included dynamic inflow model. For RCAS analysis, trim
option was used where Fz force at the hub was trimmed to the average thrust obtained in
the experiments. The aerodynamic analysis also required table-lookup for aerodynamic
coefficients at different flap deflections. This table was generated using X-FOIL results.
The table obtained was also updated with the results obtained using CFD analysis for the
purpose of comparison.
The variation of mean Fz and Mz for all the cases is shown in Figure F-16, and the
results obtained are compared with the RCAS results. Since trim analysis is performed,
the thrust predicted by RCAS is same for all the cases. The torque predicted by RCAS is
less than the experimentally obtained torque by atleast 25%.
150
145
Fz(lbf)
140 Exp
RCAS
135
Set 1 Set 2 2/rev 3/rev
-500
Exp
Mz(in-lbf)
-600 RCAS
-700
-800
Set 1 Set 2 2/rev 3/rev
Figure F-16: Comparison for Mean Loads
341
The results for Case 1 (2/rev actuation frequency) for Fz component are shown in
Figure F-17. The RCAS results show vibratory loads at the actuation frequency only, that
is, at the 2/rev frequency. The amplitude of vibratory loads predicted by RCAS is of the
same order as the loads observed in experiments. However, the experimental results show
vibratory loads at 3/rev frequency. A similar trend is observed for Case 5 (3/rev actuation
frequency) (Figure F-18) where the RCAS results show vibratory loads at the actuation
frequency only, while the experimental results show increase in vibration for loads at
3/rev and 6/rev frequency and decrease in vibratory loads at 4/rev frequency. The amount
of increase and decrease in the amplitude depends upon the baseline case selected for the
analysis.
Similarly, in case of Mz component, the RCAS results in Figure F-19 and Figure F-20
show vibratory loads at the actuation frequency only and the amplitude of vibratory loads
predicted by RCAS is of the same order of magnitude as the 1/rev loads observed in
experiments.
342
Experimental Result RCAS Prediction
4 4
4 FFT Case1-FFT Base1 4 FFT Case5 -FFT Base1
f=2/rev
2 3.5 2 3.5
Fz(lbf)
Fz(lbf)
0 0
3 3
-2 -2
2.5 2.5
Fz Force(lb)
Fz Force(lb)
0 5 10 0 5 10
Freq(/rev) 2 Freq(/rev) 2
4 4
1.5 1.5
2 2
Fz(lbf)
Fz(lbf)
1 1
0 0
0.5 -2
-2 FFT Case1-FFT Base2 FFT Case5-FFT Base20.5
0 5 10 0 0 5 0 10
1 2 3 4 5 6 1 2 3
Oscillation Frequency (/rev) Oscillation
Figure F-17: Comparison for Fz component for Case 1

4 4
-FFT Base1 4 FFT Case5-FFT Base1f=2/rev f=3/rev
ase1
2 3.5 3.5
Fz(lbf)
0 3 3
-2
2.5 2.5
Fz Force(lb)
Fz Force(lb)
5 10 0 5 10
eq(/rev) 2 Freq(/rev) 2
4 1.5 1.5
2
Fz(lbf)
1 1
0
-2 0.5 0.5
ase1
-FFT Base2 FFT Case5-FFT Base2
5 10 0 0 5 10 0
1 2 3 4 5 6 1 2 3 4 5 6
eq(/rev) Freq(/rev)
Oscillation Frequency (/rev) Oscillation Frequency (/rev)
Figure F-18: Comparison for Fz component for Case 5
343
30 22 30 22
20 20 20 20
Mz(in-lbf)
Mz(in-lbf)
f=2/rev
18 18
10 10
16 16
0 0
14 14
-10 -10
0 5 10 0 5 1
Mz (in-lbf)
Mz (in-lbf)
12 12
10 10
30 30
FFT Case1-FFT Base2 8 FFT Case5-FFT Base2
8
20 20
Mz(in-lbf)
Mz(in-lbf)
6 6
10 10
4 4
0 0
2 2
-10 -10
0 5 10 0 0
1 2 3 4 5 6
5 0
1 2
1
Oscillation Frequency (/rev) Oscilla
Figure F-19: Comparison for Mz component for Case 1

30 22 22
FFT Base1 FFT Case5-FFT Base1
20 20 20
Mz(in-lbf)
f=2/rev f=3/rev
10 18 18
16 16
0
14 14
-10
10 0 5 10
Mz (in-lbf)
Mz (in-lbf)
12 12
ev) Freq(/rev)
10 10
30
FFT Base2 8 FFT Case5-FFT Base2 8
20
Mz(in-lbf)
6 6
10
4 4
0
2 2
-10
10 0 0 5
1 2 3 4 5 6
10 0
1 2 3 4 5 6
ev) Freq(/rev)
Oscillation Frequency (/rev) Oscillation Frequency (/rev)
Figure F-20: Comparison for Mz component for Case 5
344
As described earlier, a frequency sweep analysis was performed to determine the
transfer function between the actuation voltage and different responses. For these tests,
the frequency of flap actuation was varied from 20 to 105 Hz in an interval of 10 seconds
while the amplitude of input voltage was held constant and three sets of data were
collected. Here four different cases were considered, namely, a) both flaps oscillating in
phase, b) both flaps oscillating out of phase, c) only flap 1 was actuated and d) only flap 2
was actuated.
The transfer function obtained between the input voltage and the flap deflection for
Flap 1 and Flap 2 is shown in Figure F-21 and Figure F-22, respectively. In both the
cases, the results show the trend observed earlier where the amplitude of flap oscillation
increased with an increase in the actuation frequency. The results obtained for Flap 2 are
more consistent while the results for Flap 1 vary significantly depending upon the case.
Also, there is very good coherence throughout the range of actuation frequency indicating
good signal quality from the flap sensors. The variation of phase is similar for Flap 1 and
Flap 2 for all the cases and the mean value of phase angle is close to 90 deg.
The transfer function obtained for the hub loads Fz and Mz is shown in Figure F-23
and Figure F-24, respectively. Unlike the results obtained from the flap deflection sensor,
the results obtained from the load cell show significant amount of noise in the system due
to which the coherence for both these transfer functions is very poor. The Fz component
shows high amplitude at 3/rev frequency only, as it was observed in the experimental
results obtained earlier.
345
Flap 1(deg/V)
0.01
Both Flaps (In Phase)
Both Flaps (Out of Phase)
0.005
Flap 1 Only
0
1 2 3 4 5 6 7
1
Coherence
0.5
0
1 2 3 4 5 6 7
Phase (deg)
100
0
-100
1 2 3 4 5 6 7
Frequency(/rev)
Figure F-21: Flap Deflection to Input Voltage Transfer Function for Flap 1
Flap 2(deg/V)
0.02
0.01
Flap 2 Only
0
1 2 3 4 5 6 7
1
Coherence
0.5
0
1 2 3 4 5 6 7
Phase (deg)
100
0
-100
1 2 3 4 5 6 7
Frequency(/rev)
Figure F-22: Flap Deflection to Input Voltage Transfer Function for Flap 2
346
0.02
Fz(lbf/V)
0.01
0
1 2 3 4 (In Phase) 5
Both Flaps 6 7
1 Both Flaps (Out of Phase)
Coherence
Flap 1 Only
0.5 Flap 2 Only
0
1 2 3 4 5 6 7
Phase (deg)
100
0
-100
1 2 3 4 5 6 7
Frequency(/rev)
Figure F-23: Hub Thrust (Fz) to Input Voltage Transfer Function

0.1
Mz(in-lbf/V)

Flap 1 Only
0.05
Flap 2 Only
0
1 2 3 4 5 6 7
1
Coherence
0.5
0
1 2 3 4 5 6 7
Phase (deg)
100
0
-100
1 2 3 4 5 6 7
Frequency(/rev)
Figure F-24: Hub Torque (Mz) to Input Voltage Transfer Function
F.5 Conclusion
Finally, a summary of different issues experienced in the experimental analysis of dual
active flaps is provided along with possible corrections that can be made for future
experiments. The issues are listed in the order of their importance.
Issue 1: Baseline vibration in hub loads measurement
347
Since the tests were done in hover condition, it was expected that the vibratory loads
in the absence of any flap motion (baseline vibratory loads) would be very small.
However, during the experiments, it was noticed that the baseline vibratory loads were of
the same order of magnitude as the expected response loads due to flap motion (for a
torsionally stiff blade fabricated for these experiments).
Possible Causes
a) Ground Effect: The tests described in this thesis were conducted in the spin test
stand located in Aerospace Department at the University of Michigan. The rotor
blades have a diameter of 10ft and the blades are located at a height of 7.5ft above
the ground. The standard recommended distance between the blades in spin test
stand and ground plane in order to avoid ground effect is around 1.5D (15ft in the
current setup). As a result of close proximity of the blades to the ground, the
blades experience large unsteady motion of air. Even though the setup includes a
bell mouth around the spin test stand and a wire mesh above it in order to
smoothen the flow coming on to the rotor blade, hub loads still showed sufficient
vibratory amplitude.
b) Proximity to walls around the room: As mentioned earlier, the spin test stand is
enclosed in a closed room. Hence, there is a significant amount of reflection of the
vortices generated by blade motion from the four walls enclosing the room.
Preliminary analysis performed at GT using CFD (see Figure F-25) showed that
the 4/rev vibratory loads are generated due to walls around the spin test stand. The
amplitude of vibratory load corresponding to 4/rev frequency shown in Figure
348
F-25 (ΔCT ~ 0.075x10-3 , that is, ΔFZ ~ 7.1 lbf) is very close to the value obtained
experimentally at 1336 RPM (ΔFZ ~ 6.0 lbf).
Corrections:
1) Increasing the height of the spin test stand above the ground can result in some
improvement in the baseline vibratory loads. The increase in vibratory loads at the
hub due to proximity to the ground plane can be investigated in more details using
RCAS.
2) The four walls around the room lead to the recirculation of air inside the room. It
is difficult to quantify the effect of these four walls using aeroelastic codes like
RCAS and AVINOR. However, the effect of walls can be determined using CFD
as demonstrated in the results shown in Figure F-25. Thus, in order to minimize
these effects, it is important to perform these experiments in open conditions or in
a wide wind tunnel. However, before performing any experiment, it is always
useful to quantify the baseline vibration using CFD and experimental analysis.
3) One way to reduce the recirculation of air in the test stand room is to keep both
the doors open in spin test stand room. (provided it is safe to do it)
Cost Analysis
The main sources of vibration are the four walls enclosing the spin test stand and
proximity to ground. Hence moving the spin test stand to a new facility would be useful.
However, before finalizing the facility, detailed analysis need to be conducted to
determine baseline vibration in the test facility and expected vibratory loads from the flap
motion. In the current setup, the only possible way to minimize baseline vibration is to
perform the experiment with open doors. Increasing the height of the spin test stand
349
would also be useful in minimizing the baseline vibration. For increasing the height,
using a long shaft between motor and blade would be more useful than raising the height
of the entire test stand.
Figure F-25: Variation of CT for Baseline Conditions [177]
Issue 2: Poor performance of flaps (flap deflection) at high RPM
It was observed during the experiments that the flap deflections decrease significantly
(amplitude less than 0.5 deg) when the RPM is above 1000.
Possible causes:
a) Gaps between the airfoil and flap
350
The flap was designed to have a certain gap between the main blade and the
leading edge of the flap in the chordwise direction. However during the
manufacturing process, there was small variation in size of these gaps. Also, the
analysis performed in CFD to determine the flap-hinge location did not account
for gaps along the spanwise direction. Since the flap supports are installed at the
ends of the flap, there is some gap between the main blade and the flap in the
spanwise direction. Even though an attempt was made to cover up these gaps
using E-Glass tabs, it was not exactly flush with the flap. As a result of this, there
was a reduction in the flap effectiveness and an increase in the additional
aerodynamic forces acting on the flaps which increase with RPM. This can result
in a decrease in flap deflection obtained.
b) Large friction at high RPM
During the design process, special care was taken to reduce friction from the flap
actuation mechanism. All the supporting parts were fabricated with steel to ensure
that there is very little compliance in the system; and flap and its supports can
sustain high centrifugal force. However, the reduction in flap deflection indicates
that there is a large increase in friction with the increase in RPM (besides the
increase in aerodynamic forces against which the flap needs to move).
Correction:
a) In the current setup, it would be difficult to cover up the gaps along the spanwise
direction. However, for future flap installation, it would be more suitable to have
the flap supported in the middle so that gaps at the end can be minimized.
351
b) Also, it is important to account for the effect of gaps (in both spanwise and
chordwise directions) on the unsteady loads acting on flaps and on reduction in
the flap effectiveness. This requires detailed CFD analysis of the flap region in
rotating conditions.
c) The use of bearing in the clevis region would be useful for reducing some of the
friction effects. However, it was not possible to include the bearing in the current
design due to limited space available and size constraint imposed by maximum
length of moment arm.
Cost Analysis:
In the short term (1-2 weeks), a better way to cover the gaps in both spanwise and
chordwise direction can be designed without getting in contact with flaps. However,
before going ahead with that, detailed aerodynamic analysis needs to be performed in
order to determine effect of gap size on flap effectiveness. For future active flap
experiments, the flap hinge mechanism needs to be redesigned to reduce gaps and include
ball bearings in the clevis. In the current flap-actuation mechanism designed, many
constraints were imposed due to the requirement of prestress for piezoelectric actuator
that was generated using the prestress wire in active flap. However, other actuator
available (like Cedrat) can generate required prestress from the actuator frame and hence
do not require any external mechanism to generate prestress.
Issue 3: Very small effect of flap motion on vibratory loads at the hub
Flap oscillations at different actuation frequencies led to very small increase in the
vibratory loads at the hub at that frequency.
Possible Cause
352
a) Large torsion stiffness of the blade
The main aim of the experimental analysis was to determine the unsteady
aerodynamic drag associated with the flap motion. In order to minimize the
contribution from blade elasticity to loads generated at the hub (aeroelastic loads),
the active blade was designed to be stiff in torsion. The blade designed had a
torsional stiffness of 5.88/rev at 1336 RPM. Due to the reduction in flap motion at
high RPM (Issue 1), the operating RPM was changed to 900 in order to ensure
sufficient flap deflection. However, decreasing the RPM caused the blade torsion
frequency to become 8.88/rev. Thus, due to the large torsional stiffness of the
blade, the twisting moment generated by the active flaps was not sufficient to
twist the blade and thereby generate aerodynamic loads.
b) Gap between the airfoil and flap in the spanwise direction.
As discussed previously, the 3D air flow over the gaps can result in reduction in
the flap effectiveness. The current 2D codes can only account for gaps in the
chordwise directions.
c) Walls enclosing the spin test stand
The spin test stand used for experiments is enclosed inside a room and as a result,
there is recirculation of air after reflection from the walls. In the CFD study at
GT[177], it was shown that there is approximately 22% reduction in CT because
of the shroud and walls around the spin test stand. The effect of walls on the
vibratory loads generated at the hub due to flap oscillations is yet to be quantified.
Correction:
353
a) Reducing blade torsional stiffness at the root: This can be achieved by using a
torsional spring/coupling between the blade root and root attachment on the test
stand. This might result in a small increase in radius of the blade. However, the
increase in loads using this technique will be due to aeroelastic effects and might
be detrimental to original goal of the experiments.
Cost Analysis:
In short term, a torsional spring can be used to reduce torsional stiffness of the blade
locally at root. However, for future active flap experiments, blades need to be designed
appropriately.
Issue 4: Large variability in results (for similar flap deflection and operating RPM)
Vibratory hub loads generated at the rotor hub showed significant variation in the
mean value and amplitude of vibration, even during a single run. During the tests, the
data was collected for 150 revolutions in steady state condition at a sampling rate of 1000
samples/sec. In order to analyze the data, this data was split in to 10 sets of 15 revolutions
each. The mean value and per rev harmonic loads for each of these intervals were
obtained to determine the fluctuations in the test stand hub loads. Results obtained for the
mean value of loads indicated the presence of lower order harmonics (less than 1 Hz
frequency) in the hub loads.
Possible Causes:
1) This can be attributed to the presence of ground and walls around the spin test
stand. It may be possible to capture some of these effects with CFD analysis;
however, the analysis needs to be runs for atleast 200 or more revolutions to
capture these effects.
354
Correction:
1) This can also be fixed by using a wide wind tunnel or by performing the tests in
open conditions.
Issue 5: Uncontrollable flap deflections
During the experimental runs for the baseline case, the flap position was adjusted to 0
deg at 100RPM and then the RPM was increased to operating RPM (900 RPM). While
increasing the RPM, there was small variation in the flap position (between ±1 deg). For
the cases where flap was actuated, the input voltage was fixed to 800 Vp-p with an offset
of 400V in order to maximize the amplitude of flap oscillation. However, due to this
input voltage, it was difficult to control the mean value of flap oscillation.
Possible Causes:
a) Due to increase in dynamic pressure with increase in RPM, the aerodynamic
forces acting on the flap hinge change. As a result of this, the mean value of the
flap position changes with an increase in RPM for the baseline cases.
b) A significant amount of static friction was observed in the flap hinge mechanism.
With the same amount of prestress applied (by adjusting the clevis), different flap
deflections were obtained (variation of 1-2 deg). Also, the waveform generator
used for generating the input signal for active flaps produced small spikes. These
spikes also resulted in shift in the mean position of the flap.
c) The range for voltage supply that can be given to the X-frame actuator was fixed
by the manufacturer (from 0V to 800V). Thus, in order to maximize the amplitude
of flap deflection, complete range of voltage supply had to be utilized. If offset
355
voltage was varied to adjust the mean position of flap oscillation, it would have
resulted in reduction in the amplitude of flap oscillation.
Corrections:
1) Feedback control for controlling flap deflection: In order to accurately control the
flap deflection during the tests, a feedback loop is required. In the recent tests
with active flaps (at Boeing and Sikorsky), a controller based on HHC algorithm
is used to control the flap position.
2) There is a need to minimize static friction in the flap hinge mechanism by
appropriate lubrication.
3) There is a need to redesign the flap-actuation mechanism such that higher flap
deflection could be obtained. This would allow the controller to control both the
mean value and amplitude of flap deflection.
Cost Analysis:
Developing a feedback controller with basic PID approach would be useful in
controlling the flap deflection when the flaps are not oscillating. Also, an advanced
controller based on HHC would be required to control the flap deflection in rotating
conditions.
Issue 6: Decrease in flap deflection with time
Flap deflection was observed to reduce with time during the experimental runs. At full
RPM, sufficient flap deflection could only be obtained for less than 2 minutes. After each
run, rotor had to be stopped and lubrication had to be applied. In some cases, flap had to
be removed from the blade and installed again.
356
Possible Cause:
a) Loss of lubrication: The increase in friction force can also occur due to the loss of
lubrication which occurs because of large centrifugal force.
b) Small deformation of flap supports: The large centrifugal force acting on the flap,
can lead to small deformation of the flaps and its support over a period of time.
This might cause some components of the flap and its support to get in contact
with non-moving parts and rub against each other. This can result in significant
increase in friction and reduce the flap deflection.
Correction:
1) Lubrication: During the tests, it was observed that using lubrication after every
run helped in reducing the friction effects.
2) Reinstallation of the flaps on active blade: The flap lubrication was effective only
for 3-4 runs and after that; even lubrication did not produce any improvement in
the flap deflection. In this case, flaps had to be removed from the setup and
reinstalled again. This helped in recovering the flap deflection.
3) The need for lubrication needs to be minimized in future tests to ensure longer test
time by reducing the number of moving parts and by use of elastic hinges.
Issue 7: Unexpected large 1/rev component in Mz
The experimental results obtained show significant 1/rev component in M z. Also, this
component increased when flaps were oscillated. The amplitude of 1/rev component was
larger when the tests were conducted at 2 deg collective setting as compared to tests
conducted at 6.5 deg collective.
357
Issue 8: Decrease in hub load response with increase in flap actuation frequency
Computational results obtained from RCAS and AVINOR showed an increase in
vibratory loads with increase in the actuation frequency till it approaches the first torsion
frequency. However, the experimental results obtained showed that flaps are effective in
generating vibratory loads at the hub only when the actuation frequency is between 2/rev
to 4/rev even though a higher amplitude of flap deflection is obtained at higher flap
actuation frequencies.
358
REFERENCES
359
[1] A. C. Veca. "Vibration Effects on Helicopter Reliability and Maintainability."
Sikorsky Aircraft Division, AD-766 307, April, 1973.
[2] R. Heffernan, D. Precetti and W. Johnson. "Vibration Analysis of the SA349/2

Helicopter." NASA, Technical Memorandum 102794, 1991.
[3] G. Reichert. "Helicopter Vibration Control: A survey." Vertica, Vol. 5, No. 1,

1981, pp. 1-20.
[4] O. Dieterich. "Application of Modern Control Technology for Advanced IBC

Systems." 24th European Rotorcraft Forum, Marseilles, France, 15-17
September, 1998.
[5] E. Breitbach and A. Buter. "The main Sources of Helicopter Vibration and Noise
Emissions and Adaptive Concepts to reduce them." Journal of Structural Control,
Vol. 3, No. 1-2, June, 1996, pp. 21-32.
[6] H. Strehlow, R. Mehlhose and P. Znika. "Passive and Active Vibration Control
Activities in the German Helicopter Industry." Aerospace Technology Exhibition
and Conference (AEROTECH), Birmingham, 14-17 January, 1992.
[7] K. S. Brentner and F. Farassat. "Modeling Aerodynamically Generated Sound of

Helicopter Rotors." Progress in Aerospace Sciences, Vol. 39, 2003, pp. 83-120.
[8] F. H. Schmitz. "Rotor Noise." Aeroacoustics of Flight Vehicles, Theory and

Practice - Vol 1: Langley Research Center, 1995, pp. 65-49.
[9] B. Edwards and C. Cox. "Revolutionary Concepts for Helicopter Noise Reduction
- S.I.L.E.N.T Program." NASA, CR-2002-211650, 2002.
[10] W. Gerstenberger. "The Rotary Round Table: Practical Aspects of Vibration

Control." Journal of the American Helicopter Society, Vol. 2, No. 3, 1957, pp. 3-
4.
[11] R. G. Loewy. "Helicopter Vibrations: A Technological Perspective." Journal of

American Helicopter Society, Vol. 29, No. 4, 1984, pp. 4-30.
[12] R. L. Bielawa. "Analytical investigation of Helicopter Rotor Blade appended

Aeroelastic devices." NASA, Technical Report 166525, 1984.
[13] P. Konstanzer, B. Enenkl, P.-A. Aubourg and P. Cranga. "Recent advances in

Eurocopter’s Passive and Active Vibration Control." American Helicopter Society
64th Annual Forum, Montreal, Canada, April 29-May 1, 2008.
360
[14] M.-N. H. Hamouda and G. A. Pierce. "Helicopter Vibration Suppression Using
Simple Pendulum Absorbers on the Rotor Blade." Journal of American
Helicopter Society, Vol. 29, No. 1, 1994, pp. 19-29.
[15] R. Ganguli. "Survey of Recent Developments in Rotorcraft Design Optimization."

Journal of Aircraft, Vol. 41, No. 3, 2004, pp. 493-507.
[16] P. P. Friedmann. "Helicopter Vibration Reduction Using Structural Optimization

with Aeroelastic/Multidisciplinary Constraints." Journal of Aircraft, Vol. 28, No.
1, 1991, pp. 8-21.
[17] R. Celi. "Recent Applications of Design Optimization to Rotorcraft - A survey."

Journal of Aircraft, Vol. 36, No. 1, 1999, pp. 176-189.
[18] P. P. Friedmann and T. A. Millot. "Vibration Reduction in Rotorcraft Using

Active Control: A Comparison of various Approaches." Journal of Guidance,
Control and Dynamics, Vol. 18, No. 4, 1995, pp. 664-673.
[19] C. Kessler. "Active Rotor Control for Helicopters: Motivation and Survey on
Higher Harmonic Control." CEAS Aeronautical Journal, Vol. 1, No. 1, 2011, pp.
3-22.
[20] C. Kessler. "Active Rotor Control for Helicopters: Individual Blade Control and
Swashplateless Rotor Designs." CEAS Aeronautical Journal, Vol. 1, No. 1-4,
2011, pp. 23-54.
[21] S. P. King and A. E. Staple. "Minimization of Helicopter Vibration Through

Active Control of Structural Response." Rotorcraft Design Operations, Vol. CP-
423, AGARD, October, 1986.
[22] J. T. Pearson, R. M. Goodall and I. Lyndon. "Active Control of Helicopter

Vibration." Computing and Control Engineering Journal, Vol. 5, 1994, pp. 277-
284.
[23] R. C. Cribbs, P. P. Friedmann and T. Chiu. "Coupled Helicopter Rotor/Flexible

Fuselage Aeroelastic Model for Control of Structural Response." AIAA Journal,
Vol. 38, No. 10, 2000, pp. 1777-1787.
[24] A. Z. Lemnios and A. F. Smith. "An Analytical Evaluation of the Controllable

Twist Rotor Performance." Kaman Aerospace Corporation, USAAMRDL
Technical Report 72-16, May, 1972.
[25] A. Z. Lemnios. "Wind Tunnel Investigation of the Controllable Twist Rotor

Performance and Dynamic Behavior." Kaman Aerospace Corporation,
USAAMRDL Technical Report 77-10, 1977.
361
[26] J. Shaw. "Higher Harmonic Blade pitch Control for Helicopter Vibration
Reduction: A Feasibility Study." MIT, ASRL-TR-150-1, 1968.
[27] J. Shaw and N. Albion. "Active Control of the Helicopter Rotor for Vibration
Reduction." Journal of American Helicopter Society, Vol. 26, No. 3, 1981, pp.
32-39.
[28] K. Nguyen and I. Chopra. "Effects of Higher Harmonic Control on Rotor

Performance and Control Loads " Journal of Aircraft, Vol. 29 No. 3, 1992, pp.
336-342.
[29] E. R. Woods, R. W. Powers, J. H. Cline and C. E. Hammond. "On Developing

and Flight Testing a Higher Harmonic Control System " Journal of American
[30] C. E. Hammond. "Wind Tunnel Results Showing Rotor Vibratory Loads

Reduction Using Higher Harmonic Blade Pitch " Journal of American Helicopter
Society, Vol. 28, No. 1, 1983, pp. 10-15.
[31] K. Nguyen, M. Betzina and C. Kitaplioglu. "Full-Scale Demonstration of Higher

Harmonic Control for Noise and Vibration Reduction on the XV-15 Rotor."
American Helicopter Society 56th Annual Forum, Virginia Beach, Virginia, May
2-4, 2000.
[32] J. Shaw, N. Albion, E. J. Hanker and R. S. Teal. "Higher Harmonic Control: Wind
Tunnel Demonstration of Fully Effective Vibratory Hub Force Suppression."
Journal of American Helicopter Society, Vol. 34, No. 1, 1989, pp. 14-25.
[33] M. Kretz. "Research in Multicyclic and Active Control of Rotary Wings."

Vertica, Vol. 1, No. 2, 1976, pp. 95-105.
[34] N. D. Ham. "Helicopter Individual-blade Control Research at MIT 1977-1985."

Vertica, Vol. 11, No. 1-2, 1987, pp. 109-122.
[35] S. A. Jacklin, A. Blaas, D. Teves and R. Kube. "Reduction of Helicopter BVI

Noise, Vibration, and Power Consumption Through Individual Blade Control."
Proceedings of the 51st Annual Forum of the American Helicopter Society, Fort
Worth, Texas, May 9-11, 1995.
[36] S. A. Jacklin, A. Haber, G. d. Simone, T. Norman, C. Kitaplioglu and P. Shinoda.

"Full-Scale Wind Tunnel Test of an Individual Blade Control System for a UH-60
Helicopter ", American Helicopter Society 58th Annual Forum, Montreal, Canada,
11-13 June, 2002.
[37] U. T. P. Arnold and D. Fürst. "Closed loop IBC results from CH-53G flight tests."
Aerospace Science and Technology, Vol. 9, 2005, pp. 421-435.
362
[38] V. Giurgiutiu. "Review of Smart-Materials Actuation Solutions for Aeroelastic
and Vibration Control." Journal of Intelligent Material Systems and Structures,
Vol. 11, July, 2000, pp. 525-544.
[39] V. Giurgiutiu. "Active-Materials Induced-Strain Actuation for Aeroelastic

Vibration Control." The Shock and Vibration Digest, Vol. 32, No. 5, September,
2000, pp. 355-368.
[40] V. Giurgiutiu. "Power and Energy issues in the Induced strain actuation for
Aerospace Adaptive Control." AIAA/ASME/AHS Adaptive Structures Forum, Salt
Lake City, UT, 18-19 April, 1996.
[41] V. Giurgiutiu. "Recent Advances in Smart Material Rotor Control Actuation."

41st AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials
Conference Atlanta, Georgia, 3-6 April, 2000.
[42] F. Straub. "A Feasibility Study of using Smart Materials for Rotor Control."
Smart Materials and Structures, Vol. 5, 1996, pp. 1-10.
[43] F. Straub, M. A. Ealey and L. M. Schetky. "Application of Smart Material to

Helicopter Rotor Active Control." SPIE`s Symposium on Smart Structures and
Materials, San Diego, 3-6 March, 1997.
[44] I. Chopra. "Status of Application of Smart Structures Technology to Rotorcraft

Systems." Journal of American Helicopter Society, Vol. 45, No. 4, 2000, pp. 228-
252.
[45] P. P. Friedmann. "Rotary-wing Aeroelastic Scaling and its Implications for

Adaptive-materials-based Actuation." SPIE Conference on Smart Structures and
lntegrated Systems, Newport Beach, California, 1 March, 1999.
[46] O. Dieterich, B. Enenkl and D. Roth. "Trailing Edge Flaps for Active Rotor
Control Aeroelastic Characteristics of the ADASYS Rotor System." American
Helicopter Society 62nd Annual Forum, Phoenix, AZ, May 9-11, 2006.
[47] P. P. Friedmann. "Rotary-Wing Aeroelasticity: Current Status and Future

Trends." AIAA Journal, Vol. 42, No. 10, 2004, pp. 1953-1972.
[48] P. P. Friedmann, M. de Terlizzi and T. F. Myrtle. "New Developments in

Vibration Reduction with Actively Controlled Trailing Edge Flaps."
Mathematical and Computer Modeling, Vol. 33, 2001, pp. 1055-1083.
[49] P. P. Friedmann. "Vibration Reduction In Rotorcraft Using Actively Controlled

Flaps - From Theoretical Concept To Flight Ready Hardware." AHS 4th
363
Decennial Specialists Conference on Aeromechanics, San Francisco, CA, 21-23
January, 2004.
[50] S. Bieniawski and I. Kroo. "Flutter Suppression Using Micro-Trailing Edge

Effectors ", 44th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics,
and Materials Conference, Norfolk, Virginia, April 7-10, 2003.
[51] S. Bieniawski and I. Kroo. "Development and Testing of an Experimental

Aeroelastic Model with Micro-Trailing Edge Effectors ", 41st Aerospace Sciences
Meeting and Exhibit, Reno, Nevada, 6-9 January, 2003.
[52] J. Coder, M. Maughmer and P. Martin. "CFD Investigation of Unsteady

Rotorcraft Airfoil Aerodynamics: MiTEs and Dynamic Stall." 49th AIAA
Aerospace Sciences Meeting including the New Horizons Forum and Aerospace
Exposition, Orlando, Florida, 4-7 January, 2011.
[53] L. Liu, A. K. Padthe and P. P. Friedmann. "A Computational Study of Microflaps

with Application to Vibration Reduction in Helicopter Rotors." 50th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference, Palm Springs, California, 4-7 May, 2009.
[54] A. K. Padthe, L. Liu and P. P. Friedmann. "Numerical Evaluation of Microflaps

for On Blade Control of Noise and Vibration." 52th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, Denver, Colorado, 4-
7 April, 2011.
[55] C. Matalanis, B. Wake, D. Opoku, B.-Y. Min, N. Yeshala and L. Sankar.

"Aerodynamic Evaluation of Miniature Trailing-Edge Effectors for Active Rotor
Control " Journal of Aircraft, Vol. 48, No. 3, 2011, pp. 995-1004.
[56] P. Anusonti-Inthra, F. Gandhi and M. Frecker. "Design of a Conformable Rotor

Airfoil using Distributed Piezoelectric Actuation." ASME International
Mechanical Engineering Congress, Washington, D.C., 15-21 November 2003.
[57] S. Saito, N. Kobiki and Y. Tanabe. "Application of an Active Device for

Helicopter Noise Reduction in JAXA." Fluid Dynamics Research, Vol. 42, No. 1,
2010, pp. 1-24.
[58] J. P. Rodgers. Development of an Integral Twist-Actuated Rotor Blade for

Individual Blade Control. PhD Dissertation, Massachusetts Institute of
Technology, Boston, 1999.
[59] D. Thakkar and R. Ganguli. "Active Twist Control of Smart Helicopter Rotor - A
Survey." Journal of Aerospace Sciences and Technologies, Vol. 57, No. 4, August
2005, pp. 1-20.
364
[60] H. Monner, J. Riemenschneider, S. Opitz and M. Schulz. "Development of Active
Twist Rotors at the German Aerospace Center (DLR) ", 52nd
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials
Conference, Denver, Colorado, 4-7 April, 2011.
[61] I. Chopra. "Review of State of Art of Smart Structures and Integrated Systems "
AIAA Journal, Vol. 40, No. 11, 2002, pp. 2145-2187.
[62] S. Shin. Integral Twist Actuation of Helicopter Rotor Blades for Vibration
Reduction. PhD Dissertation, Massachusetts Institute of Technology, Boston,
MA, August, 2001.
[63] D. Kumar, B. Glaz, J. Mok, P. P. Friedmann and C. Cesnik. "Determination of

Optimum Camber Distribution in Rotating wings with Deformable Airfoils for
Vibration Reduction and Performance Enhancement using Surrogate Modeling."
Proceedings of European Rotorcraft Forum, Paris, France, 7-9 September, 2010.
[64] C. K. Maucher, B. A. Grohmann, P. Jänker, A. Altmikus, F. Jensen and H. Baier.

"Actuator Design for the Active Trailing edge of a Helicopter Rotor Blade." 33rd
European Rotorcraft Forum, Kazan, Russia, 11-13 September, 2007.
[65] C. E. S. Cesnik, S. Shin and M. L. Wilbur. "Dynamic Response of Active Twist

Rotor Blades." Smart Materials and Structures, Vol. 10, 2001, pp. 62-76.
[66] C. E. S. Cesnik and S. J. Shin. "On the Modeling of Integrally Actuated

Helicopter Blades." International Journal of Solids and Structures, Vol. 38, 2001,
pp. 1765-1789.
[67] C. E. S. Cesnik and S. J. Shin. "On the Twist Performance of a Multiple-cell

Active Helicopter Blade." Smart Materials and Structures, Vol. 10, No. 1, 2001,
pp. 53-61.
[68] S. Shin, C. E. S. Cesnik, W. K. Wilkie and M. L. Wilbur. "Design and

Manufacturing of a Model-scale Active Twist Rotor Prototype Blade." Journal of
Intelligent Material Systems and Structures, Vol. 19, No. 12, 2008, pp. 1443-
1456.
[69] S. Shin, C. E. S. Cesnik and S. R. Hall. "Closed Loop Control Test of the
NASA/Army/MIT Active Twist Rotor for Vibration Reduction." Journal of
[70] R. Palacios and C. E. S. Cesnik. "Cross-Sectional Analysis of Nonhomogeneous

Anisotropic Active Slender Structures." AIAA Journal, Vol. 43, No. 12, July,
2005, pp. 2624-2638.
[71] C. E. S. Cesnik and R. Palacios. "UM/VABS Release 1.35, User Manual." 2008.
365
[72] R. Palacios and C. E. S. Cesnik. "Geometrically Nonlinear Theory of Composite
Beams with Deformable Cross Sections." AIAA Journal, Vol. 46, No. 2, 2008, pp.
439-450.
[73] G. L. Ghiringhelli, P. Masarati and P. Mantegazza. "Characterisation of

Anisotropic, Non-Homogeneous Beam Sections with Embedded Piezo-Electric
Materials." Journal of Intelligent Material Systems and Structures, Vol. 8, No. 10,
October, 1997, pp. 842-858.
[74] G. L. Ghiringhelli, P. Masarati and P. Mantegazza. "Analysis of an Actively

Twisted Rotor by Multibody Global Modeling." Composite Structures, Vol. 52,
2001, pp. 113-122.
[75] S. Glukhikh, E. Barkanov, A. Kovalev, J. Riemenshcneider and P. Wierach.

"Design of Helicopter Rotor Blades with Actuators Made of a Piezomacrofiber
Composite." Mechanics of Composite Materials, Vol. 44, No. 1, 2008, pp. 57-64.
[76] F. Hoffmann, S. Opitz and J. Riemenschneider. "Validation of Active Twist

Modeling Based on Whirl Tower Tests." American Helicopter Society 65th
Annual Forum, Grapevine, Texas, 27-29, May, 2009.
[77] T. H. Brockmann and R. Lammering. "Beam Finite Elements for Rotating

Piezoelectric Fiber Composite Structures." Journal of Intelligent Material Systems
and Structures, Vol. 17, 2006, pp. 431-448.
[78] C. E. S. Cesnik, R. S. Park and R. Palacios. "Effective Cross section Distribution

of Anisotropic Piezocomposite Actuators for Wing twist." Proceedings of SPIE,
Vol. 5056, 2003, pp. 21-32.
[79] M. K. Sekula, M. L. Wilbur and W. T. Yeager. "Aerodynamic Design Study of an

Advanced Active Twist Rotor." American Helicopter Society 4th Decennial
Specialist Conference on Aeromechanics San Francisco, California, January 21-
23, 2004.
[80] M. K. Sekula, M. L. Wilbur and W. T. Yeager. "A Parametric Study of the

Structural design of an Advanced Active Twist Rotor." American Helicopter
Society 61st Annual Forum, Grapevine, Texas, June 1-3, 2005.
[81] M. L. Wilbur and M. K. Sekula. "The Effect of Tip Geometry on Active-Twist

Rotor Response ", American Helicopter Society 61st Annual Forum, Grapevine,
Texas, 1-3 June 2005.
[82] W. Johnson. "CAMRAD-II Comprehensive Analytical Model of Rotor

Aerodynamics and Dynamics." Johnson Aeronautics, Volume II: User's Manual,
1984.
366
[83] R. P. Thornburgh, A. R. Kreshock and M. L. Wilbur. "Structural Optimization of
Active-Twist Rotor Blades." American Helicopter Society, 67th Annual Forum,
Virginia Beach, VA, 3-5 May, 2011.
[84] J. Mok. Design Optimization for Active Twist Rotor Blades. PhD Dissertation,
University of Michigan, Ann Arbor, Michigan, 2010.
[85] C. E. S. Cesnik, J. Mok, A. Parekh and S. Shin. "Optimization Design Framework

for Integrally Twisted Helicopter Blades." 45th AIAA/ASME/ASCE/ASC
Structures, Structural Dynamics and Materials Conference, Palm Springs,
California, 19-22 April, 2004.
[86] C. E. S. Cesnik, J. Mok, J. A. Morillo and A. S. Parikh. "Design Optimization of

Active Twist Rotor Blades." 30th European Rotorcraft Forum, Marseilles,
France, 2005.
[87] P. Masarati, M. Morandini, J. Riemenschneider, P. Wierach, S. Gluhih and E.

Barkanov. "Optimal design of an Active Twist 1:2.5 Scale Rotor Blade."
European Rotorcraft Forum, Florence, Italy, September 13-15, 2005.
[88] A. Kovalos, E. Barakanov and S. Gluhihs. "Active Twist Model Rotor Blades
with D-Spar Design." Transport, Vol. XXII, No. 1, 2007, pp. 38-44.
[89] E. Barakanov, S. Gluhih and A. Kovalev. "Optimal Design of the Active Twist
for Helicopter Rotor Blades with C-Spar." Mechanics of Advanced Materials and
Structures, Vol. 15, No. 3-4, 2008, pp. 325-334.
[90] B. Glaz, P. P. Friedmann and L. Liu. "Vibration Reduction and Performance

Enhancement of Helicopter Rotors Using an Active/Passive Approach." 49th
Conference, Schaumburg, IL, 7-10 April, 2008.
[91] J. Zhang, E. C. Smith and K. W. Wang. "Active-Passive Hybrid Optimization of

Rotor Blades with Trailing Edge Flaps." Journal of the American Helicopter
Society, Vol. 49, No. 1, 2004, pp. 54-65.
[92] P. P. Friedmann and D. Hodges. "Rotary Wing Aeroelasticity -A Historical

Perspective and Some Current Trends." 44th AIAA/ASME/ASCE/AHS Structures,
Structural Dynamics, and Materials Conference, Norfolk, Virginia, 7-10 April,
2003.
[93] A. Datta and W. Johnson. "An Assessment of the State-of-the-art in

Multidisciplinary Aeromechanical Analyses." AHS Specialist's Conference on
Aeromechanics, San Francisco, CA, 23-25 January, 2008.
367
[94] W. Johnson and A. Datta. "Requirements for Next Generation Comprehensive
Analysis of Rotorcraft." AHS Specialist's Conference on Aeromechanics, San
Francisco, CA, 23-25 January, 2008.
[95] M. E. Gunduz. Software Integration for Automated Stability Analysis and Design
Optimization of a Rotor Blade. PhD Dissertation, Georgia Institute of
Technology, Atlanta, GA, 2010.
[96] B. Glaz. Active/Passive Optimization of Helicopter Rotor Blades for Improved

Vibration, Noise, and Performance Characteristics. PhD Dissertation, University
of Michigan, Ann Arbor, 2008.
[97] B. Glaz, P. P. Friedmann, L. Liu, D. Kumar and C. E. S. Cesnik. "The AVINOR

Aeroelastic Simulation Code and its Application to Reduced Vibration Composite
Rotor Blade Design." 50th AIAA/ASME/ASCE/AHS/ASC Structures. Structural
Dynamics, and Materials Conference, Palm Springs, California, 4-7 May, 2009.
[98] J. Ku, V. Volovoi and D. Hodges. "Multilevel-Multiphase Optimization of

Composite Rotor Blade with Surrogate Model." 48th
Conference, Honolulu, Hawaii, 23-26 April 2007.
[99] L. Li, V. Volovoi and D. Hodges. "Cross-Sectional Design of Composite Rotor

Blades." Journal of the American Helicopter Society, Vol. 53, No. 1, 2008, pp.
240-251.
[100] A. S. Khalid. Development and Implementation of Rotorcraft Preliminary Design

Method using Multidisciplinary Design Optimization. PhD Dissertation, Georgia
Institute of Technology, Atlanta, Georgia, December, 2006.
[101] K. B. Collins. A Multi-fidelity Framework for Physics based Rotor Blade

Simulation and Optimization. PhD Dissertation, Georgia Institute of Technology,
Atlanta, Georgia, 2008.
[102] K. K. Saijal, R. Ganguli and S. R. Viswamurthy. "Optimization of Helicopter

Rotor Using Polynomial and Neural Network Metamodels." Journal of Aircraft,
Vol. 48, No. 2, 2011, pp. 553-566.
[103] S. R. Viswamurthy and R. Ganguli. "Optimal Placement of Piezoelectric Actuated

Trailing-Edge Flaps for Helicopter Vibration Control." Journal of Aircraft, Vol.
46, No. 1, January-February, 2009, pp. 244-253.
[104] D. Patt, L. Liu and P. P. Friedmann. "Rotorcraft Vibration Reduction and Noise
Prediction Using a Unified Aeroelastic Response Simulation." Journal of the
368
[105] D. Patt, L. Liu and P. P. Friedmann. "Simultaneous Vibration and Noise
Reduction in Rotorcraft Using Aeroelastic Simulation." Journal of the American
[106] L. Liu, P. P. Friedmann, I. Kim and D. S. Bernstein. "Rotor Performance

Enhancement and Vibration Reduction in Presence of Dynamic Stall Using
Actively Controlled Flaps." Journal of the American Helicopter Society, Vol. 53,
No. 4, 2008, pp. 338-350.
[107] J.-S. Kim. Design and Analysis of Rotor Systems with Multiple Trailing edge
Flaps and Resonant Actuators. PhD Dissertation, Pennsylvania State University,
Pennsylvania, 2005.
[108] S. R. Viswamurthy and R. Ganguli. "Using the Complete Authority of Multiple

Active Trailing-edge Flaps for Helicopter Vibration Control." Journal of
Vibration and Control, Vol. 14, No. 8, 2008, pp. 1175-1199.
[109] U. Dalli and Ş. Yüksel. "Identification of Flap motion parameters for Vibration
reduction in Helicopter rotors with Multiple active Trailing edge flaps." Shock
and Vibration, Vol. 18, No. 5, 2010, pp. 727-745.
[110] K. Noburu, S. Saito, T. Fukami and T. Komura. "Design and Performance

Evaluation of Full Scale On-board Active Flap System." American Helicopter
Society 63rd Annual Forum, Virginia Beach, Virginia, May 1-3, 2007.
[111] M. V. Fulton. "Aeromechanics of the Active Elevon Rotor." American Helicopter

Society 61st Annual Forum, Grapevine, Texas, June 1-3, 2005.
[112] E. F. Prechtl. Design and Implementation of a Piezoelectric Servo-Flap Actuation

System for Helicopter Rotor Individual Blade Control. PhD Dissertation,
Massachusetts Institute of Technology, Boston, 2000.
[113] B. K. S. Woods, C. S. Kothera, J. Sirohi and N. M. Wereley. "Pneumatic

Artificial Muscles for Trailing Edge Flap Actuation: A Feasibility Study." Smart
Materials and Structures, Vol. 20, No. 10, 2011, pp. 1-15.
[114] F. Straub, D. K. Kennedy, D. B. Domzalski, A. A. Hassan, H. Ngo and T. S.

Birchette. "Smart Material-actuated Rotor Technology – SMART." Journal of
Intelligent Material Systems and Structures, Vol. 15, April, 2004, pp. 249-260.
[115] F. Straub, R. A. Vaidyanathan, T. S. Birchette and B. H. Lau. "SMART Rotor

Development and Wind Tunnel Test." 35th European Rotorcraft Forum,
Hamburg, Germany, 22-25 September, 2009.
[116] P. Lorber, J. O'Neill, B. Hein, B. Isabella, J. Andrews, M. Brigley, J. Wong, P.

LeMasurier and B. Wake. "Whirl and Wind Tunnel Testing of the Sikorsky
369
Active Flap Demonstration Rotor." American Helicopter Society 67th Annual
Forum, Virginia Beach, VA, May 3-5, 2011.
[117] H. Mainz, B. G. van der Wall, P. Leconte, F. Ternoy and H.-M. d. Rochettes.
"ABC Rotor Blades: Design, Manufacturing and Testing." 31st European
Rotorcraft Forum, Florence, Italy, September 13-15, 2005.
[118] Y. Delrieux, A. L. Pape, P. Leconte, P. Crozier, B. Gimonet and H.-M. d.

Rochettes. "Wind-Tunnel Assessment of the Concept of Active Flaps on a
Helicopter Rotor Model." American Helicopter Society 63rd Annual Forum,
Virginia Beach, VA, May 1-3, 2007.
[119] E. F. Prechtl and S. R. Hall. "Design of a High Efficiency, Large Stroke,

Electromechanical Actuator." Smart Materials and Structures, Vol. 8, 1999, pp.
13-30.
[120] N. A. Koratkar and I. Chopra. "Analysis and Testing of Mach-Scaled Rotor with
Trailing-Edge flaps." AIAA Journal, Vol. 38, No. 7, 2000, pp. 1113-1124.
[121] B. Roget and I. Chopra. "Wind Tunnel testing of Rotor with Individually
Controlled Trailing-Edge Flaps for Vibration reduction." Journal of Aircraft, Vol.
45, No. 3, May-June, 2008, pp. 868-879.
[122] J. H. Milgram, I. Chopra and F. Straub. "Rotors with Trailing Edge Flaps:
Analysis and Comparison with Experimental Data." Journal of the American
[123] S. Dawson, F. Straub, E. Booth and M. Marcolini. "Wind Tunnel Test of an

Active Flap Rotor : BVI Noise and Vibration Reduction." 51st Annual Forum of
the American Helicopter Society, Fort Worth, Texas, May, 1995.
[124] M. V. Fulton and R. Ormiston. "Small-Scale Rotor Experiments with On-blade

elevons to Reduce Blade Vibratory Loads in Forward Flight." American
Helicopter Society 54th Annual Forum, Washington, DC, May 20-22, 1998.
[125] L. R. Centolanza and E. C. Smith. "Induced-Shear Piezoelectric Actuators for

Active Twist Rotor Blades." 43rd AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, Denver, Colorado, 22-25 April,
2002.
[126] J. W. Clement, D. Brei, A. J. Moskalik and R. Barrett. "Bench-top

Characterization of an Active rotor blade Flap system incorporating C-block
Actuators." AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and
Materials Conference and Exhibit, 39th, and AIAA/ASME/AHS Adaptive
Structures Forum, Long Beach, CA, April 20-23, 1998.
370
[127] S. R. Hall and E. F. Prechtl. "Development of a Piezoelectric Servoflap for
Helicopter Rotor Control." Smart Materials and Structures, Vol. 5, 1996, pp. 26-
34.
[128] F. Straub, H. Ngo, V. R. Anand and D. B. Domzalski. "Development of a

Piezoelectric Actuator for Trailing edge Flap Control of full scale Rotor Blades."
Smart Materials and Structures, Vol. 10, 2001, pp. 25-34.
[129] T. Lee and I. Chopra. "Design of Piezostack-driven Trailing-edge Flap Actuator

for Helicopter Rotors." Smart Materials and Structures, Vol. 10, 2001, pp. 15-24.
[130] H. Ghiasi, D. Pasini and L. Lessard. "Optimum Stacking Sequence Design of

Composite Materials Part I: Constant Stiffness Design." Composite Structures,
Vol. 90, No. 1, 2009, pp. 1-11.
[131] H. Ghiasi, K. Fayazbakhsh, D. Pasini and L. Lessard. "Optimum Stacking

Sequence Design of Composite Materials Part II: Variable Stiffness Design."
Composite Structures, Vol. 93, No. 1, 2010, pp. 1-13.
[132] Z. Gürdal, S. IJsselmuiden and J. v. Campen. "Composite Laminate Optimization

with Discrete Variables." Encyclopedia of Aerospace Engineering: John Wiley &
Sons, Ltd, 2010.
[133] G. J. Kennedy and J. J. R. A. Martins. "A Laminate Parametrization Technique

for Discrete Ply Angle Problems with Manufacturing Constraints." Structural and
Multidisciplinary Optimization, 2012.
[134] G. J. Kennedy and J. J. R. A. Martins. "A Regularized Discrete Laminate

Parametrization Technique with Applications to Wing-Box Design Optimization."
53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and
Materials Conference, Honolulu, Hawaii, 23 - 26 April, 2012.
[135] S. Murugan, R. Ganguli and D. Harursampath. "Surrogate based Design

Optimization of Composite Airfoil Cross section for Helicopter Vibration
Reduction." The Aeronautical Journal, Vol. 116, No. 1181, 2012, pp. 709-725.
[136] S. D. Guido, H. J. M. Geijselaers and A. d. Boer. "Continuous-Discrete Variable

Optimization on Composite using Kriging Surrogate Model." 16th International
Conference on Composite Structures, Porto, Portugal, June 28-30, 2011.
[137] J. N. Kudwa. "Overview of the DARPA Smart Wing Project." Journal of

Intelligent Material Systems and Structures, Vol. 15, 2004, pp. 261-267.
[138] S. Kota, J. Hetrick, O. Russel, D. Paul, E. Pendleton, P. Flick and C. Tilmann.

"Design and application of compliant mechanisms for morphing aircraft
structures." Proceedings of SPIE, Vol. 5054, No. 24, 2003, pp. 24-33.
371
[139] M. J. Santer and S. Pellegrino. "Topology Optimization of Adaptive Compliant
Aircraft Wing Leading Edge." 48th AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics & Materials (SDM) Conference Honolulu, Hawaii, 23-26
April, 2007.
[140] L. Saggere and S. Kota. "Static Shape Control of Smart Structures Using
Compliant Mechanisms " AIAA Journal, Vol. 37, No. 5, May 1999, pp. 572-578.
[141] F. Gandhi and P. Anusonti-Inthra. "Skin Design studies for Variable Camber
Morphing Airfoils." Smart Materials and Structures, Vol. 17, 2008, pp. 1-8.
[142] O. K. Rendinotis, L. N. Wilson., D. C. Lagoudas and M. M. Khan. "Development

of a Shape-Memory-Alloy Actuated Biomimetic Hydrofoil." Journal of
Intelligent Material Systems and Structures, Vol. 13, 2002, pp. 35-49.
[143] J. D. Bartley-Cho, D. P. Wang, C. A. Martin, J. N. Kudwa and M. N. West.

"Development of High-rate, Adaptive Trailing Edge Control Surface for the
Smart Wing Phase 2 Wind Tunnel Model " Journal of Intelligent Material
Systems and Structures, Vol. 15, 2004, pp. 279-291.
[144] F. Gandhi, M. Frecker and A. Nissly. "Design Optimization of a Controllable

Camber Rotor Airfoil." AIAA Journal, Vol. 46, No. 1, 2008, pp. 142-153.
[145] S. Thepvongs, C. E. S. Cesnik, R. Palacios and D. A. Peters. "Finite-State

Aeroelastic Modeling of Rotating Wings with Deformable Airfoils." American
Helicopter Society 64th Annual Forum, Montreal, Canada, April 29-May 1, 2008.
[146] S. Thepvongs, J. R. Cook, C. E. S. Cesnik and M. J. Smith. "Computational

Aeroelasticity of Rotating Wings with Deformable Airfoils." American Helicopter
Society 65th Annual Forum, Grapevine, Texas, May, 2009.
[147] P. J. Röhl, D. Kumar, P. Dorman, M. Sutton and C. Cesnik. "A Composite Rotor
Blade Structural Design Environment for Aeromechanical Assessments in
Conceptual and Preliminary Design." 68th American Helicopter Society
International Annual Forum, Fort Worth, Texas, 1-3 May, 2012.
[148] P. J. Röhl, P. Dorman, M. Sutton, D. Kumar and C. Cesnik. "A Multidisciplinary

Design Environment for Composite Rotor Blades." 53rd
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Material
Conference, Honolulu, Hawaii, 23 - 26 April 2012.
[149] D. Kumar, C. Cesnik, P. J. Röhl and M. Sutton. "Optimization Framework for the
Dynamic Analysis and Design of Active Twist Rotors." 68th American Helicopter
Society International Annual Forum, Fort Worth, TX, 1-3 May, 2012.
372
[150] D. Kumar and C. Cesnik. "New Hybrid Optimization for Design of Active Twist
Rotor." 54th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and
Materials Conference, Boston, MA 8-11 April, 2013.
[151] D. Kumar and C. Cesnik. "New Strategy for Design of Composite Rotor Blade
with Active Flaps." American Helicopter Society 69th Annual Forum and
Technology Display, Phoenix, Arizona May 21-23, 2013.
[152] D. Kumar and C. Cesnik. "Performance Enhancement in Dynamic Stall Condition

using Active Camber Deformation." American Helicopter Society 69th Annual
Forum and Technology Display, Phoenix, Arizona, May 21-23, 2013.
[153] P. J. Röhl, P. Dorman, M. Sutton, C. E. S. Cesnik and D. Kumar. "IXGEN- A

Modeling Tool for the Preliminary Design of Composite Rotor Blade." AHS
Future Vertical Lift Design Conference San Francisco, 18-20 January 2012.
[154] H. Saberi, M. Khoshlahjeh, R. Ormiston and M. Rutkowski. "Overview of RCAS

and Application to Advanced Rotorcraft Problems." AHS 4th Decennial Specialist
Conference on Aeromechanics, San Francisco, California, January 21-23, 2004.
[155] R. Palacios. Asymptotic Models of Integrally-Strained Slender Structures for

High-Fidelity Nonlinear Aeroelastic Analysis. PhD Dissertation, University of
Michigan, Ann Arbor, 2005.
[156] "RCAS User Manual." 2011.
[157] W. G. Bousman. "Aerodynamic Characteristics of SC1095 and SC1094 R8

Airfoils." NASA, TP-2003-212265, December, 2003.
[158] K. Deb, A. Pratap, S. Agarwal and T. Meriyan. "A Fast and Elitist Multiobjective
Genetic Algorithm: NSGA-II." IEEE Transactions on Evolutionary Computation,
Vol. 6, No. 2, 2002, pp. 182-197.
[159] D. R. Jones, M. Schonlau and W. J. Welch. "Efficient Global Optimization of

Expensive Black Box Functions." Journal of Global Optimization, Vol. 13, 1998,
pp. 455-492.
[160] A. Forester, A. Sobester and A. Keane. Engineering design via Surrogate

Modeling: A Practical Guide. AIAA: John Wiley and Sons, Ltd, 2008.
[161] N. V. Queipo, R. T. Haftka, W. Shy, T. Goel, R. Vaidyanathan and P. K. Tucker.

"Surrogate based Analysis and Optimization." Progress in Aerospace Sciences,
Vol. 41, 2005, pp. 1-28.
[162] J. Sacks, W. J. Welch, T. J. Mitchell and H. P. Wynn. "Design and Analysis of

Computer Experiments." Statistical Science, Vol. 4, No. 4, 1989, pp. 409-435.
373
[163] S. N. Lophaven, H. B. Nielsen and J. Sondergaard. "A Matlab Kriging Toolbox,
version 2.0." Informatics and Mathematical Modeling, Technical Report IMM-
TR-2002-12, 2002.
[164] B. Glaz, P. P. Friedmann and L. Liu. "Helicopter Vibration Reduction throughout

the Entire Flight Envelope Using Surrogate-Based Optimization." Journal of
American Helicopter Society, Vol. 54, No. 1, 2009, pp. 12007-1-12007-15.
[165] C. E. S. Cesnik, S. Shin, W. K. Wilkie, M. L. Wilbur and P. H. Mirick.

"Modeling, Design and Testing of the NASA/ARMY/MIT Active Twist Rotor
Prototype Blade." American Helicopter Society 55th Annual Forum Montreal,
Canada, May 25-27, 1999.
[166] K. Deep, K. P. Singh, M. L. Kansal and C. Mohan. "A Real coded Genetic
Algorithm for solving Integer and Mixed-integer Optimization Problems."
Applied Mathematics and Computation, Vol. 212, No. 2, 2009, pp. 505-518.
[167] D. A. Peters, M. A. Hsieh and A. Torrero. "A State Space Airloads Theory for
Flexible Airfoils." American Helicopter Society 63rd Annual Forum, Phoenix,
AZ, May, 2006.
[168] D. A. Peters and C. J. He. "Finite State Induced Flow Models Part II: Three-
Dimensional Rotor Disk." Journal of Aircraft, Vol. 32, No. 2, March-April, 1995,
pp. 323-333.
[169] L. A. Ahaus. An Airloads Theory for Morphing Airfoils in Dynamic Stall with
Experimental Correlation. PhD Dissertation, Washington University in St. Louis,
Saint Louis, Missouri, May, 2010.
[170] J. Martin and T. Simpson. "Use of Kriging Models to Approximate Deterministic

Computer Models." AIAA Journal, Vol. 43, No. 4, 2005, pp. 853-863.
[171] R. Palacios and C. E. S. Cesnik. "Low-Speed Aeroelastic Modeling of Very

Flexible Slender Wings with Deformable Airfoils." 49th
Conference, Schaumburg, IL, 7-10 April 2008.
[172] A. Forester, A. Sobester and A. Keane. Engineering Design via Surrogate

Modeling : A Practical Guide. John Wiley and Sons, Ltd, 2008.
[173] M. J. Sasena. Flexibility and Efficiency Enhancements for Constrained Global

Design Optimization with Kriging Approximations. PhD Dissertation, University
of Michigan, Ann Arbor, 2002.
374
[174] S. R. Hall, T. Tzianetopoulou, F. Straub and H. Ngo. "Design and Testing of a
double X-frame Piezoelectric Actuator." Proceedings of SPIE, Vol. 3985, 2000.
[175] J.-L. Petitniot, H.-M. d. Rochettes and P. Leconte. "Experimental Assessment and
Further Development of Amplified Piezo Actuators for Active Flap Devices."
Actuator 2002, 8th International Conference on New Actuators, Bremen,
Germany, 10-12 June, 2002.
[176] E. R. Muir, L. Liu, P. P. Friedmann and D. Kumar. "Effect of Piezoelectric

Actuator Hysteresis on Helicopter Vibration and Noise Reduction." Journal of
Guidance, Control and Dynamics, Vol. 35, No. 4, 2012, pp. 1299-1311.
[177] J. Cook and M. J. Smith. "Effect of Testing Facility" Personal Communication,

Georgia Institute of Technology, March, 2010
375

Design and Analysis of Composite Rotor Blades For ActivePassive Vibration Reduction

Uploaded by

Copyright:

Available Formats

Design and Analysis of Composite Rotor Blades For ActivePassive Vibration Reduction

Uploaded by

Document Information

Original Description:

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

Design and Analysis of Composite Rotor Blades For ActivePassive Vibration Reduction

Uploaded by

Copyright:

Available Formats

Design and Analysis of Composite Rotor Blades for

Active/Passive Vibration Reduction

A dissertation submitted in partial fulfillment

Professor Carlos E. S. Cesnik, Chair

I would like to extend my sincere thanks to my family, teachers, friends and

colleagues who helped me immensely during my days in the graduate school.

course of my long and uneventful (fortunately!) doctoral research. I appreciate the

graduate school to help me learn better, to grow as a researcher and to build my

acknowledge encouragement and guidance provided by Prof. Friedmann during my

implementation of the basic surrogate-based optimization technique. I also benefitted

tensile testing of fabricated blade sections.

Ananthraman, Dipankar Dua, Dr. Amit Kaushik and my lunch group.

Tech/University of Michigan/Washington University Vertical Lift Rotorcraft Center of

and Engineering Command (AMRDEC), Aviation Applied Technology Directorate

(AATD) through SBIR Phase II contract W911W6-10-C-0030.

ACKNOWLEDGEMENTS ............................................................................................ iii

LIST OF FIGURES .......................................................................................................... x

LIST OF TABLES ....................................................................................................... xviii

LIST OF APPENDICES .............................................................................................. xxii

ABSTRACT .................................................................................................................. xxiv

1.1 Introduction to Vibration in Helicopters ................................................................ 1

1.2 Methods for Vibration Reduction ........................................................................... 5

1.3 Literature Review Relevant to the Thesis ............................................................. 14

1.4 Objectives of the Dissertation ................................................................................ 27

CHAPTER 2. A MULTIDISCIPLINARY DESIGN ENVIRONMENT FOR

2.1 Introduction ............................................................................................................. 34

2.2 Design Approach ..................................................................................................... 37

2.3 Application Example: Vibratory Hub Load Minimization ................................ 44

2.4 Optimization Studies for Vibration Reduction .................................................... 52

2.5 Concluding Remarks .............................................................................................. 64

CHAPTER 3. OPTIMIZATION FRAMEWORK FOR THE DYNAMIC

3.1 Optimization Framework ....................................................................................... 67

3.2 Numerical Studies ................................................................................................... 71

3.3 Optimization Results .............................................................................................. 75

3.5 Optimization at 5/rev Actuation Frequency (Max θ5/rev) .................................... 95

3.6 Optimization at a Range of Actuation Frequencies (Max θ345/rev) ...................... 97

3.7 Effect of Advance Ratio ........................................................................................ 101

3.8 Concluding Remarks ............................................................................................ 103

CHAPTER 4. MIXED-VARIABLE OPTIMIZATION FOR DESIGN OF

4.1 Architecture for New Mixed-variable Optimization Framework .................... 107

4.2 Optimization with Normalized Ply Thickness .................................................... 114

4.3 Optimization Study with 8 Design Variables ..................................................... 130

4.5 Post Processing of Optimization Results............................................................. 142

4.6 Concluding Remarks ............................................................................................ 148

CHAPTER 5. NEW STRATEGY FOR DESIGN OF COMPOSITE ROTOR

5.1 Aeroelastic Analysis with Active Flaps ............................................................... 152

5.2 Baseline Rotor Blade............................................................................................. 153

5.3 Preliminary Analysis ............................................................................................ 156

5.5 Optimization Results ............................................................................................ 163

5.6 Concluding Remarks ............................................................................................ 171

CHAPTER 6. PERFORMANCE ENHANCEMENT AND VIBRATION

6.1 Optimization Framework ..................................................................................... 175

6.2 Camber Deformation Shape ................................................................................ 182

6.3 Preliminary Numerical Results ........................................................................... 184

6.4 Optimization Problem Definition ........................................................................ 190

6.5 Concluding Remarks ............................................................................................ 202

CHAPTER 7. CONCLUSIONS AND FUTURE WORK ...................................... 204

7.1 Summary ................................................................................................................ 204