Post on 24-Mar-2015
High-Lift Airfoil Separation Control with Dielectric Barrier Discharge Plasma Actuators
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University
By
Jesse C. Little
Graduate Program in Mechanical Engineering
The Ohio State University
2010
Dissertation Committee:
Igor Adamovich
Michael Dunn
James Gregory
Mo Samimy, Advisor
Andrea Serrani
Copyright by
Jesse C. Little
2010
ii
Abstract
This work examines the performance of dielectric barrier discharge (DBD)
plasma actuators for controlling separation from the leading edge and trailing edge flap
shoulder of a supercritical high-lift airfoil. DBD plasma actuators driven by both typical
AC voltages (AC-DBD) and more developmental nanosecond duration pulses (NS-DBD)
are investigated. Characterization of the two actuators shows that very different behavior
is created when exciting the plasma discharge using these two waveforms. The AC-DBD
plasma actuator functions through electrohydrodynamic effects that introduce zero net
mass, but nonzero net momentum into the flow. Conversely, the electrohydrodynamic
effects of the NS-DBD are quite weak suggesting thermal effects from rapid localized
heating by the plasma are responsible for control authority. The performance of both
devices as separation control actuators is tested on a high-lift airfoil system.
The AC-DBD is effective for controlling turbulent boundary layer separation
from a deflected trailing edge flap between Reynolds numbers of 240,000 and 750,000.
Momentum coefficients for the AC-DBD plasma actuator are generally an order of
magnitude lower than those usually employed for such studies yet control authority is
still realized through amplification of natural vortex shedding from the flap shoulder. The
corresponding lift enhancement is primarily due to upstream effects from increased
circulation around the entire model rather than full separated flow reattachment to the
iii
deflected flap surface. Lift enhancement via instability amplification is found to be
relatively insensitive to changes in angle of attack provided that the separation location
and underlying dynamics do not change. Control authority decreases with increasing
Reynolds number and flap deflection highlighting the necessity for further optimization
of AC-DBD plasma actuators for use in realistic takeoff and landing transport aircraft
applications. As a whole, these findings compare favorably to studies on a similar high-
lift platform using piezoelectric driven zero net mass flux actuation.
The NS-DBD plasma actuator is ineffective for controlling separation from the
deflected trailing edge flap. However, the device is found to be superior to the tested AC-
DBD plasma actuators for controlling leading edge separation and rivals the performance
of a passive droop by extending the stall angle by six degrees in the Reynolds number
range 750,000-1,000,000. Detailed flow diagnostics show the NS-DBD plasma actuator
functions as an active trip for pre-stall incidence angles and generates coherent spanwise
vortices that entrain freestream momentum into the separated region at post-stall angles.
These structures are generated across all surveyed frequencies, but optimal dimensionless
frequencies for controlling separation are in the range four to six depending on the
incidence angle. The contrasting performance of the NS-DBD plasma actuator at the
leading and trailing edge in comparison to the AC-DBD is discussed and
recommendations for future work are provided.
iv
Acknowledgments
The world class facilities, technical guidance, interesting projects and professional
development opportunities provided to me by Prof. Mo Samimy have been invaluable. I
will be forever grateful for his tutelage. I wish to acknowledge Prof. Igor Adamovich for
being the driving force behind much of the NS-DBD plasma work. Thanks to my
dissertation committee for providing comments that have improved the quality of this
manuscript and challenged me to explore additional related avenues of study. I have had
the opportunity to work with incredible people during my time at the OSU GDTL. A list
of these names would require its own appendix. The stimulating discussions and unique
ideas proposed by these individuals have substantially influenced my growth as a
researcher. I will always be thankful for these interactions and the friendships that have
subsequently developed. I wish to specifically acknowledge Dr. Munetake Nishihara and
Dr. Keisuke Takashima for acquiring the electrical measurements and schlieren images
used in the actuator characterization as well as their work on the NS-DBD power supply
development. Thanks to Kristine McElligott for assisting in the wind tunnel experiments.
A very special recognition is reserved for my family and especially my wife, Misty.
Thank you for everything. Aspects of this work have been supported by AFRL, DAGSI,
The Boeing Company and the Howard D. Winbigler Professorship at OSU.
v
Vita
October 24, 1980 ………………………………………………. Born Athens, OH, USA 2002-2004……………………………………………Undergraduate Research Associate
Department of Mechanical Engineering The Ohio State University
2003-2004…………………………………………………………………Research Intern Aerosol and Process Technologies Battelle Memorial Institute Columbus, OH, USA 2004…………………………………………………………B.S. Mechanical Engineering The Ohio State University 2005…………………………………………………………M.S. Mechanical Engineering The Ohio State University 2009…………………………………………………………Graduate Teaching Associate Department of Mechanical Engineering The Ohio State University 2004-present…………………………………………………Graduate Research Associate Department of Mechanical Engineering The Ohio State University
Publications
Book Chapters 1. Samimy, M., Debiasi, M., Caraballo, E., Serrani, A., Yuan, X., Little, J., and Myatt,
J., “Reduced-order Model-based Feedback Control of Subsonic Cavity Flows,” In Active Flow Control, R. King, editor, Volume 95 of Notes on Numerical Fluid Mechanics and Multidisciplinary Design (NNFM), pages 211-229. Springer Verlag, 2007. ISBN: 978-3-540-71438-5.
vi
Refereed Journal Articles 1. Little, J., and Samimy, M., “Control of Separation from the Flap of a High-Lift
Airfoil using DBD Plasma Actuation,” submitted to AIAA Journal, January 2010. 2. Little, J., Nishihara, M., Adamovich, I. and Samimy, M., “High-Lift Airfoil Trailing
Edge Separation Control Using a Single Dielectric Barrier Discharge Plasma Actuator,” Experiments in Fluids, DOI 10.1007/s00348-009-0755-x, 2009.
3. Yuan, X., Caraballo, E., Little, J., Debiasi, M., Serrani, A., Ozbay, H., Myatt, J. and Samimy, M., “Feedback Control Design for Subsonic Cavity Flows,” Applied and Computational Mathematics, Vol. 8, No. 1, 2009, pp. 70-91.
4. Malone, J., Debiasi, M., Little, J. and Samimy, M., “Analysis of the Spectral Relationships of Cavity Tones in Subsonic Resonant Cavity Flows,” Physics of Fluids, Vol. 21, No. 055103, 2009.
5. Caraballo, E., Little, J., Debiasi, M., and Samimy, M., “Development and Implementation of an Experimental Based Reduced-order Model for Feedback Control of Subsonic Cavity Flows,” Journal of Fluids Engineering, Vol. 129, No. 7, pp. 813-824, 2007.
6. Little, J., Debiasi, M., Caraballo, E., and Samimy, M., “Effects of Open-loop and Closed-loop Control on Subsonic Cavity Flows,” Physics of Fluids, Vol. 19, No. 6, 065104, 2007.
7. Samimy, M., Debiasi, M., Caraballo, E., Serrani, A., Yuan, X., Little, J., and Myatt, J. H., “Feedback Control of Subsonic Cavity Flows Using Reduced-order Models,” Journal of Fluid Mechanics, Vol. 579, pp. 315-346, 2007.
8. Yan. P., Debiasi, M., Yuan, X., Little, J., Özbay, H., and Samimy, M., “Experimental Study of Linear Closed-Loop Control of Subsonic Cavity Flow,” AIAA Journal, Vol. 44, No. 5, pp. 929-938, 2006.
Fields of Study
Major Field: Mechanical Engineering
Studies in: Aerodynamics, Experimental Techniques, Flow Control, Fluid Mechanics, Optical Diagnostics, Turbulence
vii
Table of Contents
Abstract ............................................................................................................................... ii
Acknowledgments.............................................................................................................. iv
Vita ...................................................................................................................................... v
Publications ......................................................................................................................... v
Fields of Study ................................................................................................................... vi
Table of Contents .............................................................................................................. vii
List of Tables ...................................................................................................................... x
List of Figures .................................................................................................................... xi
Nomenclature ................................................................................................................... xxi
Chapter 1: Introduction ....................................................................................................... 1
Chapter 2: Background ....................................................................................................... 5
2.1 Separation Control..................................................................................................... 5
2.1.1 LE Separation Control ........................................................................................ 6
2.1.2 TE Separation Control ........................................................................................ 8
2.2 Plasma Actuators ..................................................................................................... 10
viii
2.2.1 AC-DBD Plasma Actuators .............................................................................. 10
2.2.2 NS-DBD Plasma Actuators .............................................................................. 17
2.3 Separation Control with Dielectric Barrier Discharge Plasma Actuators ............... 20
Chapter 3: Experimental Facilities and Measurement Techniques ................................... 24
3.1 Wind Tunnel ............................................................................................................ 24
3.2 Airfoil Model........................................................................................................... 36
3.3 Plasma Actuator Hardware...................................................................................... 40
3.4 Diagnostics .............................................................................................................. 43
3.4.1 Static Pressure................................................................................................... 43
3.4.2 Fluctuating pressure .......................................................................................... 43
3.4.3 PIV .................................................................................................................... 49
3.4.4 Accuracy ........................................................................................................... 57
3.5 Test Conditions ....................................................................................................... 59
Chapter 4: Dielectric Barrier Discharge Plasma Actuators .............................................. 60
4.1 AC DBD Plasma Actuator Design .......................................................................... 60
4.2 AC-DBD Plasma Actuator Characterization........................................................... 66
4.3 NS DBD Plasma Actuator Design .......................................................................... 87
4.4 NS-DBD Plasma Actuator Characterization ........................................................... 87
Chapter 5: Baseline Verification and Characterization .................................................. 103
ix
Chapter 6: Trailing Edge Separation Control ................................................................. 119
6.1 Single AC-DBD Plasma Actuator Control Results ............................................... 119
6.2 Perspective on OSU Single AC-DBD Control Results ......................................... 147
6.3 Attempts to Extend AC-DBD Plasma Control Authority ..................................... 148
6.4 Single NS-DBD Separation Control Results ......................................................... 153
Chapter 7: Leading Edge Separation Control ................................................................. 156
7.1 Single DBD Plasma Actuator Control Results: Downstream Orientation ............ 157
7.2 Single DBD Plasma Actuator Control Results: Upstream Orientation ................. 174
7.3 Comparison of Single NS-DBD Plasma Actuators to Passive Flow Control ....... 196
Chapter 8: Summary and Conclusions ............................................................................ 199
Chapter 9: Discussion of Future Work ........................................................................... 205
References ....................................................................................................................... 208
x
List of Tables
Table 3.1: PIV Uncertainty ............................................................................................... 58
Table 4.1: Properties of various dielectric materials (Roth and Dai 2006). ..................... 64
Table 4.2: Momentum and power characteristics for 3 kHz actuation at 20 kVpp............ 80
Table 4.3: Momentum and power characteristics for 3 kHz actuation at 20 kVpp with
various burst frequencies at dc=50%. ............................................................................... 81
Table 4.4: Electrical properties of NS-DBD plasma on a 46 cm long flat plate actuator. 90
Table 4.5: Electrical properties of AC-DBD plasma on a 46 cm long flat plate actuator. 91
Table 5.1: Boundary layer properties at x/c=0.70 measured using PIV. ........................ 118
xi
List of Figures
Figure 1.1: Simplified high-lift airfoil system with deflected LE droop and TE simple
flap. ..................................................................................................................................... 2
Figure 1.2: Typical high-lift airfoil system with deflected multi-element LE slat and TE
flap (Lin and Dominik 1997). ............................................................................................. 2
Figure 1.3: Further simplified high-lift airfoil system with deflected TE flap. .................. 3
Figure 2.1: Typical asymmetric DBD plasma actuator geometry (Corke et al. 2010). .... 11
Figure 2.2: Simultaneous traces of voltage and current for a typical AC-DBD plasma
actuator. ............................................................................................................................. 12
Figure 2.3: High-speed photography of the forward stroke (a) and reverse/back stroke (b)
of a typical AC-DBD plasma actuator (Enloe et al. 2008a). ............................................ 14
Figure 2.4: Schematic of negatively charged species movement for the forward (a) and
reverse/back stroke (b) in a typical AC-DBD plasma actuator (Enloe et al. 2008a). ....... 14
Figure 2.5: Visual appearance of a typical ~10 cm long AC-DBD plasma actuator
operating in the normal glow regime (a) and at maximum thrust (b) (Thomas et al. 2009)
........................................................................................................................................... 15
Figure 3.1: Axial fan used in the subsonic recirculating wind tunnel. ............................. 25
Figure 3.2: Operator keypad for the subsonic recirculating wind tunnel. ........................ 26
Figure 3.3: Subsonic recirculating wind tunnel contraction section. ................................ 26
xii
Figure 3.4: Flow conditioning screens at the subsonic recirculating wind tunnel
contraction entrance. ......................................................................................................... 27
Figure 3.5: Safety catch screens downstream of the subsonic recirculating wind tunnel
test section. ........................................................................................................................ 27
Figure 3.6: Turning cascades used in the elbows of the subsonic recirculating wind
tunnel................................................................................................................................. 28
Figure 3.7: Gravity filter, heat exchanger, modulating valve, flow meter and drain line on
the subsonic recirculating wind tunnel. ............................................................................ 29
Figure 3.8: Test section arrangement with wall plugs on both infield and outfield sides. 31
Figure 3.9: Modified pitot-static probe assembly. ............................................................ 32
Figure 3.10: Dimensionless dynamic pressure profiles measured 0, 30.5, 61 and 85 cm (0,
12, 24 and 33.5 in) (left to right) from the test section inlet. ............................................ 34
Figure 3.11: Near-wall dimensionless dynamic pressure profiles measured at 0, 30.5, 61
and 85 cm (0, 12, 24 and 33.5 in) (left to right) from the test section inlet. ..................... 34
Figure 3.12: Modular instrument panel for the subsonic recirculating wind tunnel. ........ 36
Figure 3.13: 2D profile of the airfoil in cruise configuration showing the approximate
location of static pressure taps and high bandwidth pressure transducers near the airfoil
centerline. .......................................................................................................................... 38
Figure 3.14: OSU version of the simplified high-lift EET airfoil with TE flap deflected.39
Figure 3.15: Model bending produced by loading at high Re and α with DBD plasma at
the LE. ............................................................................................................................... 40
Figure 3.16: Setup used for distinguishing flow phenomena from EMI. ......................... 47
xiii
Figure 3.17: Example of pressure spectra used for distinguishing flow phenomena from
EMI. .................................................................................................................................. 47
Figure 3.18: Example of randomly occurring voltage spikes in pressure traces due to
EMI. .................................................................................................................................. 48
Figure 3.19: PIV experimental setup for LE airfoil measurements. ................................. 50
Figure 3.20: Example of laser reflections from the model surface and their effect on PIV
data. ................................................................................................................................... 52
Figure 3.21: Sample raw PIV data images for LE (a) and TE (b) experiments. ............... 52
Figure 3.22: Experimental setup for flat plate PIV measurements. .................................. 54
Figure 4.1: Airfoil damage due to arc formation. ............................................................. 63
Figure 4.2: Degradation of the dielectric surface after substantial plasma run-time. ....... 66
Figure 4.3: Simultaneous current and voltage traces for a typical AC-DBD plasma
actuator. ............................................................................................................................. 67
Figure 4.4: Example of Q-V data for AC-DBD plasma. .................................................. 69
Figure 4.5: Dissipated power per unit length as a function of applied voltage. ............... 70
Figure 4.6: AC-DBD time-averaged induced velocity magnitude, W
, in quiescent air
(m/sec)............................................................................................................................... 72
Figure 4.7: Modulation waveforms for AC-DBD plasma actuation................................. 73
Figure 4.8: Phase-averaged AC-DBD plasma induced velocity fields, V , (left, m/sec) and
vorticity, Ω , (right sec-1) in quiescent air for modulation using AM (top) and various BM
(2nd from top to bottom) waveforms. ................................................................................ 75
xiv
Figure 4.9: Mean (a) and rms (b) velocity profiles, U, at x=20 mm for AC-DBD plasma
operating in quiescent air for various BM frequencies at 50% dc. ................................... 78
Figure 4.10: Exponential profiles used to interpolate momentum values for various AC-
DBD plasma modulation frequencies. .............................................................................. 80
Figure 4.11: Mean and oscillatory momentum values for the AC-DBD plasma induced
flow in quiescent air for modulation using AM and various BM waveforms. ................. 82
Figure 4.12: Q-V diagram for AC-DBD plasma actuation using AM at 100 Hz. ............ 84
Figure 4.13: AC-DBD plasma integrated energies for various modulation waveforms... 86
Figure 4.14: Typical voltage, current (a) and power, energy (b) traces for an NS-DBD
plasma actuator. ................................................................................................................ 89
Figure 4.15: Startup vortex (b) and dielectric charging effect (c) created by NS-DBD
plasma actuator. ................................................................................................................ 93
Figure 4.16: Mean velocity profiles, U, for NS and AC-DBD plasma operating in
quiescent air with 2 kHz carrier frequency using 100% (a) and 50% (dc) at 90 Hz
measured 20 mm downstream. ......................................................................................... 95
Figure 4.17: Schlieren imaging of compression waves generated by NS-DBD plasma
actuator viewed along the major axis of the actuator. ...................................................... 97
Figure 4.18: Schlieren photography of vertical density gradients for NS-DBD plasma
generated compression waves viewed along the minor axis of the actuator. ................... 98
Figure 4.19: Schlieren photography of horizontal density gradient for NS-DBD plasma
generated compression waves viewed along the minor axis of the discharge. ................. 98
Figure 5.1: Simplified high-lift airfoil with LE droop and TE flap. ............................... 103
xv
Figure 5.2: Baseline CL vs. α for sample Re and δf. ........................................................ 105
Figure 5.3: Baseline CP behavior for Re=410k and δf=20o at various α. ........................ 105
Figure 5.4: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.95 for Re=410k
and α=0o at various δf. ..................................................................................................... 106
Figure 5.5: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.95 for Re=410k
and δf=20o at various α. ................................................................................................... 107
Figure 5.6: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.90 for α=0o and
δf=30o at various Re. ....................................................................................................... 107
Figure 5.7: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.40 for Re=750k,
δf=0 at various α. ............................................................................................................. 108
Figure 5.8: Baseline comparison of OSU and NASA (Melton et al. 2006) results for
tripped and untripped behavior of CL vs. α for Re=750k and various δf. ........................ 109
Figure 5.9: OSU (a) and NASA (b) CP curves at Re=750k, α=0 and δf=0 (Melton et al.
2003). .............................................................................................................................. 109
Figure 5.10: OSU (a) and NASA (b) stall characteristics at Re=410k and δf =12o (Melton
2006). .............................................................................................................................. 110
Figure 5.11: Re and δf effects on OSU airfoil CP distributions. ...................................... 113
Figure 5.12: OSU (a) and NASA (b) PSD of fluctuating pressure, cp, at Re=240k, α =0, δf
=45, δs =-25 in (Melton et al. 2006). ............................................................................... 114
Figure 5.13: 3D behavior of CP for Re=410k, α=0 and δf=20 (a), δf =30 (b) and δf =40 (c).
......................................................................................................................................... 115
xvi
Figure 5.14: PIV-measured boundary layer velocity profile with power law fit for
Re=240k, α=0, δf=30. ...................................................................................................... 117
Figure 6.1: Example of a DBD plasma actuator applied at the flap shoulder (x/c=0.77) of
the simplified NASA EET airfoil, δf=30o. ...................................................................... 120
Figure 6.2: Effects of passive (power off) DBD plasma actuator on CP (a) and PSD of
fluctuating pressure, cp, at x/c=0.90 (b). ......................................................................... 121
Figure 6.3: Effect of dimensionless actuation frequencies FL,m+ (a) and F* (b) on ΔCL at
α=0o for various Re and δf. .............................................................................................. 123
Figure 6.4: Baseline and controlled CP behavior for Re=240k, α=0o and δf =20. .......... 125
Figure 6.5
: Baseline and controlled (dc=50%) time-averaged streamlines for Re=240k,
α=0o and δf =20o. ............................................................................................................. 127
Figure 6.6: Baseline and controlled (dc=50%) time-averaged normalized vorticity, *
Ω ,
for Re=240k, α=0o and δf =20o. ....................................................................................... 129
Figure 6.7: Baseline and controlled time-averaged vorticity magnitude, *Ω at x/c=1.1 for
Re=240k, α=0o and δf =20o. ............................................................................................ 130
Figure 6.8: Phase-averaged normalized vorticity fields (ΔΦ=π), *
Ω , for Re=240k, α=0o
and δf =30o forced at FL+=8.5, FL,m
+=0.4 (F*=1). .......................................................... 132
Figure 6.9: Baseline and controlled PSD of fluctuating pressure, cp, measured at x/c=0.90
Re=240k, α=0o and δf=30o. ............................................................................................. 133
Figure 6.10: Spatial correlation of the v component of velocity, Rvv, at y/c=yT for
Re=240k, α=0o and δf=30o for baseline and controlled (dc=50%) flows. ....................... 133
xvii
Figure 6.11: First four normalized baseline and controlled (AM) POD modes of the v
component of velocity, ( )1 4 y−∗ϕ , for Re=240k, α=0o and δf=30o. ................................. 136
Figure 6.12: Modal energy (a) and cumulative energy (b) for baseline and controlled
(AM) POD modes for Re=240k, α=0o and δf=30o. ......................................................... 137
Figure 6.13: Effect of modulation waveform on ΔCL for Re=240k, α=0o and variable δf
when forcing at FL+=8.5, F*=1. ...................................................................................... 139
Figure 6.14: ΔCL and <J/ρ> as a function of BM dc for Re=240k, α=0, δf=30, FL+=8.5,
FL,m+=0.4, F*=1. ............................................................................................................. 140
Figure 6.15: Effect of Re, α, and δf on ΔCL for forcing at F*=1 using dc=50%. ............ 143
Figure 6.16: Mean U velocity profiles (a) and Reynolds stresses uu (b) and vv (c) used
in CD calculations for Re=240k, α=0 and δf=20 at x/c=1.1, 1.2, 1.3, 1.4 and 1.5 (left to
right). ............................................................................................................................... 146
Figure 6.17: DBD plasma actuators straddling the flap shoulder. .................................. 149
Figure 6.18: Normalized time-averaged vorticity magnitude profiles, *Ω at x/c=1.1 for
Re=240k, α=0, δf=20 for baseline and AC-DBD actuators operating in phase at FL+ =12.7
and F*=1. ........................................................................................................................ 149
Figure 6.19: Effect of reversing AC-DBD momentum direction on CP. ........................ 151
Figure 6.20: First four normalized baseline and controlled (AM) POD modes of the v
component of velocity, ( )1 4 y−∗ϕ , for Re=240k, α=0o and δf=30o with reverse actuator at
x/c=0.77. .......................................................................................................................... 152
Figure 6.21: NS-DBD effect on CP for Re=240k, α=0 and δf=30. .................................. 154
xviii
Figure 6.22: NS-DBD effect on CP for Re=240k, α=0 and δf=30 with increased energy
input. ............................................................................................................................... 155
Figure 7.1: DBD plasma actuator mounted at the LE of the OSU simplified NASA EET
airfoil. .............................................................................................................................. 158
Figure 7.2: CL curves for the OSU airfoil with and without passive LE actuator and Re
overshoot compared to NASA (Melton et al. 2006). ...................................................... 159
Figure 7.3: Various examples of baseline (plasma off) behavior observed when an
actuator is applied to the LE. .......................................................................................... 161
Figure 7.4: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at
α=10o. .............................................................................................................................. 163
Figure 7.5: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at
α=12o. .............................................................................................................................. 164
Figure 7.6: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at
α=14o. .............................................................................................................................. 166
Figure 7.7: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at
α=16o. .............................................................................................................................. 168
Figure 7.8: Effect of NS-DBD plasma actuation at x/c=0 on CP for Re=750k at α=16o
showing preference for higher frequency forcing. .......................................................... 170
Figure 7.9: Comparison of the effects of NS (a) and AC (b) DBD plasma on CP for
Re=750k. ......................................................................................................................... 172
Figure 7.10: Effect of Fc+ of AC and NS-DBD plasma at x/c=0 on suction side CP at
x/c=0.05. .......................................................................................................................... 174
xix
Figure 7.11: Photograph of the DBD plasma actuator mounted near the airfoil LE in
reverse arrangement ds=6 mm from x/c=0. .................................................................... 176
Figure 7.12: Effect of AC and NS-DBD plasma for reverse actuator arrangement ds=6
mm from x/c=0. ............................................................................................................... 177
Figure 7.13: Effect of Fc+ of NS-DBD plasma for reverse actuator arrangement ds=6 mm
from x/c=0 on suction side CP at x/c=0.05 at Re=750k. ................................................. 178
Figure 7.14: Effect of Fc+ of NS-DBD plasma at for reverse actuator arrangement ds=6
mm from x/c=0 on suction side CP at x/c=0.05 at Re=750k and Re=1000k. .................. 179
Figure 7.15: Time-averaged behavior of CP (a) and normalized vorticity, *
Ω , for baseline
(b) and NS-DBD forcing at Fc+=4 (c) at Re=750k, α=12o. ............................................. 181
Figure 7.16: Phase-averaged normalized velocity fluctuations, *v , for NS-DBD plasma
forcing at Fc+=4 (c) at Re=750k, α=12o. ......................................................................... 182
Figure 7.17: Time-averaged behavior of CP (a) and normalized vorticity, *
Ω , for baseline
(b) and NS-DBD forcing at Fc+=4 (c) at Re=750k, α=14o. ............................................. 183
Figure 7.18: Phase-averaged normalized velocity fluctuations, *v (a), swirling strength,
*ciλ , (b) and vorticity,
*Ω , (c) for NS-DBD forcing at Fc
+=4 for Re=750k, α=14o. ...... 186
Figure 7.19: Time-averaged behavior of CP (a) and normalized vorticity, *
Ω , for baseline
(b) and NS-DBD forcing at Fc+=0.6 (c) at Re=750k, α=16o. .......................................... 188
Figure 7.20: Phase-averaged normalized velocity fluctuations, *v (a) and swirling
strength, *ciλ , (b) for NS-DBD forcing at Fc
+=0.6 for Re=750k, α=16o. ........................ 189
xx
Figure 7.21: Time-averaged behavior of CP (a) and normalized vorticity, *
Ω , for baseline
(b) and NS-DBD forcing at Fc+=5.6 (c) at Re=750k, α=16o. .......................................... 191
Figure 7.22: Phase-averaged normalized velocity fluctuations, *v (a) and swirling
strength, *ciλ , (b) for NS-DBD forcing at Fc
+=5.6 for Re=750k, α=16o. ........................ 192
Figure 7.23: Time-averaged behavior of CP (a) and normalized vorticity, *
Ω , for baseline
(b) and NS-DBD forcing at Fc+=11.3 (c) at Re=750k, α=16o. ........................................ 194
Figure 7.24: Phase-averaged normalized velocity fluctuations, *v (a) and swirling
strength, *ciλ , (b) for NS-DBD forcing at Fc
+=11.3 for Re=750k, α=16o. ...................... 195
Figure 7.25: Effect of AFC with NS-DBD plasma actuation on CL for the OSU airfoil.197
Figure 7.26: Effect of PFC with using LE droop on CL for the simplified NASA EET
airfoil (Melton et al. 2005) .............................................................................................. 198
xxi
Nomenclature
a = POD modal amplitude c = model chord, 25.4 cm CD = sectional drag coefficient Cdp = sectional pressure drag coefficient CE = power coefficient, P/q∞U∞c CL = sectional lift coefficient CP = pressure coefficient cP = fluctuating pressure coefficient ˆPc = Fourier transform of cP denoted by ^ Cv = Specific heat at constant volume Cµ = momentum coefficient, J/q∞c <Cµ> = oscillatory momentum coefficient, <J>/q∞c Cµ,tot = total momentum coefficient, Cµ + <Cµ> Cµ,Δp = modified momentum coefficient based on pressure difference, Δpz/q∞c ds = arc length measured along airfoil suction surface from x/c=0 E = energy f = frequency fc = plasma carrier frequency fm = modulation frequency fTE = characteristic frequency measured on the flap Fc
+ = reduced frequency based on model chord, fc/U∞
FL+ = reduced frequency based on flap length, fL/U∞
FL,m+ = reduced modulation frequency based on flap length, fmL/U∞
F* = normalized modulation frequency, fm/fTE h = dielectric thickness J = time-averaged actuator induced momentum <J> = oscillatory actuator induced momentum k = wind tunnel calibration constant l = actuator length L = flap length, 6.35 cm m = mass N = sample size P = time-averaged power consumed by plasma p = static pressure po = stagnation pressure
xxii
Δp = pressure difference across compression wave Q = charge q = dynamic pressure, ρU2/2 q∞ = freestream dynamic pressure, ρU∞
2/2 Rvv = Normalized spatial correlation of v Re = chord based Reynolds number, U∞c/ν ΔT = temperature rise due to plasma heating t = time u = fluctuating streamwise velocity U = streamwise velocity U = time-averaged velocity denoted by _ U = phase-averaged velocity denoted by ~ U* = normalized velocity, U/ U∞ Urms = rms of streamwise velocity U∞ = freestream velocity Vac = AC voltage v = fluctuating vertical velocity V = instantaneous vertical velocity W
= 2D vector field [U,V] xsp = length of separation region x = two dimensional space (x,y) x/c = normalized streamwise coordinate y/c = normalized vertical coordinate yT = y/c coordinate of the tail of the deflected trailing edge flap z = width of fluid heated by plasma α = angle of attack in degrees δf = flap deflection angle in degrees ε = dielectric constant λci = swirling strength λci
* = normalized swirling strength, λcic/U∞ ρ = density ν = kinematic viscosity δ* = boundary layer displacement thickness θ = boundary layer momentum thickness σ = sample standard deviation H = shape factor, δ*/ θ Ω = spanwise component of vorticity Ω* = normalized spanwise vorticity, Ωc/U∞ Φ = POD mode Φ* = Normalized POD mode, Φ/U∞ CI = 95% confidence interval PSD = power spectral density TE = trailing edge
xxiii
LE = leading edge PFC = passive flow control AFC = active flow control AM = amplitude modulation with sine wave BM = burst modulation with square wave DBD = dielectric barrier discharge AC = alternating current DC = direct current NS = nanosecond pulse LAFPA = localized arc filament plasma actuator EHD = electrohydrodynamic EMI = electromagnetic interference PIV = particle image Velocimetry ZNMF = zero net mass flux HV = high voltage dc = duty cycle
1
Chapter 1: Introduction
High-lift airfoils typically employ leading edge (LE) droops or slats and trailing
edge (TE) flaps that can be deflected during take-off or landing and stowed during cruise
(Figure 1.1). The former acts to extend the stall angle while the latter acts to increase the
airfoil camber by creating greater CL at a given α at the expense of decreasing the stall
angle (Hoener and Borst 1975). Consequently, the two devices are often used in tandem.
Simple flaps can impose a penalty due to flow separation that occurs when the
momentum of fluid in the boundary layer is not sufficient to overcome wall friction and
the adverse pressure gradient encountered as it travels over the deflected surface.
Traditional methods of eliminating flow separation on high-lift airfoils utilize multi-
element slats and flaps that allow mixing of fluid between the pressure (bottom) and
suction (top) sides like those shown in Figure 1.2. These systems are effective for
augmenting lift, but increase mechanical complexity, manufacturing cost, weight and
parasitic drag even when stowed during cruise. A system study indicates that significant
decreases in manufacturing cost, weight and drag could be realized if the complex multi-
element high-lift system is simplified like Figure 1.1 while maintaining CL,max (McLean
et al. 1999). To obtain similar performance, passive flow control (PFC) previously in the
2
form of slots must be replaced by some active flow control (AFC) device that would not
reintroduce additional detrimental factors.
Figure 1.1: Simplified high-lift airfoil system with deflected LE droop and TE simple flap.
Figure 1.2: Typical high-lift airfoil system with deflected multi-element LE slat and TE flap (Lin and Dominik 1997).
This works explores the use of two types of dielectric barrier discharge (DBD)
plasma actuators for controlling LE and TE separation on a further simplified high-lift
airfoil (Figure 1.3). The most common of these is the DBD plasma actuator driven by an
AC voltage waveform (AC-DBD). This device has become very popular in recent years
3
due to its simple construction, lack of moving parts, fast time response and low power.
AC-DBD plasma is well-established as a flow control actuator at airfoil LEs for
freestream velocities up to 30 m/sec (Moreau 2007). Its use on the more challenging
problem of controlling flow separation at the TE flap shoulder has not been fully
explored and is one of the primary focuses of this work. The control of flow separation
over a deflected TE flap using more common piezoelectric type actuators has been
documented and serves as a basis for judging the AC-DBD plasma on a similar scaled
version of the same airfoil model (Pack et al. 2002; Melton et al. 2003; Melton et al.
2004; Melton et al. 2005; Melton et al. 2006; Melton et al. 2007).
Figure 1.3: Further simplified high-lift airfoil system with deflected TE flap.
DBD plasma actuators driven by repetitive nanosecond pulses (NS-DBD) are less
established, but appear quite promising for flow control applications at higher speeds.
The construction of this actuator is identical to the AC version, but it has shown airfoil
LE separation control authority in isolated tests up to Mach 0.74 which is well beyond
any published results of AC-DBD plasma control authority (Roupassov et al. 2009). The
4
efficacy of this unique device is primarily explored at the LE of a supercritical airfoil and
compared to the more established AC-DBD as well as published results for a common
PFC device on a similar scaled version of the same model.
A characterization of both types of DBD plasma actuators is performed in
quiescent air to highlight the substantially different behavior created by producing the
discharge with these radically different input waveforms. In addition to the obvious
metric of CL, detailed flow diagnostics reveal the physics behind separation control and
lift enhancement using both types of DBD plasma actuators. This work is intended to
provide a benchmark comparison of AC-DBD plasmas to piezoelectric actuators for
controlling flow separation over a deflected TE flap while clearly documenting the
potential of NS-DBD plasma actuators for LE airfoil separation control.
5
Chapter 2: Background
Separation control is a broad and widely studied topic and thorough reviews on its
various applications have been published (Gad-el-Hak and Bushnell 1991; Greenblatt and
Wygnanski 2000). Passive separation control techniques which generally constitute
geometric changes such as slotted LE slats and TE flaps are employed on many
operational aircraft. The slotted portions of the PFC devices are necessary to energize the
boundary layer on the suction surface via mixing with high momentum fluid from the
pressure surface allowing it to follow the curvature of the deflected system. These control
elements are effective for reducing or eliminating separation during takeoff and landing
when aircraft speeds are low and high-lift is required. However, they are expensive to
manufacture, heavy, mechanically complex and introduce parasitic drag effects even
when stowed during cruise. Despite these drawbacks, the benefits of PFC outweigh the
incurred cost created by their application to the aerodynamic surface.
2.1 Separation Control
Active separation control has gained popularity in recent years due to its potential
for maintaining or enhancing the benefits of passive control techniques without the
penalty associated with many of the detrimental factors listed above. The main difference
between PFC and AFC strategies is that the latter can be turned on and off by command
6
at time scales consistent with relevant flow dynamics. This also gives that potential for
implementation in a feedback control system that coupled with adequate sensors and
controller could create even greater benefits in flight efficiency and maneuverability. A
complete review of active separation control is a subject in itself. Rather, the following
background information focuses on separation control studies that examine the effect of
nominally two-dimensional actuation on two dimensional airfoil models. It should be
noted that governing parameters and their optimal values such as frequency scale
favorably from 2D airfoil to 3D wing configurations at modest sweep angles (Seifert and
Tillman 2009).
2.1.1 LE Separation Control
Technological advances over the last few decades have allowed researchers to
more fully explore the wide parameter space associated with this research topic.
Accordingly, significant advances in the understanding of separated flow phenomena in
response to actuation have followed. Among the most widely accepted is that unsteady
actuation via pulsed blowing, pulsed suction or both (zero net mass flux (ZNMF)) is
more effective than steady forcing (Seifert et al. 1996). The range of effective
dimensionless frequencies associated with separation control is on the order of unity. The
dimensionless frequency often termed reduced frequency, F+, or more commonly
Strouhal number is defined as:
spfxF
U+
∞
= 2.1
7
where f is the forcing frequency, xsp is the length of the separated region and U∞ is the
freestream velocity (Darabi and Wygnanski 2004; Glezer et al. 2005). This parameter
underscores the importance of the characteristic length scale of separated flow
phenomena, xsp, which is generally assumed to be the length of the separation zone over
the body in question (Seifert et al. 1996). This variable also depends on the response of
the flow to actuation (Wygnanski 2004). Physically, the reduced frequency of unity
requires that a perturbation must be introduced during the time that the freestream flow
propagates over the separated region. The importance of actuator location is closely
related to this expression since the shear layer created between the freestream and the low
speed separated region by nature selectively amplifies small perturbations if they are
introduced near its receptivity region. The optimum choice of this location for unsteady
actuation is generally at or slightly upstream of the separation line. This ensures that the
shear layer is excited by the control perturbations near its receptivity point. Successful
introduction of such forcing creates large spanwise vortices that develop via the Kelvin-
Helmholtz instability. These vortices encourage momentum transport between the
freestream and separated region thus reattaching the flow (Darabi and Wygnanski 2004;
Melton et al. 2005). Separation control by forcing at higher frequencies (F+ > 10) has
been classified as a different phenomenon characterized by enhanced dissipation. The
more desirable aspect of this forcing is a reduction in flow unsteadiness and subsequent
dynamic loads on the aerodynamic surfaces under the separated region (Amitay and
Glezer 2002).
8
2.1.2 TE Separation Control
Studies of specifically TE airfoil separation control are less prevalent, but two
major efforts in recent years have resulted in additional understanding of this system. The
ADVINT program on the Boeing Tilt-Wing SSTOL transport (Kiedaisch et al. 2006;
Nagib et al. 2006; Smith et al. 2006; Kiedaisch et al. 2007; Nagib et al. 2007; DeSalvo et
al. 2010) and a parallel work by NASA (Pack et al. 2002; Melton et al. 2003; Melton et
al. 2004; Melton et al. 2005; Melton et al. 2006; Melton et al. 2007) have produced
significant developments using AFC via both nonzero and ZNMF periodic excitation
(Seifert and Tillman 2009). Perhaps the most significant of these is that more momentum
input is required in comparison with LE control (Melton et al. 2006). This is
commensurate with the existence of a thicker, turbulent boundary layer that develops
along the main element of the airfoil. Consequently, simply tripping the boundary layer
approaching the flap shoulder, which can be effective for LE separation control at low α,
is not sufficient for controlling separation at high Re and flap deflection, δf. These results
also support LE separation control findings that show greater centripetal acceleration
created by airfoil surface curvature requires additional momentum for realizing similar
control authority (Greenblatt and Wygnanski 2003).
Because of these challenges, it is difficult to fully attach the flow over a simple
TE flap. Instead, separation control and circulation control are inherently linked in these
systems (Cerchie et al. 2006). Evidence of this effect is observed when increases in CL
are attributed to upstream effects rather than complete or partial reattachment to the flap
surface (Kiedaisch et al. 2006; Melton et al. 2006). This is especially true for low
9
frequency forcing (F+~0.3-1.5) if separation is not radically delayed (Melton et al. 2004).
Such behavior does not occur in more canonical studies of separation control for wall-
mounted humps (Seifert and Pack 2002). Additionally, the control authority for a given δf
is highly sensitive to actuation location in regions of high surface curvature encountered
near the flap shoulder (Melton et al. 2006). This is especially true for typical ZNMF type
actuation issuing from small often 2D slots on the model surface. Such results can be
traced back to the importance of separation location on the actuator placement in that
control introduced at or slightly upstream of the separation point generally produces the
greatest effect (Greenblatt and Wygnanski 2000).
For purely increasing CL, high frequency excitation (F+ > 10) is less efficient in
terms of both momentum and power than low frequency modulated versions of the same
signal at F+~0.3-1.5 (Seifert et al. 1996; Melton et al. 2006). The latter mechanism relies
on the existence of natural instabilities in the flow field which can selectively amplify
small disturbances if introduced near a region of strong receptivity (Darabi and
Wygnanski 2004). In either the low or high frequency forcing cases above, amplifying the
disturbance generally serves to increase the control authority (Melton et al. 2004;
Kiedaisch et al. 2006; Melton et al. 2006), but the scaling laws governing these effects
are not well-defined (Seifert and Tillman 2009). Additional open questions on the
directional effects of the momentum addition as well as the size and shape of the slot or
hole for which the momentum is introduced remain. It also appears that flows with large
separated regions such as those over deflected TE flaps may benefit substantially from
3D forcing and the subsequent production of streamwise vorticity (DeSalvo et al. 2010).
10
2.2 Plasma Actuators
The recent interest in plasma actuators for aerodynamic flow control is motivated
by their simple construction, lack of moving parts, fast time response and low power. The
two types of plasma actuators used in this work are believed to function through
electrohydrodynamic (EHD) and thermal mechanisms. The former relies on momentum
transfer from charged species to neutral air molecules while the latter works through
Joule heating effects. These two very different mechanisms are produced using a DBD
arrangement. The construction of the device is identical in both cases, but the input
waveform is substantially different. DBD plasmas are relatively new to the aerodynamic
community, but have long been used in a variety of industrial applications such as ozone
generation (Kogelschatz 2003).
2.2.1 AC-DBD Plasma Actuators
The DBD plasma actuator for aerodynamic flow control is usually composed of
two electrodes separated by a dielectric material arranged in an asymmetric fashion
shown in Figure 2.1. Application of a sufficiently high voltage (≥5 kV) AC (1-10 kHz)
signal between the electrodes weakly ionizes the air over the dielectric covering the
encapsulated electrode. The dielectric barrier allows the generation of a large volume of
plasma by preventing the discharge from collapsing into an arc. The DBD plasma
actuator is a self limiting device in that the accumulation of charged particles onto the
dielectric surface opposes the electric field requiring consistently higher voltages to
sustain the discharge. This is circumvented using an AC waveform which, because of a
change in polarity, creates movement of charged species back and forth between the
11
exposed electrode and the dielectric surface at the AC driving frequency. The movement
of these charged particles transfers momentum to the flow via ion-neutral collisions. In
quiescent conditions, the asymmetric plasma actuator creates suction above the exposed
electrode and a pseudo wall jet over and downstream of the covered electrode that has
been shown to follow a skewed Gaussian profile (Hoskinson et al. 2008). Maximum near
wall streamwise velocities generated by a single actuator in air are less than 10 m/sec
measured a few millimeters from the wall just downstream of the plasma extent (Moreau
2007; Corke et al. 2009).
Figure 2.1: Typical asymmetric DBD plasma actuator geometry (Corke et al. 2010).
The induced flow is predominantly directed away from the exposed electrode due
to the asymmetry of the actuator geometry and behavior of the discharge over the two
waveform half cycles (Enloe et al. 2004a; Enloe et al. 2008b). In air, this force
production is dominated by interactions from negatively charged species, most notably
12
the negative oxygen ion (Enloe et al. 2006; Kim et al. 2007; Font et al. 2010). Each cycle
of the AC waveform produces a dominant velocity pulse due to movement of these
species. Time-resolved force measurements show that the dominant momentum
producing phase of the waveform is associated with the negative going half of the
negative half cycle and that this can produce up to 97% of the generated thrust (Enloe et
al. 2008b). Referring to Figure 2.2, this region is seen as the portion of the AC waveform
with negative polarity and negative slope.
Figure 2.2: Simultaneous traces of voltage and current for a typical AC-DBD plasma actuator.
13
This portion of the discharge cycle is characterized by more uniform behavior
where negative species are deposited on to the dielectric surface. This is often referred to
as the forward stroke with respect to electron movement (Corke et al. 2010). The more
diffuse behavior can be seen in light emission, current measurements and high speed
photography which all show more uniform low intensity microdischarges in comparison
to the positive half cycle. The reverse or back discharge is substantially more filamentary
in nature with concentrated streamers at distinct, but unpredictable locations.
High speed photography shown in Figure 2.3 has revealed this behavior in detail.
The current traces in Figure 2.2 correlate with this behavior as the current spikes in the
reverse/back discharge case are less uniform and reach peaks of >200 mA while the
forward stroke is more uniform with lower intensity peaks. It has been postulated that this
asymmetry is created by the inability of the dielectric surface to freely give electrons to
the bare electrode similar to the schematic in Figure 2.4 (Corke et al. 2010). This
asymmetric behavior has motivated the use of a negative sawtooth high voltage signal
which acts to extend the lifetime of the momentum generating portion of the waveform
thereby increasing the generated thrust (Enloe et al. 2004a; Balcon et al. 2009).
14
Figure 2.3: High-speed photography of the forward stroke (a) and reverse/back stroke (b) of a typical AC-DBD plasma actuator (Enloe et al. 2008a).
Figure 2.4: Schematic of negatively charged species movement for the forward (a) and reverse/back stroke (b) in a typical AC-DBD plasma actuator (Enloe et al. 2008a).
a)
b)
15
It is generally accepted that AC-DBD plasma body forces are voltage driven
phenomenon as both the thrust and dissipated power are proportional to Vac7/2 when the
device is operating in the normal glow regime similar to Figure 2.5a (Corke et al. 2009).
Beyond this region, the dissipated power tends to increase while the induced flow
saturates at some maximum value determined by the excitation waveform and actuator
construction (Forte et al. 2007). The power increase seen beyond the Vac7/2 regime is
believed to be dissipated through heating in randomly occurring thick filaments that are
constant in space and are easily visible with the naked eye (Figure 2.5b). These should
not be confused with the microdischarges in Figure 2.3. Note that the localized heating in
these thick filaments eventually leads to a breakdown of the dielectric material and the
formation of an arc filament which eliminates velocity production.
Figure 2.5: Visual appearance of a typical ~10 cm long AC-DBD plasma actuator operating in the normal glow regime (a) and at maximum thrust (b) (Thomas et al. 2009)
16
There is an optimal carrier frequency, fc, for DBD plasma induced thrust that is
dependent on both the ambient gas and the bulk capacitance of the dielectric, ε/h, where ε
and h are the dielectric constant and thickness respectively. In addition, for a given
dielectric operating at its optimized frequency, the thrust created from the actuator is
dependent on the bulk capacitance of the dielectric (Corke et al. 2009). For a given
voltage input, dielectrics with larger bulk capacitance tend to produce greater thrust when
operated in the Vac7/2 region due to an enhanced electric field. Consequently, an idealized
dielectric will have a large dielectric constant and a small thickness to create a larger bulk
capacitance. The caveat here is that dielectrics with smaller thickness generally have
lower dielectric strength and cannot withstand high voltages without entering the arc
regime. The use of relatively thick materials increases dielectric strength at the expense
of requiring higher voltages to initially ionize the gas and sustain the discharge in the
Vac7/2 region. This results in high values of thrust, but an increased actuator thickness and
voltage requirement (Corke et al. 2009). Using this knowledge, the generated thrust from
an AC-DBD plasma actuator can be increased by an order of magnitude using thicker
dielectric materials with lower dielectric constants and higher AC voltages (Thomas et al.
2009).
As previously discussed, the force production of AC driven DBD plasma
actuators is strongly dependent on the oxygen content of the environment (Enloe et al.
2006; Kim et al. 2007) However, ambient conditions such as pressure and humidity also
play a role. The dependence of thrust generation on these parameters has not been
exhaustively studied, but early reports suggest the actuator may still be viable under these
17
variable conditions (Abe et al. 2008; Benard et al. 2009a). Still, considerably more work
must be done to fully characterize their performance across a wide range of flight
conditions. Other aspects such as the effect of charge accumulation on the dielectric
surface appear to be crucial, but conflicting reports exist on the wheter this has a positive
or negative effect on the induced flow (Opaits et al. 2009b; Font et al. 2010). Despite
these many open questions, strides continue to be made for implementation as AC-DBD
plasma actuators have been used with varying degrees of success in feedback control and
flight testing (Patel et al. 2007; Sidorenko et al. 2008).
Numerous models have been developed for DBD plasma actuators that include
varying degrees of complexity in simulating plasma chemistry (Golubovskii et al. 2002;
Singh and Roy 2008; Likhanskii et al. 2009). These approaches are certainly useful, but
depending on the level of complexity can be computationally expensive. In many
practical flow control applications an empirical space-time lumped element circuit model
appears sufficient for predicting body forces that can be used in the Navier Stokes
equations for design purposes (Jayaraman and Shyy 2008; Rizzetta and Visbal 2009;
Corke et al. 2010).
2.2.2 NS-DBD Plasma Actuators
The AC-DBD plasma actuator is currently undergoing rapid development, but the
most important aspect of this remains the low momentum of the induced flow of the
device which limits control authority at high speed. This continues to be explored through
optimization of a single actuator, the use of multiple actuators and more novel
arrangements like sliding discharges that rely on additional DC bias voltages (Forte et al.
18
2007; Thomas et al. 2009; Corke et al. 2010). The common goal in all these cases is to
increase the velocity generated by the device. This is certainly a useful endeavor, but all
flow control actuators that rely on ZNMF momentum introduction eventually lose
effectiveness as requirements increase at progressively higher flow speeds.
This has motivated the study of actuation techniques based on other mechanisms,
namely thermal effects. Flow control using this method of actuation can be developed via
bulk or localized heating and was originally intended for use in supersonic flows to
weaken or manipulate shock structure (Palm et al. 2003; Adelgren et al. 2005). The
former requires the production of a very large volume of diffuse plasma which is
expensive from a power standpoint and difficult if not impossible to sustain at
atmospheric pressure. The latter can be produced using high power lasers or microwave
beams, but a more practical method is through an arc discharge between pin electrodes.
This has been extensively explored with localized arc filament plasma actuators
(LAFPAs) in high speed jets (Samimy et al. 2007). The control mechanism in this case is
believed to be a thermal effect that generates compression waves. The flow amplifies
these LAFPA generated perturbations when introduced at the correct frequency and
mode. LAFPAs also have the advantage of high bandwidth which allows them to be used
to excite and manipulate a variety of flow instabilities (Utkin et al. 2007).
These devices appear nearly ideal for exciting various instabilities in free jets.
However, they generate a considerable amount of heat requiring they be housed in high
temperature non-conducting materials such as ceramics. Additionally, multiple arc
filaments are required for control authority in most systems. For, example eight actuators
19
distributed around the circumference of a one inch diameter jet have considerable control
authority. For transport aircraft separation control applications, placing multiple pin
electrodes along an aircraft wing may not be practical.
The use of AC-DBD plasmas for flow control via thermal mechanisms was
originally considered, but this has since been dispelled by multiple studies (Enloe et al.
2004a; Jukes et al. 2006; Sung et al. 2006). Instead, a different high voltage waveform in
the form of very short pulses is employed. The construction of this device is identical to
the AC-DBD plasma actuator, but the input waveform now has pulse width of a few tens
of nanoseconds with pulse voltages in the range 10-50 kV. The NS-DBD plasma actuator
relies on a very short rise time to generate rapid localized heating by the plasma. The
dielectric barrier along with the short pulse width serves to prevent the discharge from
collapsing into an arc. This allows the thermal effect to become distributed along the span
of the device. The discharge can now be sustained on simple plastic dielectric surfaces
rather than the more extreme ceramics required in the arc filament case.
The physics of NS-DBD plasma actuators as flow control devices have not been
extensively studied. This stems from the developmental nature of this type of discharge
as well as the difficulty in procuring power supplies capable of producing such
waveforms. EHD effects produced by these devices appear quite weak and compression
waves generated by NS-DBD plasma actuators have been detected in both modeling and
experiments in still air (Roupassov et al. 2009). Details of these effects as well as
quantitative measurements of heating and compression wave strength remain open for
investigation, but estimates of near surface temperature rises of ~200 K corresponding to
20
compression waves at Mach 1.13 have been suggested (Roupassov et al. 2009). There
have also been attempts to combine both AC and NS-DBD waveforms in hopes of
increasing EHD effects, but these results are still very preliminary (Opaits et al. 2009b).
2.3 Separation Control with Dielectric Barrier Discharge Plasma Actuators
Early examples of AC-DBD plasma actuators as flow control devices
demonstrated their potential for boundary layer and LE airfoil separation control
applications (Roth et al. 2000; Post and Corke 2004). More recently, experimental studies
using such devices have broadened to include jet mixing, cavity tone attenuation, noise
control and aero-optics (Benard et al. 2007; Chan et al. 2007; Freeman and Catrakis
2008; Thomas et al. 2008). While these new applications have gained popularity in recent
years, the majority of DBD plasma work is still applied to separation control. These
actuators are particularly appealing for this application due to the nature of their induced
flow when arranged asymmetrically. In this configuration, the induced flow produces a
wall jet with maximum velocity a few mm away from the surface which is often
amenable for influencing boundary layers.
The mechanism responsible for separation control by AC-DBD plasma is most
often associated with the wall jet generation described above, but whether this results in
boundary layer tripping, energizing or amplification of instabilities depends on the flow
system under consideration. For separation control explicitly, the state of the boundary
layer (laminar or turbulent) just upstream of the actuator will also play a role.
21
Like other methods of periodic excitation (Greenblatt and Wygnanski 2000), AC
driven DBD plasma actuators are often most effective for airfoil separation control and
lift enhancement when excitation is created with F+ on the order of unity (Huang et al.
2006; Greenblatt et al. 2008; Patel et al. 2008). To operate in this fashion, the actuator
must be powered with a sufficiently high fc to produce the plasma (1-10 kHz) and
modulated at a lower frequency to excite the longer wavelength instabilities associated
with most separated flow dynamics. This behavior is analogous to ZNMF type actuation
created by piezoelectric diaphragms which produce the highest intensity fluctuations
when excited near the resonant frequency of the disc and/or cavity which is often on the
order of a few kHz. Studies of separation control with AC-DBD plasma actuators often
assume the flow does not respond to perturbations created by the high frequency carrier
signal, but instead responds as if exposed to steady wall jet (Thomas et al. 2008). For
many low-speed applications, this is true because the instabilities involved are not
receptive to high frequency perturbations and instead feel their effect as a quasi-steady
phenomenon. However, it has been confirmed that in air, the movement of charged
species in the plasma creates a dominant velocity pulse at the fc of plasma generation and
thus suggests the possibility of using AC-DBD plasma actuators for high frequency
forcing applications (Glezer et al. 2005) if sufficient amplitude can be produced
(Takeuchi et al. 2007; Boucinha et al. 2008).
In terms of frequency, the excitation signals driving piezoelectric and AC-DBD
plasma of actuators are quite similar, but the characteristics of the momentum production
and delivery are fundamentally different. In a typical ZNMF type actuator, periodic
22
blowing and suction is 180o out of phase and occurs through a slot or orifice mounted at
an angle relative to the freestream or surface. AC-DBD plasma actuators are completely
surface mounted and the exact location at which the plasma actuator accomplishes
control is not immediately obvious although actuators placed at or slightly upstream of
the separation location give favorable results (Huang et al. 2006; Sosa et al. 2007;
Jolibois et al. 2008). This appears consistent with modeling results that show the highest
force density associated with such devices is near the edge of the exposed electrode
(Enloe et al. 2004b; Corke et al. 2007). As described above, there are periodic blowing
and suction effects for a AC-DBD plasma actuator, but these predominantly occur during
the same phase of the input waveform, do not emanate from a common location (i.e. slot)
and are far from equal in magnitude. This behavior causes AC-DBD plasma to perform
more similarly to a pulsed blowing device, while still retaining ZNMF properties.
The use of AC-DBD plasma actuators for airfoil separation control at locations
other than the LE has been limited. Studies have reported that actuators placed near the
TE of airfoils can produce effects similar to simple flaps with deflections of a few
degrees (Vorobiev et al. 2008). This results in a uniform increase in CL across all α and a
slight reduction in minimum drag coefficient, CD, at Re on the order of 105 corresponding
to velocities of a few tens of meters per second (He et al. 2009).
While the potential of AC-DBD plasma actuation for separation control is
apparent, they also possess some drawbacks. They have primarily been limited to
relatively low speed (U∞ < 30 m/s) low Re (~105) applications such as those associated
with micro air vehicles due to the weak flows induced by the plasma discharge
23
(Greenblatt et al. 2008). More recently, claims of control authority for freestream
velocities as high as 60 m/sec with Re=106 have been presented in the literature (Patel et
al. 2008). To date, DBD plasma actuators have not produced sufficient momentum to
eliminate separation for flows over simple deflected flaps at Re > 105 unless the
freestream velocity is quite low (Mabe et al. 2009), although novel arrangements of these
devices show promise (Poggie et al. 2010). Recent publications suggesting momentum
production can be increased by an order of magnitude using thicker dielectric materials
with lower dielectric constants and higher AC voltages imply that the technology is
continuing to mature and may allow use in transport aircraft applications (Thomas et al.
2009). These devices also present the possibility of placing multiple actuators in various
locations and orientations on the airfoil to allow spatially distributed forcing over 180
degrees referenced to the model surface which can produce streamwise vorticity and
further freestream momentum entrainment (Porter et al. 2009).
NS-DBD plasma actuator testing in flow control applications has been limited to
one isolated publication (Roupassov et al. 2009). This work demonstrated the efficacy of
NS-DBD plasma actuators for controlling flow separation on an airfoil LE up to Mach
0.74. This flow speed is substantially greater than any associated with AC-DBD control
authority in existing literature for airfoils. Unfortunately, attempts at replicating this
success have been unsuccessful. As such, it remains an exciting, but not well-documented
actuation technique.
24
Chapter 3: Experimental Facilities and Measurement Techniques
3.1 Wind Tunnel
A Gottingen-type, closed, recirculating wind tunnel (SN: 55706) manufactured by
Engineering Laboratory Design, Inc. (ELD) is used as the test bed for this work. Viewed
from above, the standard version of this tunnel creates counterclockwise flow. At the
request of the GDTL, the tunnel was modified by the manufacturer to produce clockwise
flow. This places the large cross section elements upstream of the test section in a more
convenient location maximizing practical laboratory work space. The various wind tunnel
components were shipped in a dedicated trailer and assembled on site. Overall
dimensions of the assembled tunnel are 9.8 x 2.2 x 4.1 m3 (32.2 x 7.2 x 13.5 ft3) with a
test section centerline height of 1.4 m (4.6 ft).
Air flow in the tunnel is created by an axial fan powered with a 200 hp variable
speed AC induction motor (Figure 3.1). The fan speed is set using an operator keypad on
the tunnel outfield just downstream of the test section (Figure 3.2). The contraction
before the test section has a symmetrical cross section and an area ratio of 6.25:1 (Figure
3.3). The primary diffuser downstream of the test section expands with a total angle of
6°. This assembly allows continuously variable air velocity in the tunnel from 3-90 m/s
25
(10-300 ft/s). Flow conditioning upstream of the test section includes a hexagonal cell
aluminum honeycomb (Figure 3.4). High porosity stainless steel screens are mounted
downstream of the test section as a safety catch (Figure 3.5). Four high efficiency turning
cascades fabricated of galvanized steel are installed in each of the four tunnel elbows
(Figure 3.6). This assembly results in freestream turbulence specifications on the order of
0.25% with +/-1% variation in mean freestream velocity across the tunnel span measured
15 cm (6 in) from the test section inlet. These measurements were performed by factory
personnel at ELD (ELD 2006).
Figure 3.1: Axial fan used in the subsonic recirculating wind tunnel.
26
Figure 3.2: Operator keypad for the subsonic recirculating wind tunnel.
Figure 3.3: Subsonic recirculating wind tunnel contraction section.
27
Figure 3.4: Flow conditioning screens at the subsonic recirculating wind tunnel contraction entrance.
Figure 3.5: Safety catch screens downstream of the subsonic recirculating wind tunnel test section.
28
Figure 3.6: Turning cascades used in the elbows of the subsonic recirculating wind tunnel.
The tunnel is also equipped with a commercial aluminum fin/copper tube, double
row heat exchanger with electronic modulating valve and set point controller located on
the wind tunnel operator keypad. The heat exchanger is positioned upstream of last elbow
before the nozzle contraction (Figure 3.7). This arrangement allows the tunnel freestream
operating temperature to be maintained at +/- 1 °C from the ambient when supplied a
sufficient source of cooling water (max 189 lpm (50 gpm)). Cooling water is taken from
the lines normally supplying the compressors at the Aeronautical and Astronautical
Research Laboratory (AARL) at OSU by opening a valve on the east wall of room 129. A
significant amount of slag tends to build in these lines necessitating a gravity and course
mesh filter upstream of the heat exchanger to prevent fin clogging. No detrimental effects
29
on cooling either the wind tunnel or the compressors have been observed when operating
both simultaneously.
Figure 3.7: Gravity filter, heat exchanger, modulating valve, flow meter and drain line on the subsonic recirculating wind tunnel.
During the course of the plasma actuator experiments, the modulating valve
associated with cooling the facility was found to be highly sensitive to electromagnetic
interference (EMI). The effect of EMI is to fully open the valve which can drop the air
temperature in the tunnel below the desired set point. To prevent a change in tunnel
30
temperature during plasma tests, the modulating valve is always forced to 100% by
setting the temperature to an unrealistically low value (i.e. 50 F). The temperature of the
tunnel is measured by a thermocouple placed just before the first elbow downstream of
the test section.
The 61 x 61 x 122 cm3 (2 x 2 x 4 ft3) facility test section is constructed of
optically accessible super abrasion resistant acrylic walls that are 25.4 mm (1 in) thick.
The infield side wall of the standard test section is fitted with a blank 30.5 cm (12 in)
diameter port that is located 30.5 cm (12 in) from the test section floor and 61 cm (24 in)
downstream of the test section entrance. The infield plug is equipped with a protractor for
setting model angles of rotation and attack using manual thumb screws. The standard
outfield wall has no such port. To facilitate use with 2D full-span airfoil models, the
outfield wall panel has been replaced with a mirror image of the infield wall complete
with an additional wall plug port shown in Figure 3.8.
31
Figure 3.8: Test section arrangement with wall plugs on both infield and outfield sides.
A two axis traversing assembly (Velmex, Inc. Unislide) driven by a DC stepper
motor and controller is mounted on top of the test section allowing introduction of
instrumentation such as pitot-static and hot wire probes through a 109 cm (42.8 in) long
high density nylon brush seal along the tunnel ceiling centerline. The use of such
instrumentation requires an aluminum support to prevent deflection of the probes at high
tunnel speeds. It is also necessary to modify the downstream side of the support with
some device to prevent tuning of the probe assembly to the freestream velocity. Without
this addition, the probe can vibrate significantly similar to a high amplitude tuning fork.
The final arrangement for the 91.44 cm (36 inch) pitot-static probe is shown in Figure
3.9. In this case, the probe is mounted on the front of the support and a small piece of
32
round stock is used on the downstream side. Both are secured to the support using
aluminum tape.
Figure 3.9: Modified pitot-static probe assembly.
Two piezometer rings consisting of 4 static pressure taps located on the centerline
of each panel are installed 50.8 mm (2 in) from both the contraction entrance and exit.
Pressures at these locations are used to set the tunnel operating conditions. Vertical
profiles of the dynamic pressure, q, measured along the tunnel centerline have been
acquired using a pitot-static probe at streamwise locations of 0, 30.5, 61 and 85 cm (0,
12, 24 and 33.5 in) relative to the test section entrance for velocities in the range 10-90
33
m/s with a 10 m/s increment (Figure 3.10). Maximum boundary layer thickness on the
tunnel floor is less than 20 mm (0.8 in) for all velocities and locations surveyed (Figure
3.11). The profiles have been offset for clarity by a dimensionless dynamic pressure,
q/q∞, of 0.5 for each 12 inches of streamwise location. Average freestream measurements
of q across the tunnel length surveyed compared to outputs from the piezometric rings
upstream and downstream of the contraction show that the tunnel has a calibration
constant, k, of 1.05 calculated from
o
qkp p
∞=−
3.1 where q∞ is the dynamic pressure measured with the pitot-static probe and po and p are
the static pressures measured at the contraction entrance and exit respectively (Rae and
Pope 1984). This factor has been employed in plotting the profiles in Figure 3.10 and
Figure 3.11 such that the q/q∞ is essentially unity in the freestream. Note that the profile
furthest downstream (x=33.5 cm) is shifted less than other cases to represent the actual
streamwise development of the boundary layer. While tunnel specifications state low
speed velocities such as 3 m/sec are obtainable, it is not advisable to acquire data below
10 m/sec due to significant flow unsteadiness. It is also clear that the boundary layer is
transitional on the tunnel floor for freestream velocities of 30 m/sec and below, however
this does not affect the freestream velocity profile.
34
Figure 3.10: Dimensionless dynamic pressure profiles measured 0, 30.5, 61 and 85 cm (0, 12, 24 and 33.5 in) (left to right) from the test section inlet.
Figure 3.11: Near-wall dimensionless dynamic pressure profiles measured at 0, 30.5, 61 and 85 cm (0, 12, 24 and 33.5 in) (left to right) from the test section inlet.
35
Components of the wind tunnel not in use during airfoil and traverse experiments
include a blank acrylic top panel, blank acrylic outfield wall panel and blank wall plug.
Two additional minor modifications to the tunnel include threaded ports in the test
section floor for mounting various assemblies and a port upstream of the tunnel
contraction for introducing particle seed used in optical diagnostics. Additional details on
the wind tunnel including electrical information, installation, operation and maintenance
instructions can be found in the ELD Inc. supplied manual (ELD 2006).
A modular instrument panel near the tunnel operation station houses additional
hardware (Figure 3.12). The panel includes two sets of differential static pressure
transducers (Omega Engineering, Inc. PX655-25DI and PX655-5DI) used for setting the
tunnel test conditions and three process meters (Omega Engineering, Inc. DP-25-E-A) for
displaying the output. The Velmex controller for the 2D traverse is also housed on the
instrument panel and can be controlled via computer or manually. The anemometer for
use in hot wire experiments (TSI 1750) is mounted on the instrument panel as well as
Scanivalve pressure sensors (DSA-3217/16px 5 psid), quick connectors for use with 1.6
mm (0.063) inch inner diameter polyurethane static pressure tubing and an five port
Netgear ethernet switch used for routing information to a dedicated wind tunnel
computer. Standard manometers serve as a means of verifying the tunnel operating
conditions.
36
Figure 3.12: Modular instrument panel for the subsonic recirculating wind tunnel.
3.2 Airfoil Model
A simplified high-lift version of the NASA Energy Efficient Transport (EET)
airfoil has been chosen as the test model based on recommendations from the initial
sponsor (AFRL). The 2D EET airfoil was thoroughly examined (Lin and Dominik 1997),
but more recently significant studies on active separation control with synthetic jets have
been completed for a similar simplified version (Pack et al. 2002; Melton et al. 2003;
Melton et al. 2004; Melton et al. 2005; Melton et al. 2006; Melton et al. 2007). The OSU
version has a chord of 25.4 cm (10 in) and fully spans the 61 cm (24 in) test section in a
37
horizontal configuration. The model is equipped with a deflectable (0-60°) TE flap that is
25% of the airfoil chord, but for simplicity lacks the leading edge droop used by NASA.
It is constructed of a nylon compound (Duraform GF) and has been fabricated using
selective laser sintering (SLS) technology.
The model is assembled in three elements consisting of the suction surface,
pressure surface and flap. The model elements are fabricated by General Pattern, Inc. and
assembled and instrumented by ELD according to specifications provided by OSU. There
are 45 staggered static pressure taps located near the test section centerline and 15 static
pressure taps at ¼ and ¾ spans used for checking flow three dimensionality. The urethane
tubing described above connects the pressure taps from the infield side of the tunnel to
the static pressure sensors mounted on the instrument panel. Care is taken to ensure no
kinks or leaks exist in this tubing before testing.
The model is also instrumented with 7 high bandwidth 2.5 mm (~0.1 in) diameter
Kulite pressure transducers (XCQ-080-25A and LQ-062-25A) flush mounted near the
centerline. Figure 3.13 shows the airfoil profile and the location of static pressure taps
and transducers near the centerline. There were significant issues during the pressure
transducer installation as many were broken by ELD, Inc. due to tight space constraints in
the model. This was especially difficult for the miniature sensors (LQ-062-25A) in the
TE flap. Details on model construction and disassembly can be found in the ELD
supplied manual (ELD 2007).
38
Figure 3.13: 2D profile of the airfoil in cruise configuration showing the approximate location of static pressure taps and high bandwidth pressure transducers near the airfoil
centerline.
Independent settings for the incidence and flap deflection angles are set manually
using separate wall plugs with protractors on the tunnel infield side. The zero angle of
attack position is scribed on the infield wall plug that is attached to airfoil model. This
scribe aligned with 20 degrees on the uncalibrated protractor sets the zero angle of attack
position. The flap protractor is calibrated separately and documented on the flap wall
plug. A digital photograph showing the airfoil with trailing edge flap deflected and flap
wall plug is shown in Figure 3.14.
39
Figure 3.14: OSU version of the simplified high-lift EET airfoil with TE flap deflected.
The model has been extensively tested at multiple incidence and flap angles up to
Re=750k and has been run less extensively at Re=1000k at α as high at 18o. Attempts to
further increase Re at high α are not advisable based on visual observations of model
bending and flutter (Figure 3.15). Operation at higher Re and low α may be possible, but
this has not been attempted. The model is not designed for use in dynamic α studies.
40
Figure 3.15: Model bending produced by loading at high Re and α with DBD plasma at the LE.
3.3 Plasma Actuator Hardware
Input signals for the AC DBD plasma actuators are generated using a dSpace DSP
1103 board and hardware system. The dSpace system was originally purchased for use in
feedback control, but serves only as a function generator for this work. Outputs of the
dSpace D/A board include both the full modulated excitation signal supplied to the high
voltage power supply as well as the modulation signal alone for use in phase-locked data
acquisition timing.
Signals generated by dSpace are used as inputs to a Powertron Model 1500S AC
power supply (1500 W) and step-up high voltage transformer. Two identical power
supply models have been used, but the majority of single actuator AC DBD plasma tests
have been performed using amplifier 1 as labeled. Amplified signals from the power
41
supply are sent to a low power (200 W) high voltage (0-20 kVrms) transformer designed
to operate in the frequency range of 1-5 kHz. Ballast resistors were originally employed
in the plasma circuit, but were found to be unnecessary and subsequently removed. The
output of the transformer is connected directly to the high voltage electrode. No
impedance matching or EMI hardening circuit elements are employed. The high voltage
15 kVDC rated lead wires must be insulated from the laboratory floor and ground wire as
a weak discharge can ensue even through the insulation. This is accomplished using
sheets of cardboard. A safety circuit consisting of various fuses was investigated to
protect the airfoil from damage in the event of arc formation, but close monitoring of the
signal during operation using a Tektronix P6015A high voltage probe and a Tektronix
TDS 220 oscilloscope was found to be sufficient. The AC power supply and hardware is
located behind the wind tunnel operator station to facilitate one person operation.
High voltage nanosecond pulses are generated using an in-house constructed
pulser. The pulser is a magnetic compression type currently capable of repetition rates up
to 10 kHz. Peak voltage developed on an open load can reach ~18 kV with maximum
pulse energy of ~100 mJ. Pulses can be produced with either positive or negative
polarity, but only the positive pulse is examined in this work. The device is relatively
inexpensive in comparison to commercial pulsers of similar specification and can be
repaired using mostly off-the-shelf components. Dimensions of the pulser are
approximately 40 x 40 x 40 cm3. A 650 V DC Power Supply (Sorenson DCR 600-4.5B)
is used to power the system and a Tektronix AF6310 function generator provides input
signals. The pulser is located in the wind tunnel infield during operation due to the
42
requirements that lead wires be of equal length and as short as possible to minimize pulse
propagation effects and the possibility of reduced peak voltages (Opaits et al. 2009a).
This requires two operators during testing.
AC voltage measurements are acquired and monitored at the secondary side of the
high voltage transformer with a Tektronix P6015A high voltage probe. A LeCroy current
probe (CP031) is also employed. The power dissipated by the AC-DBD is calculated with
the Lissajous figure using charge-voltage measurements (Falkenstein and Coogan 1997).
A 47 nF capacitor is connected in series with the covered ground electrode in this case.
The voltage across the capacitor is measured using a Tektronix P6111B voltage probe.
The corresponding signals are monitored on a LeCroy Waverunner 6050A oscilloscope,
but the actual power calculation is performed offline. The AC-DBD plasma actuator is
operated using a 1-3 kHz sinusoidal carrier frequency with voltage levels up to 30 kVpp
and various modulation waveforms and frequencies. Electrical measurements for the NS-
DBD are acquired at the high voltage electrode in order to most accurately resolve the
actual voltage developed on the load. EMI is relatively minor when all external
equipment is properly grounded. The only devices that show considerable EMI influence
are the modulating valve for the heat exchanger and the differential pressure sensors used
for monitoring the tunnel air speed. Neither of these is essential once test conditions are
established.
43
3.4 Diagnostics
3.4.1 Static Pressure
Measurements of static pressure from ~1 mm (0.040 inch) inner diameter taps on
the airfoil model surface are acquired using three Scanivalve digital pressure sensor
arrays. Values of dimensionless pressure (CP) are averaged over 50 samples acquired at
10 Hz near the model centerline and used to calculate sectional lift coefficient (CL) using
numerical integration according to:
Pp pC
q∞
∞
−=
3.2
1
, ,0
L P pressure P suctionside side
xC C C dc
≈ − ∫
3.3
where p is the static pressure on the surface, p∞ is the freestream static pressure, q∞ is the
freestream dynamic pressure and x/c is the chordwise position.
3.4.2 Fluctuating pressure
Kulite pressure transducers installed in the model are powered using an in-house
constructed signal conditioner that amplifies each sensor output by 1000. The signal
conditioner has a variable low pass filter, but noise levels are found to be substantially
reduced when a separate Kemo filtering unit is employed (Benchmaster 21M). The Kemo
unit has variable low/high pass filter settings, but the majority of tests have been
performed using a low pass filters at 10 kHz. Attempts to filter out the AC plasma carrier
frequency during low frequency modulating signal testing were met with limited success.
44
The resulting pressure traces are sampled simultaneously at 50 kHz using a National
Instruments PCI-6143 data acquisition board. Average spectra are calculated from 32
blocks of 8192 samples which results in a frequency resolution of approximately 6 Hz. A
Hanning window function is applied to each block.
The power spectral density (PSD) for each block of dimensionless fluctuating
surface pressure, cP, is calculated using
( ) ( )P j P j Pc t C t C= − 3.4
( ) ( ) 2 ( 1)( 1),
1
, ,
ˆ ( ) ,
k [1, (N+1)/2] and ( 1) /
sp
sp sp
Ni j k
P x k P jj
x k x s
c F w j c t e
F k F N
π+ − − −
=
+ +
=
∈ = −
∑ 3.5
( ) ( ) ( ) ( )[ ]
, , , ,
2
1
ˆ ˆ2 ,
1 for 1, / 2 with ( ( ))
sp sp sp spx k P x k P x k x s
N
j
PSD F c F c F CNF
k N C w jN
+ ∗ + + +
=
=
∈ = ∑3.6
where, PC is the time-averaged surface pressure, w(j) is the Hanning window function, tj
is the time index of the jth sample, Fx,s + is the dimensionless sampling frequency, Fx,k
+ is
the measured frequency and N is the number of samples acquired. The dimensionless
units of PSD are then expressed as cP2/
spxF + in the corresponding figures where xsp is the
length scale used for expressing the dimensionless frequency which in this work is the
45
chord, c, or flap length, L. The corresponding blocks are then averaged to obtain the
properly scaled result.
Pressure transducer measurements near plasma actuators are quite challenging
due to EMI and arc formation. During the course of this work, transducers near the TE
flap shoulder and LE have been damaged due to direct arcing from the high voltage
electrode for both AC and NS-DBD plasma. The resulting current spike can also
propagate through the transducer cabling causing damage to the individual channels of
the signal conditioner. At this time, only three transducers (x/c=0.40, 0.95 and 0.40
(pressure side)) remain operational. Less catastrophic effects have also been observed in
that noise levels for individual sensors increase with EMI exposure time such that after
months of use the transducer signal becomes excessively noisy. It is also suspected that
transducer installation during which shielding wire was pulled away from the sensor
element has caused a less robust shield effect and resulted in higher noise levels in some
sensors. Strides must continue to be made to improve pressure and velocity diagnostics
near plasma actuation.
Since significant uncertainty exists around measurements of pressure spectra in
the presence of plasma, a study was performed in an attempt to distinguish real effects
from electrical noise. Figure 3.16 shows the experimental setup in which an actuator is
placed upstream of the TE flap shoulder and an additional pressure transducer is taped to
the test section wall outfield near the high voltage cable. Figure 3.17 shows the resulting
pressure spectra calculated using this arrangement for transducer 5 (x/c=0.90). In this
case, properly magnitude scaling has not been employed as only the relative effects are of
46
interest. The baseline flow shows an oscillation near 90 Hz with a high frequency noise
floor near -15 dB beginning around 1 kHz. Actuation at 2 kHz with sinusoidal
modulation of 90 Hz increases the magnitude across all frequencies and shows a very
strong peak at the carrier frequency (2 kHz). An additional large amplitude peak is also
visible at the modulation frequency (90 Hz). With this data only, it is unclear if the 90 Hz
peak and the broadband increase are actual flow changes or merely artifacts of electrical
noise. The spectrum of the transducer outside the wind tunnel gives a representation of
EMI propagation through the air from the HV line. The 2 kHz carrier waveform is quite
apparent, but there is no peak associated with the 90 Hz modulation signal. The
broadband signal also matches the controlled spectra above ~500 Hz suggesting this is
not a real flow artifact. Finally, the plasma actuator is operated in still air (tunnel off)
using the same excitation inputs as the control case. The spectrum from transducer 5
(x/c=0.90) is also shown in the figure. It is obvious that as in the outside sensor case, the
2 kHz carrier frequency is quite strong and the broadband levels above ~200 Hz are likely
an EMI artifact only. However, there is no evidence of a strong peak at the modulation
frequency and the controlled flow is found to have a broadband level above the tunnel off
case below ~200 Hz. From this information, it can then be seen that flow behavior cannot
be distinguished from electrical noise above ~200 Hz in this case. However, both the
coherent and incoherent dynamics below 200 Hz appear quite physical. Note that these
frequency values are not constant across the parameter space surveyed and similar
analysis should be performed for each forcing case. The use of pressure spectra is only
employed briefly in Section 6.1 for a case in which this verification has been used.
47
Figure 3.16: Setup used for distinguishing flow phenomena from EMI.
Figure 3.17: Example of pressure spectra used for distinguishing flow phenomena from EMI.
48
One additional comment on the analysis of pressure traces near plasma is
warranted. The broadband increases shown in Figure 3.17 are partially attributed to an
impulse like spike in the pressure trace which, in the frequency domain, is composed of
all frequencies. Consequently, amplification across the entire frequency range is visible.
A sample of pressure data for this behavior is shown in Figure 3.18. Since only a finite
number of these impulses appear in the pressure trace, blocks of data which contain these
events can simply be neglected in the analysis since they represent unphysical flow
behavior. Note again that this is only one example of this behavior and that other cases
can contain more or less of these spikes. The source of this behavior is not known.
Figure 3.18: Example of randomly occurring voltage spikes in pressure traces due to EMI.
49
3.4.3 PIV
Two-component particle image velocimetry (PIV) is used to obtain quantitative
velocity measurements for the various flow fields discussed. Nominally submicron olive
oil seed particles are introduced upstream of the test section contraction using a 6-jet
atomizer (TSI 9306A). The beam from a dual-head Spectra Physics PIV-400 Nd:YAG is
directed over a distance of ~6 m (~20 ft) then focused using a 1 m focal length spherical
lens and two -25 mm focal length cylindrical lenses. The additional cylindrical lens is
used to increase the spreading rate of the beam in an effort to provide a uniform light
distribution over the entire field of view which for the airfoil cases is approximately 300
x 200 mm2 (12 x 8 in2). Only ~50% of the maximum laser energy is necessary for ample
signal to noise ration in this flow. The beam forming hardware is mounted on top of the
wind tunnel and can be located in two possible positions depending on the airfoil case
tested (LE or TE) by removing the various 1/4 -20 screws connecting it to the tunnel
ceiling. In the LE case, the optical assembly is positioned as far upstream as possible
while the TE case requires the assembly in the more downstream position so that the
wake can be visualized. The spanwise location of the beam can be adjusted using an
optical rail and slider while the streamwise position can be fine-tuned using thumbscrews
on the final turning optic. No serious vibrations have been encountered for the beam
forming optics on the tunnel ceiling. However, the turning optics on the structural support
near the shear layer/SWBLI facility have been found to vibrate during operation of the
AARL transonic wind tunnel. Beam alignment is ensured by first using burn paper to
match the tails of the sheets and the fine tuned by visual inspection of the particle
50
movement between two frames of the raw data. The experimental setup for airfoil LE
PIV measurements is shown in Figure 3.19. The two axis traverse assembly is also visible
in this image.
Figure 3.19: PIV experimental setup for LE airfoil measurements.
Significant laser reflections exist in the test section primarily due to reflections
from the airfoil model surface. These contaminate not only data near the airfoil surface,
but also reflect from the test section walls and contaminate other regions of the flow field.
An example of this behavior is shown in Figure 3.20. These effects are minimized by
51
adhering blue masking tape to the infield interior wall of the wind tunnel. The reflections
are further reduced by covering the tape with flat black paint (see Figure 3.15).
Reflections from the model surface can be reduced using fluorescent paint created by
solving rhodamine powder in ethanol and mixing with water soluble acrylic paint.
Rhodamine absorbs the green wavelengths (532 nm) and fluoresces red (~600 nm), but
its efficiency is only ~30%. By using a narrow band pass optical filter (CVI F10-532.0-4-
2.00) over the camera lens, a portion of the original green intensity is eliminated.
However, when the beam is directly incident on the surface, the paint loses effectiveness
due to burnoff after ~100 high power laser shots. Thus, the rhodamine paint is more
effective when the laser sheet is not directly incident on the surface. Note that thin pieces
of Kapton tape perform similarly by diffusing the laser reflections, but these can only be
used in large separated regions such that the boundary layer is not disturbed by their
addition to the surface. Additional localized regions of reflection such as white high
voltage wires can minimized using a simple black marker. By using rhodamine, Kapton
tape, paint, marker and finely tuning the beam location and intensity, these reflections are
minimized and high quality raw data is produced. Sample raw data from both the LE and
TE experiments is shown in Figure 3.21.
52
Figure 3.20: Example of laser reflections from the model surface and their effect on PIV data.
Figure 3.21: Sample raw PIV data images for LE (a) and TE (b) experiments.
a)
b)
53
Images are acquired and processed using a LaVision PIV system operating
software version DaVis 7.2. Two images corresponding to the pulses from each laser
head are acquired by a LaVision 14 bit 2048 by 2048 pixel Imager Pro-X CCD camera
equipped with a Nikon Nikkor 50 mm f/1.2 lens. The camera is mounted on an optical
table on the wind tunnel outfield using an optical rail and support. PIV measurements
over the chord of the airfoil in the LE case are acquired by viewing the light sheet
orthogonally through the main wall plug. Measurements over the TE flap and in the
airfoil wake must be acquired from a downstream angle of approximately 14° due to the
wall plug arrangement. An image correction algorithm provided in the commercial
software is applied to the data in this case. The boundary layer and flat plate PIV data are
acquired using an additional Vivitar 2x teleconvertor to increase spatial resolution at the
expense of field of view. In that latter, the flat plate substrate is mounted vertically in the
test section to reduce reflections from the surface (Figure 3.22). Finely machined
Lavision calibration plates (Type 31 and 10) are used for the airfoil and flat plate/BL
cases respectively.
54
Figure 3.22: Experimental setup for flat plate PIV measurements.
A narrow band pass optical filter (CVI F10-532.0-4-2.00) is often used when
acquiring PIV data. The effect of this component is to eliminate contamination from
ambient lighting as well as rhodamine painted surfaces. The optical filter has an
additional effect of slightly blurring the particle size on the CCD. In airfoil experiments,
this effect can be beneficial for increasing the particle size such that peak locking effects
that can occur for small particle sizes (< 1 pixel) on the CCD are minimized (Stanislas et
al. 2008). In cases where the spatial resolution is increased (boundary layer and flat
plate), the particle size can become too large due to the addition of the optical filter. For
these tests, no optical filter is used necessitating the lights in the laboratory must be
55
substantially dimmed. Additional ambient light is eliminated by draping a black felt cloth
over the infield side of the wind tunnel. For measurements of the actuator in still air, the
flow is seeded and the tunnel is energized to uniformly distribute the particles. Once
adequate seeding is ensured, the tunnel is shut off and particles remain suspended
Regions of interest are adjusted depending on the flow field in an effort to reduce
file size and processing time while maximizing sampling rate (max 10 Hz). Note that
cropping the raw image in x does not improve the sampling rate for this hardware system.
Also, the camera cropping in x is forced to be symmetric while cropping y is more
flexible. Regions outside the area of interest are excluded from the analysis using a
manually drawn data mask. For each image pair, subregions are cross-correlated using
decreasing window size (642-322 pixel2) multi-pass processing with 50% overlap in high
accuracy Whittaker reconstruction mode (Lavision 2007). The time separation between
laser pulses used for particle scattering is tuned according to the flow velocity, camera
magnification and correlation window size using the known freestream velocity and the
criteria which requires particle movement to be less than ¼ of the correlation window
size. The particle size on the CCD is in the 1-4 pixel range and the seeding density is
tuned to ensure no less than 6 particles are present in a given correlation window in all
cases. The resulting velocity fields are post-processed to remove any remaining spurious
vectors using a correlation peak ratio criteria, allowable vector range and median filter
that uses contributions from the eight closest neighboring vectors. Removed vectors are
replaced using an eight vector standard interpolation scheme and a 3x3 Gaussian
smoothing filter is also applied to the calculated velocity fields. The PIV data for used in
56
creating ensemble averages are sampled at 10 Hz. Statistics for the airfoil, boundary layer
and actuator characterization are calculated from 1000, 500 and 500 instantaneous
velocity fields respectively. Details on the Lavision algorithms are documented in the
corresponding manual (Lavision 2007).
Conditional sampling of PIV data (phase-locking) is accomplished using the
programmable timing unit of the LaVision system. In this case, the acquisition is synced
with the modulation frequency of the actuation signal. The baseline pressure signal is not
sufficiently periodic to allow this acquisition in the airfoil case. Velocity fields at various
phases of the actuator modulation frequency are investigated by stepping through the
actuator period using time delays. Phase locked data sets are averaged over at least 50
images for each phase which is found to be sufficient for resolving the primary flow
features (velocity and vorticity). Conditionally sampled (phase-locked) PIV data is
acquired at 5 Hz. The spatial resolutions of PIV data for the flat plate, boundary layer and
airfoil data sets are approximately 0.2, 0.4 and 2.4 mm, respectively which corresponds to
magnification values of ~44 and ~7 pixels/mm. Near surface measurements for the flat
plate and boundary layer data sets are obtained within 0.3 and 0.4 mm of the substrate
respectively.
The timing between the laser pulse and actuator signal must be checked for proper
synchronization. In the phase-locked case, this is obvious since the actuator and laser
signal must be have a constant delay at a given phase to ensure phase averages are
accurate. The opposite effect must occur for a true ensemble average in that the PIV
acquisition should sample over many phases of the actuation signal such that time-
57
averaged results are not biased toward any phase. Under certain circumstances the PIV
system and function generator running the actuator can run on similar clocks. What
appears to be a random acquisition can actually be inadvertently phase-locked. This
effect is known to occur when using Lavision PTU version 9 and the Tektronix AF6310
function generator used for NS-DBD plasma actuation. A simple solution to this problem
is to slightly increase or decrease the PIV acquisition rate to some non integer multiple of
the actuation signal (9.99 or 10.01 Hz). This effect does not occur for actuation used
dSpace for input signals, but for good practice one should employ this non-integer
acquisition rate in all cases.
3.4.4 Accuracy
Full scale accuracy for measurements of instantaneous velocity and the
uncertainty on mean velocity calculations are listed in Table 3.1. The former is calculated
by assuming negligible laser timing errors and a correlation peak estimation error of 0.1
pixels. This is the generally accepted error estimation for modern PIV systems (Stanislas
et al. 2008). The smaller error in the boundary layer and airfoil cases results from better
choices of the time separation between laser pulses due to the a priori knowledge of
maximum velocities in the freestream flow. The uncertainties in measurements of mean
velocity are presented using 95% confidence intervals, CI:
2
1.96CI xNσ
= ±
3.7
where x is the sample mean, σ, is the sample standard deviation and N is the sample size
(Benedict and Gould 1996).
58
The values in Table 3.1 are in terms of relative error. The actuator and boundary
layer cases are estimated using the average standard deviation and average velocity in the
near wall profiles which gives a more representative estimate than the maximum or
freestream velocity. The uncertainty in the airfoil cases is estimated using the maximum
standard deviation in the turbulent wake and the freestream velocity. A range of values
that encompasses all the flow fields surveyed is listed in the interest of brevity. The
actuator and airfoil measurements have higher values of uncertainty due to the unsteady
nature these flow fields. The V component of velocity has not been considered for the
actuator and boundary layer measurements since it is not relevant to calculations of
integral parameters.
Mean Instantaneous
U V U, V
Actuator Characterization 4-6% - 1.7%
Boundary Layer 0.5-0.7% - 0.9%
Airfoil 2-3% 2-3% 0.9%
Table 3.1: PIV Uncertainty
59
3.5 Test Conditions
Experiments on the controlling separation from the TE flap have been conducted
for Re between 240k (15 m/s) and 750k (45 m/s) for δf of 20-40o at atmospheric
temperature and pressure. These test conditions correspond to a sample published results
of NASA work (Melton et al. 2004; Melton et al. 2006), but do not contain studies on the
effect of LE droop which limits our analysis to lower α. LE separation control studies
have been performed at primarily Re=750k (45 m/sec) and 1000k (62 m/sec) with no flap
deflection. For all airfoil measurements, data is acquired by establishing a separated flow
baseline condition then energizing the actuator. For repeated samples of different forcing
cases the baseline separated condition is reestablished between consecutive control cases
to eliminate confusion on the results due to hysteresis effects.
60
Chapter 4: Dielectric Barrier Discharge Plasma Actuators
4.1 AC DBD Plasma Actuator Design
As described in Section 2.2.1, AC-DBD plasma actuators which integrate
relatively thick dielectrics (on the order of a few mm) into the model geometry create the
greatest induced flows due to their ability to withstand higher voltage inputs without
entering the Corona or streamer mode (Corke et al. 2009). These types of designs are
preferred for second generation type studies where the actuator is cast into the model
design. For this initial work it is essential to maintain both the flexibility and modular
nature of the actuators due to the variable separation location and number of parameters
(actuator location, geometry, orientation and number) which are intended for
investigation. More importantly, we wish to be able to move these devices without
modifying the airfoil. For these reasons, DBD plasma actuators whose construction is
based on thin flexible adhesive materials (i.e. tapes) are chosen. Such devices are
inexpensive, readily available and easily removable which allows the actuator location
and orientation to be varied. The thin profile and ability to conform to surface curvature
are also appealing since this allows the application of the device to the surface with
minimal alteration of the basic features of the flow field.
61
It is widely believed that the electrode material is much less important than the
dielectric when inducing flows based on DBD plasma discharges (Hoskinson et al. 2008).
Accordingly, the most common electrode used in the literature, copper, is selected for this
study and no attempt is made to optimize this choice. It has a total thickness of 0.09 mm
(0.0035 in) and is bonded with an acrylic adhesive that is 0.05 mm (0.0021 in) thick. The
exposed and covered electrodes have widths of 6.35 mm (0.25 in) and 12.7 mm (0.50 in),
respectively. A ground electrode width of 12.7 mm (0.50 in) allows the use of standard
25.4 mm (1 in) wide dielectric tapes. The ground electrode width must be large enough
so the plasma formation is not limited in the streamwise extent since this has a strong
effect on the induced flow. For the input waveforms in this work, a covered electrode
width of 12.7 mm is sufficient.
Various studies recommend the use of a slight gap (1-5 mm (0.08-0.20 in))
between exposed and covered electrodes (Roth and Dai 2006; Forte et al. 2007). The
latter work showed a modest velocity increase (50 cm/sec) from the 0 to 5 mm (0.20 in)
gap case. Because of the difficulty repeating this exact gap size for multiple iterations,
some prefer the use of no gap or slight overlap between exposed and covered electrodes
(Corke et al. 2007). The zero gap recommendation has been employed in this work.
The self-imposed mandate that the actuator materials (both electrode and
dielectric) be composed of thin adhesive tapes limits the dielectric material selection to a
few basic choices. At the onset of this study, significant work was performed to
determine the best dielectric material and even combinations of various materials for
sustaining the glow discharge and inducing flow. At this time (2006), plasma actuator
62
studies were becoming very popular, but archived literature on optimizing the device
design was not readily available and various researchers had their own preference for
materials. Limiting these choices to readily available adhesives left primarily Kapton,
Teflon and/or ultra high molecular weight (UHMW) tape as possibilities. A significant
amount of time was spent to optimize the dielectric choice based on a balance of adhesive
properties, dielectric strength and induced flow capability.
Initial designs were composed of very thick layers of dielectric with an additional
layer of protective material under the ground electrode such as mica or thermally
conductive Kapton tape in order to prevent damage to the airfoil. With experience this
became less of a concern. The voltage trace could easily be monitored in real time and
when dielectric breakdown occurred, immediately shutting down the power supply was
sufficient to prevent damage to the airfoil substrate for normal operating parameters. For
more catastrophic events with significant arc formation, a small pit a few mm in diameter
is burned in the model similar to Figure 4.1. Even in this case, the damage can be
repaired by employing simple body filler. In the final design, the ground electrode is
directly applied to the model surface.
63
Figure 4.1: Airfoil damage due to arc formation.
The induced flow created by the device was only mildly affected by the dielectric
material. Rather the adhesive properties and ability to withstand arcing were deemed
most important in this application. Teflon and UHMW adhesive tapes were found
substantially inferior to Kapton in this regard when used in thin layers. This selection can
also be supported based on material properties. Referring to Table 4.1, which summarizes
various properties of dielectric materials, it can be seen that Kapton has medium to low
dielectric constant (ε~3.5) compared to other dielectrics surveyed, but its dielectric
strength is approximately an order of magnitude greater (154 kV/mm). Assuming that the
dielectric strength is an indicator of the ability for a dielectric to maintain operation in the
normal glow regime makes this a superior choice, especially for the purposes of this work
in which very thin adhesive materials are necessary. As mentioned earlier, the adhesive
and layering ability of Kapton tape has been found to be superior to others explored. This
layering decreases the bulk capacitance, but the increased dielectric strength allows
64
application of higher voltages without entering the arc regime. The Kapton tape in this
study has thickness of 0.09 mm (0.0035 in) and dielectric strength of 10 kV. Each tape
has a silicon adhesive layer that is 0.04 (0.0015 in) thick. The effect of the silicon
adhesive on the dielectric performance is not examined here. The total thicknesses of the
dielectric and the device as a whole are 0.44 mm (.0175 in) and 0.62 mm (.025 in),
respectively, unless otherwise noted. This actuator is stable for AC-DBD plasma studies
at frequencies of 1-3 kHz at voltages of ~20 kVpp. Voltages up to 30 kVpp have been
tested, but these higher potentials are substantially more prone to arc formation and
sparking to the model surface.
Table 4.1: Properties of various dielectric materials (Roth and Dai 2006).
65
The recommended carrier frequency for the Kapton dielectric in air is 5 kHz
(Corke et al. 2009). This has not been used due to transformer power saturation issues
and limited run time at this high frequency due to higher thermal loads on the dielectric.
Results suggest that this should have little effect on the maximum body force generated
due to the relatively broad peaks associated with thrust optimized excitation frequency
(Corke et al. 2009). A typical actuator like this lasts roughly one hour of non-continuous
actual run time. Note that in practical flight applications, more robust dielectrics that are
embedded in the substrate as well as close monitoring of the dielectric fidelity will be
required. Cursory investigations of induced flow created by negative going sawtooth
waveforms showed little difference compared to sinusoidal version of the same signal.
Consequently, sinusoidal carrier frequencies are used for simplicity.
Even though Kapton is preferred in this work, it is still found to degrade with run
time. Figure 4.2 shows the dielectric surface after a significant amount of run time (~1
hr). Note the white regions in the dielectric which appear almost like traces of plasma
filaments. This changes the dielectric properties and a quantitative monitoring system for
such material changes is required for the practical application of DBD plasma actuators.
Some researchers have suggested using combinations of Kapton and Teflon to increase
plasma induced flow and allow longer run time (Poggie et al. 2010). Visual inspection of
the latter seemed to show promise, but dielectric breakdown was not found to be
substantially delayed in practice nor was the induced flow radically changed. Thus, the
extra layers of Teflon are not employed in the final design. Note that rhodamine paint
(pink region) is also visible above and below the device.
66
Figure 4.2: Degradation of the dielectric surface after substantial plasma run-time.
The application of the actuator to the airfoil is carefully done by applying the
device in the selected location using pressure transducers or static pressure taps as datums
such that each layer of tape is adhered at the proper streamwise location. The electrodes
overlap for at least 3/4 span of the airfoil (~46 cm). To prevent arcing, the high voltage
cable is connected on the tunnel outfield side while the ground is connected on the tunnel
infield side. Care is taken to use sufficient insulation at the soldered joints of the leads
especially on the high voltage side. Structural joints and steel tabulations that form the
pressure taps in the airfoil must also be insulated using Kapton tape to prevent discharge
to these regions.
4.2 AC-DBD Plasma Actuator Characterization
Figure 4.3 shows simultaneously sampled traces of current and voltage for a
typical the AC-DBD plasma actuator test. As previously discussed, the majority of the
momentum is generated by the negative going portion of the negative voltage half-cycle
when the amplitude of microdischarges is lower and the discharge is more diffuse. The
large current spikes during the positive going half-cycle of the positive waveform
67
represent the more localized filamentary discharge. This portion of the cycle has been
experimentally determined to be less efficient for generating flow in air due to intense,
but more localized streamer discharges. As such, it functions more like a reset switch for
the momentum generating portion of the waveform.
Figure 4.3: Simultaneous current and voltage traces for a typical AC-DBD plasma actuator.
The power dissipated in the discharge is most simply determined by integrating
the product of voltage and current over a given cycle with data like that shown in Figure
4.3. While the calculation is quite simple, simultaneous measurements of the voltage and
current is difficult due to the radically different time scales involved. The voltage
68
waveform has a period on the order of a few hundred microseconds depending on the
carrier frequency, but the current spikes associated with microdischarges occur on time
scales of a few tens of nanoseconds (Falkenstein and Coogan 1997). Thus, one must
either sacrifice resolution of current spikes in order to get a statistically significant
number of voltage periods or limit the number of voltage periods acquired in an effort to
resolve all the microdischarges in the current trace.
A more accepted approach is to use charge-voltage methods (Falkenstein and
Coogan 1997). In this case, a capacitor is connected in series with the ground electrode
and charge on the capacitor is measured simultaneously with the high voltage applied to
the exposed electrode. Since the high voltage waveform determines the charge
distribution on the capacitor, these two signals are on similar time scales. A plot of these
two simultaneously sampled parameters forms a parallelogram as shown in Figure 4.4
where each high voltage cycle represents one revolution around the figure. The area
enclosed by this curve is the energy dissipated per high voltage cycle. The corresponding
power, P, can then be calculated by simply multiplying by the frequency of the applied
voltage:
accycle
P f V dQ= ∫ 4.1
where f is the AC frequency, Vac is the voltage applied to the HV electrode and Q is the
charge on the capacitor.
69
-10 0 10-4
-2
0
2
4[×103]
Voltage [kV]
Cha
rge
[nC
]
41.2W
Figure 4.4: Example of Q-V data for AC-DBD plasma.
Figure 4.5 gives an example of the typical power per unit length dissipated by the
actuator and calculated using Q-V measurements as a function of the applied voltage. In
this case, the discharge is generated by a 2 kHz carrier frequency and the dielectric is
composed of 5 layers of Kapton tape. The dissipated power is found to increase
proportionally with voltage to the power 3.51, consistent with existing literature (Corke et
al. 2010), where the constant of proportionality, α, is approximately 2x10-5. The data has
been plotted on a log-log scale to emphasize the validity of the fit at higher voltages
while highlighting considerable scatter of the data below 5 kVpp. This scatter occurs over
a voltage region where no discharge exists and the Vac7/2 power law is not valid. This
region is often referred to as being characterized by dielectric heating (Roth and Dai
70
2006). The electrical power per unit length, P/l, dissipated by the discharge is 0.74 W/cm
at 20 kVpp. In later experiments, the plasma carrier frequency is increased to 3 kHz and
the dielectric thickness is decreased to 3 layers (0.27 mm (0.01 in)). This changes the
power dissipation at 20 kVpp to ~1.7 W/cm. A plot of dissipated power as a function of
applied voltage for this updated arrangement would still follow the Vac7/2 power law, but
have a different constant of proportionality.
Figure 4.5: Dissipated power per unit length as a function of applied voltage.
71
The velocity induced by a single DBD plasma actuator in still air has been
characterized using PIV. The characterization is performed by mounting actuators on a
flat plate substrate made from the same material as the airfoil model (Duraform GF
resin). The generated plasma spans approximately 16 cm (6 in) of the flat plate substrate.
The flat plate is mounted parallel to the laser sheet (vertically) to minimize reflections
from the surface (Figure 3.22). The data is acquired near the mid-plane of the device.
Figure 4.6 shows a sample time-averaged total velocity field generated by the actuator
operating with a 2 kHz sinusoidal carrier frequency at a voltage of 20 kVpp with no
modulation. The actuator schematic is to scale in the streamwise direction, but expanded
vertically to allow visualization. For example, the dielectric is composed of 5 layers of
Kapton tape which has an actual thickness of ~0.5 mm (0.018 in). Figure 4.6 is presented
to show the global characteristics of DBD plasma induced flow in still air, but subsequent
tests using thinner dielectrics have been found to produce more induced flow at the same
input voltage and frequency. As documented in the literature, a quasi-steady suction
develops above the exposed electrode, but the flow field is dominated by the near wall jet
downstream of the actuator. The actuator in Figure 4.6 has been mounted near the edge of
the flat plate (white region) hence the jet spreads on both the positive and negative sides
downstream beginning near x~50 mm since there is no shear force from the surface in
this region. Time-resolved velocity and thrust measurements in the literature show that
the AC-DBD plasma actuator in air produces a dominant velocity pulse at the high
frequency driving signal corresponding to the negative going portion of the negative half
cycle as mentioned earlier (Enloe et al. 2008b). Because of the asymmetry of both the
72
electrode geometry and force production over the two waveform half-cycles, the AC-
DBD is more similar to a pulsed wall jet than a traditional ZNMF device in that a
nonzero mean flow is established.
Figure 4.6: AC-DBD time-averaged induced velocity magnitude, W
, in quiescent air
(m/sec).
The usefulness of AC-DBD plasma for energizing boundary layers is certainly
evident from the near wall jet effect, however its efficacy as a more general flow control
actuator is dependent on the ability to influence natural flow instabilities. The majority of
experiments for separation control on airfoils deal with flow instabilities that are
approximately an order of magnitude less than frequencies associated with AC-DBD
plasma generation (1-10 kHz). To create excitation of this nature, the high frequency
carrier waveform is modulated using a low frequency signal in the range associated with
separated flow dynamics. Two examples of these modulating waveforms are shown in
Figure 4.7. The voltage traces have been acquired at the secondary (HV) side of the
transformer. The carrier frequency of the waveforms is 2 kHz and the modulation
73
frequency is 100 Hz. In Figure 4.7a, the high frequency carrier has been modulated using
a sine wave envelope, termed amplitude modulation (AM). Modulation using a variable
duty cycle as shown in Figure 4.7b allows more control over the input power and actuator
on/off time. The duty cycle shown is 50%. Modulation using this waveform is hereafter
termed burst modulation (BM).
Figure 4.7: Modulation waveforms for AC-DBD plasma actuation.
The ability of these waveforms to produce velocity fluctuations in quiescent
conditions is examined using phase-averaged velocity and vorticity data for AM and BM
of 10%, 30%, 50%, 70% and 90%. The vorticity is calculated according to:
zV Ux y
∂ ∂Ω = −
∂ ∂ 4.2
with a 2nd order accurate central difference scheme. For brevity, only one phase of the
oscillation is shown in Figure 4.8, but animations of four phases of the modulation period
a) AM b) BM
74
confirm that the structures generated by plasma convect in still air. As in Figure 4.7, the
plasma is created using a carrier frequency of 2 kHz at 20 kVpp with modulation
frequency of 100 Hz. The velocity fields are averaged over 50 instantaneous phase-
locked samples which is sufficient for resolving the primary features of the pulsed flow.
Recall that AC-DBD plasma induced flows are dominated by the near wall jet (U
component). However, the vertical component of velocity (V) gives a better emphasis of
the pulsating nature of the device when excited using these modulated waveforms. Note
that the dominant feature shown is the pulsed suction initialized near the electrode
interface that is released over the covered electrode before creating a vortex train further
downstream as confirmed by vorticity data for the same phase. The sinusoidal
modulation (AM) displays a well organized vortex train commensurate with the
modulation frequency that is sustained for approximately 40 mm before dissipating. At
the lowest duty cycle (10%), the pulsed suction is quite weak and no vortex is visible
because not enough carrier cycles are produced to create significant plasma induced
flows. The 30% duty cycle case shows greater suction and some sustained vortex
behavior. As the duty cycle is further increased, pulsed suction near the electrode
interface is stronger and covers a much larger region while the generated vortices persist
further downstream. In the 70% duty cycle case, the suction near the electrode interface
also generates a secondary flow in the form of a stationary clockwise rotating vortex that
is visible just upstream of the exposed electrode. At 90% the pulsing nature of the
actuation is essentially lost with little primary vortex generation and no recognizable
structure content in the vorticity.
75
Figure 4.8: Phase-averaged AC-DBD plasma induced velocity fields, V , (left, m/sec) and vorticity, Ω , (right sec-1) in quiescent air for modulation using AM (top) and various BM
(2nd from top to bottom) waveforms.
While the behavior of actuators in quiescent flows is not necessarily indicative of
their behavior in a flow control environment, these results suggest that greater control
authority can be achieved by changing the modulation waveform. This will be examined
in Section 6.1. Other methods of creating low frequency perturbations include providing
separate excitation signals to each electrode (Post 2004) or more simply using a very low
carrier frequency that is not optimal for plasma generation. The majority of the remaining
76
results are for BM using 50% duty cycle which is a reasonable choice for generating
control authority across the test cases surveyed.
Despite the usefulness of the wall normal velocity component for observing the
pulsing actuator behavior, the U component of the velocity is the dominant momentum
source. The momentum introduced by AC-DBD plasma is quantified using the PIV data
in Figure 4.9. In this case, the input frequency and dielectric thickness have been
optimized for practical application, run-time and flow generation. The dielectric is
composed of 3 layers of Kapton tape with total thickness of 0.27 mm (0.0105 in) and the
AC driving frequency is now 3 kHz. The plasma induced flow for this optimized actuator
has been measured for a range of forcing cases that encompass relevant excitation
frequencies for airfoil TE separation control. Profiles of U and Urms created by the
actuator 20 mm downstream of the electrode interface are shown in Figure 4.9. The
profiles have been forced to obey a no-slip condition at the wall. Plasma generation at a
frequency of 3 kHz creates a wall jet with maximum mean velocity of ~3.5 m/sec
measured ~1 mm from the surface (Figure 4.9a). Operating the actuator in burst mode
using square wave modulation at 50% duty cycle lowers the maximum induced velocity
but creates profiles that become substantially fuller as the burst frequency is decreased.
The fluctuating behavior of burst mode actuation follows a similar trend in profile
fullness, but the maximum value of the fluctuations now decreases with increasing
frequency (Figure 4.9b). Note that this behavior is also dependent on the carrier
frequency and duty cycle. The profiles in Figure 4.9 have been used to calculate both
77
mean, J/ρ and oscillatory <J/ρ> components of momentum with a constant density, ρ,
assumption (Enloe et al. 2006) using:
2
0
J U dyρ
∞
= ∫ 4.3
2
0rms
J U dyρ
∞
< >= ∫ 4.4
where U and Urms are the time-averaged and rms velocity profiles. The x=20 mm
location is chosen since it is near the discharge, but just downstream of the region where
visible radiation from the discharge can contaminate the PIV data. It should be noted that
the downstream location at which the momentum should be calculated is not well defined
for AC-DBD plasma (Pons et al. 2005; Porter et al. 2007; Greenblatt et al. 2008;
Hoskinson et al. 2008). Thus, reported values should be taken as order of magnitude
approximation.
78
Figure 4.9: Mean (a) and rms (b) velocity profiles, U, at x=20 mm for AC-DBD plasma operating in quiescent air for various BM frequencies at 50% dc.
a)
b)
79
The momentum values are cast in terms of Cµ, <Cµ> and Cµ,tot for the various Re
surveyed in Section 6.1 to allow comparison to more common flow control actuators
using:
JCq cµ∞
= 4.5
JCq cµ∞
< >< >=
4.6
,totC C Cµ µ µ= + < > 4.7
where q∞ is the freestream dynamic pressure and c is the airfoil chord. It should be noted
that each δf has an optimal control frequency of burst mode excitation ranging from 70-
400 Hz and these are not all encompassed by the data in Figure 4.9. Consequently, values
of J/ρ and <J/ρ> for a given burst frequency are interpolated using an exponential fit to
the three burst mode cases as shown in Figure 4.10. This interpolated value is then used
for expressing momentum coefficients in Table 4.3 for the various optimal forcing
frequencies. No such interpolation is necessary for the unmodulated data in Table 4.2.
80
Figure 4.10: Exponential profiles used to interpolate momentum values for various AC-DBD plasma modulation frequencies.
Re/1000 J/ρ (N/mm) <J/ ρ> (N/mm) Cμ % <Cμ> % Cμ,tot% CE%
240 11.8 1.3 0.041 0.005 0.046 15
410 11.8 1.3 0.015 0.002 0.017 3.3
750 11.8 1.3 0.005 0.001 0.006 0.57
Table 4.2: Momentum and power characteristics for 3 kHz actuation at 20 kVpp.
81
Re/1000 fm (Hz) J/ρ (N/mm) <J/ρ> (N/mm) Cμ % <Cμ> % Cμ,tot% CE%
δf=20o
240 120 11.9 2.3 0.042 0.008 0.050 7.5
410 210 9.3 1.3 0.012 0.002 0.014 1.7
750 400 5.6 0.7 0.002 <0.001 0.002 0.29
δf=30o
240 90 12.9 2.6 0.045 0.009 0.054 7.5
410 160 10.7 1.9 0.014 0.002 0.016 1.7
750 270 8.1 1.2 0.003 0.001 0.004 0.29
δf=40o
240 70 13.5 3.0 0.047 0.011 0.058 7.5
410 120 11.9 2.3 0.015 0.003 0.018 1.7
750 230 8.9 1.4 0.004 0.001 0.005 0.29
Table 4.3: Momentum and power characteristics for 3 kHz actuation at 20 kVpp with
various burst frequencies at dc=50%.
Similar momentum calculations for the conditions of Figure 4.8 have also been
performed in an effort to quantify the dependence of momentum generation on the
modulation signal. An examination of the integral values in Figure 4.11 shows some
weak correlation with the results observed in Figure 4.8. For example the oscillatory
momentum coefficient peaks at 70% dc which is consistent with observations of the
82
phase-averaged V component in Figure 4.8, especially in the oscillatory case. The drop
off above 70% dc is also captured. The mean momentum coefficient values are shown to
saturate at 50% dc. The displays one of the limitations of using such integral parameters
as a comparison of the 50% and 100% dc profiles in Figure 4.9 reveal very different
behavior. The use of momentum coefficient will be revisited in the context of control
results in Section 6.1.
Figure 4.11: Mean and oscillatory momentum values for the AC-DBD plasma induced flow in quiescent air for modulation using AM and various BM waveforms.
83
Before discussing control results, some comments on the Cμ calculations are
warranted. ZNMF actuation is generally characterized using only oscillatory momentum
coefficient <Cμ> due to the periodic blowing suction behavior. The DBD plasma actuator
is certainly a ZNMF device, but the analogy with more traditional ZNMF devices such as
those used by NASA is not warranted since the plasma also produces a nonzero net mean
flow (Greenblatt et al. 2008). Thus, the characterization of such a device is more similar
to nonzero net mass flux devices. When the wall normal plasma induced flow is
neglected as is the case here, the actuator is essentially a pulsed jet and thus its
characterization requires the inclusion of both mean and oscillatory components of
momentum (Greenblatt and Wygnanski 2000). Despite this additional mean component,
Cμ,tot is still approximately an order of magnitude less than used in the majority NASA
experiments for a scaled version of the same airfoil. It is also important to note that
values of Cμ for periodic excitation produced by piezoelectric devices in the presence of
cross-flow have been obtained by correlating the pressure in the actuator cavity with the
velocity at the slot exit in bench top experiments. The pressure in the actuator cavity can
then be monitored during cross flow experiments giving a more accurate representation
of the actual momentum production (Melton et al. 2006). Calibrations for DBD plasma
actuators performed using nonzero freestream have resulted in even lower estimates of
momentum coefficient, thus the values reported in Table 4.2 and 4.3 are most likely an
overestimate further emphasizing the low momentum nature of the device (Goksel et al.
2006). The power dissipated by the discharge is also provided in Table 4.2 and 4.3 in
dimensionless form as power coefficient, CE, using:
84
EPC
q U c∞ ∞
= 4.8
where P is the time-averaged power dissipated by the plasma. Q-V power measurements
for modulated signals are more complex than the pure sine waves discussed previously.
Figure 4.12 gives an example of the Q-V curves for a 2 kHz carrier waveform modulated
using 100 Hz AM. The parallelogram is formed due to oscillations of the carrier signal,
which when modulated using the sine wave, expands and shrinks according to the peak
voltages. This complicates the Q-V area calculation and necessitates that the modulation
frequency must be taken into account.
Figure 4.12: Q-V diagram for AC-DBD plasma actuation using AM at 100 Hz.
85
Figure 4.12 shows the integrated energy dissipated on the actuator for
approximately one period of various 100 Hz modulation waveforms. The unmodulated
signal is also included for reference. The slopes for each condition are nearly identical
and the only obvious difference is the continuity of energy addition which changes based
on the modulation signal. Note that the curves are offset on the time axis for clarity. The
energy dissipated in the modulation period is found by noting the value associated with
the maximum energy (i.e. when the curve flat lines) and the corresponding power is
found by dividing by the modulation period. From this figure, it is apparent that power
and energy calculations for modulated waveforms can be simplified by multiplying the
power of 100% duty cycle case by the duty cycle of the modulation waveform in
question. For example, the 50% duty cycle case corresponds to approximately half the
energy of the 100% duty cycle case. The AM case is a more complex since the integrated
energy tends to have rounded corners associated with the more gradual sinusoidal
envelope rather than the abrupt on/off seen in the duty cycle case. Note that slightly more
than one modulation period has been captured in this case as another cycle begins near
t=5 ms. Despite these complexities, one can see that the energy associated with AM
saturates near ~40% of the maximum power (i.e 100% duty cycle).
86
-5 0 5 10Time [ms]
0
50
100
150
200Energy [mJ]
AM=100 Hz
Duty Cycle 100%
Duty Cycle 75%
Duty Cycle 50%
Duty Cycle 25%
Duty Cycle 10%
Figure 4.13: AC-DBD plasma integrated energies for various modulation waveforms.
The simplification outlined above for calculating energy and power of modulated
signals is sufficient in this work, but for cases in which very few high frequency carrier
signals are developed (i.e. very low duty cycle), the phase of the carrier frequency at
which modulation is turned on/off becomes important due to the current behavior of the
AC-DBD actuator (see Figure 4.3). This low duty cycle complexity will also be relevant
to momentum calculations. While this is an interesting thought exercise, optimizing
power requirements to this degree is not crucial until these devices become more
established in practical applications.
87
4.3 NS DBD Plasma Actuator Design
The NS-DBD plasma actuator is constructed exactly like the AC-DBD plasma
actuator. The only additional consideration now is that the cables connecting the high
voltage and ground electrodes must be as short as possible and of similar length as
previously described. To facilitate this, both leads are connected on the tunnel infield
side. Note that a typical NS-DBD plasma actuator lasts longer roughly twice as long as an
AC-DBD actuator due to short time scales involved with the NS duration pulse which
limits the time the dielectric is actually exposed to the discharge.
4.4 NS-DBD Plasma Actuator Characterization
Characterization of nanosecond pulse (NS) DBD plasma actuators highlights the
fundamentally different nature of the device when driven by these short high voltage
waveforms. Typical traces of simultaneously sampled voltage and current for one pulse
are shown in Figure 4.14. These traces have been acquired on a 30 cm long DBD load.
The construction of the actuator in the NS-DBD case is identical to the AC-DBD. It is
composed of copper tape electrodes as described previously and 3 layers of Kapton tape
dielectric (0.27 mm (0.0105 in) thick). The voltage pulse width (<100 ns) is at least three
orders of magnitude shorter than a typical AC driven DBD plasma period. The pulse
voltage is ~15 kV for this particular load (~30 cm). For longer actuators such as those on
the airfoil (46 cm) the peak voltage is slightly reduced to 12 kV due to the additional
length of the actuator. In either case, the zero to peak voltage amplitude is not
substantially different than the AC-DBD plasma actuator, but the short pulse width and
rise time has a drastic effect on the current developed. The peak current in the discharge
88
is 50 A in the NS-DBD case compared to ~200 mA for the AC-DBD. The effect of this
high current is seen in the energy and power developed in the pulse. In the latter case,
peak powers of up to 600 kW are created. While this power is exceptionally high due to
the short pulse width and high current, the average power for the NS-DBD is slightly less
than the AC-DBD plasma actuator operating at the same frequency and similar voltage
levels.
89
-200 -100 0 100 200 300 400-10
0
10
20
-20
0
20
40
60
Vol
tage
[kV
]
Time [ns]
Cur
rent
[A]
-200 -100 0 100 200 300 400-2
0
2
4
6
8[×105]
0
5
10
15
20
Time [ns]
Pow
er [W
]
Ener
gy [m
J]
Figure 4.14: Typical voltage, current (a) and power, energy (b) traces for an NS-DBD plasma actuator.
a)
b)
90
Tables 4.4 and 4.5 compare various electrical parameters for the two discharges.
The NS-DBD pulse energy is found to vary slightly with frequency and run time. The
latter has the form of a first order response to a step input with time constant of 400-600
seconds that is likely due to thermal effects in the power supply. Given that control
experiments were run for periods of less than 600 sec, the results in Table 4.4 are likely
an upper bound. Comparison of the two tables shows that for the 46 cm long actuator, the
NS-DBD consumes slightly less power than the AC-DBD plasma actuator operating at
the same frequency. Note that none of the issues encountered with simultaneous AC-
DBD measurements are encountered for the NS-DBD plasma actuator since the current
and voltage time scales are quite similar. In all the control cases explored in Chapter 7
which include tests in the range 60-2750 Hz, CE is less than 0.6%.
Frequency (Hz) Voltage (kV) Energy (mJ/pulse) Power (W)
100 ~12 < 11 < 1.1
500 ~12 < 14 < 7
1000 ~12 < 17 < 17
Table 4.4: Electrical properties of NS-DBD plasma on a 46 cm long flat plate actuator.
91
Frequency (Hz) Voltage (kVpp) Energy (mJ/cycle) Power (W)
1000 18 26 26
2000 17 21 42
3000 18 27 81
Table 4.5: Electrical properties of AC-DBD plasma on a 46 cm long flat plate actuator.
The ability of NS-DBD plasma to induce flow in quiescent air is investigated
using PIV experiments similar to the AC-DBD case. As before, the length of the actuator
on the flat plate is ~16 cm and this smaller load allows the generation of higher NS-DBD
peak voltages. Previous literature suggests the NS-DBD produces very weak velocity in
the neutral species (Roupassov et al. 2009) although some authors have studied the NS-
DBD induced vortex as a thrust generating mechanism (Opaits et al. 2008). The latter is
imaged using PIV raw data in Figure 4.15. The flat plate is arranged vertically and the
plate surface is visible by the laser reflection near x=-10 mm. When the NS-DBD plasma
is off (Figure 4.15a), there are some very weak ambient air motions seen by the vertical
striations near the flat plate surface, but these do not influence the discharge induced
flow. Initiation of the plasma discharge, seen in Figure 4.15b as the bright region on the
left side of the image creates a very large counterclockwise rotating vortex that is
qualitatively similar to published schlieren images (Opaits et al. 2007). This vortex is
quite slow (~4 cm/sec) and only occurs during the initial plasma generation at high
frequencies (≤~1 k suggesting the ambient fluid requires multiple plasma bursts to
92
produce this response. Once it propagates out of the frame of view, no other structures
are observed in the PIV raw data. Consequently, these effects do not influence the PIV
data. The rapid heating and expulsion of fluid near the surface by the initial plasma
propagation is believed to create this behavior and this is not observed in ensemble
averages. When the plasma is turned off (Figure 4.15c), charging the dielectric and its
repulsion of the olive oil seed particles is very apparent. This is displayed as a void in
particles near the dielectric and flat plate surfaces. The only region in the image that still
has particles near the surface is the area near the exposed electrode which does not suffer
from a similar surface charging effect. This surface charge buildup has been postulated as
an important phenomenon for AC-DBD plasma induced flow since it changes the electric
field distribution, positively or negatively affecting flow generation (Opaits et al. 2009b;
Font et al. 2010), but behavior to the extent observed in Figure 4.15 is not observed for
AC-DBD PIV raw data in this work.
93
Figure 4.15: Startup vortex (b) and dielectric charging effect (c) created by NS-DBD plasma actuator.
The weak induced velocity of the NS-DBD is confirmed by the data in Figure
4.16 which shows the mean velocity created by NS-DBD operating at 2 kHz for 100%
and BM at 90 Hz with 50% duty cycle measured 20 mm downstream of the electrode
interface. The weak magnitude of the NS-DBD induced flow is highlighted by plotting
the velocity generated by an AC-DBD operating at the same frequency at 20 kVpp. In
both cases, the AC-DBD is found to generate at least five times more peak velocity than
the NS-DBD. The limitations of the AC-DBD plasma actuator due to its low momentum
nature are well-established. By comparison, the NS-DBD is even weaker such that no
control authority is likely manifested by NS-DBD generated momentum at the Re
surveyed in this work. Consequently, values for momentum and corresponding
momentum coefficient for the NS-DBD are irrelevant. The weak velocity created by the
10 mm
a) b) c)
94
NS-DBD is thought to be a consequence of the extremely short time scales associated
with the plasma discharge which are not sufficiently long to allow substantial fluid
response.
95
Figure 4.16: Mean velocity profiles, U, for NS and AC-DBD plasma operating in quiescent air with 2 kHz carrier frequency using 100% (a) and 50% (dc) at 90 Hz
measured 20 mm downstream.
a)
b)
96
In Section 7.1 the NS-DBD is shown to generate superior control authority over
AC-DBD plasma actuators for LE airfoil separation control at high speeds. The preceding
velocity measurements suggest that the momentum addition cannot be a control
mechanism for the Re surveyed here. Instead, it is postulated that a thermal mechanism is
responsible. This scenario is well established for LAFPAs that have shown control
authority in subsonic and supersonic cold and hot jet experiments (Samimy et al. 2007)
and have also recently been supported by computations (Gaitonde 2009). One of the
characteristics of this control mechanism is the generation of compression waves due to
rapid local heating by the plasma. In the case of LAFPAs, these waves propagate
spherically as if generated by point source (Samimy et al. 2010). Similar, but
substantially more complex behavior has been confirmed for NS-DBD plasma using
phase-averaged schlieren imaging.
Figure 4.17 shows the compression waves imaged by viewing the discharge along
the major axis of an actuator. In this case, the plasma forms from right to left and can be
seen as the bright region near y=1.25 mm. The behavior imaged in Figure 4.17 is
relatively complex as the wavefront is composed of both a quasi-planar and cylindrical
compression wave, but appears relatively uniform along the viewing axis. This behavior
is consistent with numerical results of (Roupassov et al. 2009). However, the schlieren
technique by nature performs a spatial average normal to the field of view. The three
dimensionality of the discharge is observed by performing schlieren measurements
viewing the plasma from behind the high voltage electrode along its minor axis (Figure
4.18 and Figure 4.19). It is now apparent that what appeared to be relatively two
97
dimensional behavior is substantially more complex in that the cylindrical compression
wave along the span of the device is actually the summation of multiple spherical
compression waves created by localized regions of the discharge. Figure 4.18 shows
measurements taken with the schlieren knife edge vertically. This is done to highlight
density gradients along the y-coordinate and the quasi-planar shock behavior is well
imaged using this technique. Rotating the knife edge allows visualization of density
gradients along the x coordinate. In this case (Figure 4.19), the quasi-planar compression
wave is no longer visible. Note that the spherical compression waves are visible for the
knife edge in both orientations.
Figure 4.17: Schlieren imaging of compression waves generated by NS-DBD plasma actuator viewed along the major axis of the actuator.
98
Figure 4.18: Schlieren photography of vertical density gradients for NS-DBD plasma generated compression waves viewed along the minor axis of the actuator.
Figure 4.19: Schlieren photography of horizontal density gradient for NS-DBD plasma generated compression waves viewed along the minor axis of the discharge.
Δt=20 μs Δt=130 μs Δt=240 μs
Δt=49 μs Δt=159 μs Δt=269 μs
99
While the compression wave behavior away from the surface is visually
compelling, the efficacy of the device for flow control is more likely dependent on the
compression wave strength in the boundary layer of the flow field to be controlled. For
example, in the airfoil results that follow the maximum boundary layer thickness (~7
mm) is encountered on the TE and the LE should be considerably thinner. Thus it is more
appropriate to concentrate on the NS-DBD behavior near the surface. At the onset of the
NS pulse, both the planar and localized compression waves are visible, but the latter
appears to be the stronger of the two as it propagates slightly faster as seen from the
curved region protruding out of the quasi-planar wavefront in Figure 4.18. As the time
delay increases, the spherical and planar waves merge such that the only evidence of the
spherical pattern is the various circular arcs beneath planar wave. Recall that these are
phase-averaged data which mean the local “hot spots” in the discharge that produce these
stronger spherical wavefronts must be somewhat spatially stationary. At this time, it is
not possible to predict the location of these strong filaments, however it is postulated that
these intense regions could be manipulated by shaping the electrodes, perhaps in a
sawtooth fashion, to create stronger regions of electric field.
To truly understand this device, quantitative near surface high spatial resolution
measurements of temperature rise and compression wave strength are necessary.
Measurements of the latter a few mm from the surface using existing data show
approximately sonic compression wave speeds (Mach 1) which are consistent with
LAFPA generated compression waves. However, existing literature and limited near
surface data suggest that the compression wave strength here may be substantially
100
greater. A rough estimate of this effect can be made based on the energy deposited by a
pulse, ~15 mJ from Table 4.4. Let us assume that 100% of this pulse energy is deposited
into a two dimensional near surface gas layer as heat. The temperature rise associated
with this heating can be estimated simply by:
v
ETmC
∆ = 4.9
where ΔT is the temperature rise, E is the energy provided by the pulse, m is the mass of
the heated air and Cv is specific heat of air at constant volume which is used based on the
assumption that the fluid does not have sufficient time to expand during the heating
process (~50 ns). The mass of the air, m, or more importantly the volume of air used to
calculate this mass requires another assumption. The compression wave behavior near the
surface appears to show a wavefront that originates near the electrode interface rather
than over the entire extent of the 12.7 mm ground electrode. This heated region has been
estimated at 0.4 x 0.4 mm2 in the literature based on ICCD images of the NS-DBD
discharge (Roupassov et al. 2009). With this and the span of the device for the airfoil
cases in Chapters 6 and 7, (46 cm) the temperature rise near the surface is estimated to be
~235 K consistent with the literature (Roupassov et al. 2009). In reality, this is likely an
upper bound since the heating efficiency of the pulse energy is certainly not 100%. Also,
the two dimensional plasma assumption and the simplified calculation of heated air mass
are clearly not valid based on the schlieren images of Figures 4.18 and 4.19.
Using a similar temperature rise of ~200 K, compression wave strength near the
surface has been estimated at Mach 1.13 using one dimensional theory (Roupassov et al.
2009). This is again probably an upper bound since it is unlikely that a Mach 1.13
101
compression wave dissipates to approximately Mach 1 over one mm of propagation as
calculations from schlieren images like those in Figure 4.17 suggest. Still this may
provide some insight for quantifying a momentum or better amplitude coefficient for
these thermal perturbations. If we assume a Mach 1.1 compression wave that behaves
like a normal shock in atmospheric pressure (~100 kPa), the pressure difference across
the wave Δp is ~25 kPa using simple one dimensional normal shock relations. This can
be cast as a momentum coefficient analogous to Equation 4.6:
, ppzC
q cµ ∆∞
∆= 4.10
where z is the width of the heating region, assumed to be 0.4 mm and the supscript, Δp,
denotes the coefficient is defined based on a pressure difference rather than a
conventional momentum value, J. Using the pressure difference across a Mach 1.1
compression wave at atmospheric pressure, 25 kPa, and Re=750k, consistent with test
cases from Chapter 7, , pCµ ∆ =~4%. This is a reasonable number compared to more
standard ZNMF devices (i.e. synthetic jets) with substantial control authority (Cμ~1%).
This very simplified analysis may hint at the robustness of thermal effects for controlling
high speed flows. For example, even a very weak shock (Mach 1.02) for the Re=750k
airfoil case results in , pCµ ∆ =~0.6%.
An open question is whether the strength of the near surface compression wave
can be increased using bursts of these short duration pulses similar to the AC-DBD
signals of Figure 4.7. For this to occur, localized heating the near surface gas layer should
occur on a time scales shorter than the acoustic time scale. Again using an estimate of a
102
0.4 mm gas layer for air at 300 K, this heating should occur in less than ~1.2 µs which
means that if bursts of the NS-DBD pulses are used, they should probably be provided at
frequencies greater than 800 kHz.
It should be noted that similar compression wave behavior does not exist in the
literature for AC-DBD plasma actuators and thermal effects from these devices have been
ruled out as a control mechanism by multiple studies (Enloe et al. 2004b; Jukes et al.
2006; Sung et al. 2006). The compression wave behavior described above is also quite
repeatable for bursts of multiple pulses suggesting the dielectric charging has little
influence on the thermal mechanism.
103
Chapter 5: Baseline Verification and Characterization
Supercritical airfoils like the simplified NASA EET model are prone to stall at
relatively low α due to a small radius of curvature at the LE (Abbot and vanDoenhoff
1959). In practice, this is circumvented using a deflectable LE droop or slotted slat which
allows extension of the stall angle to higher α by promoting boundary layer transition and
weakening the adverse pressure gradient near the LE (Figure 5.1). This PFC device has a
significant effect on the CP distribution around the model as suction peaks are now
developed at the LE and over slat/droop shoulder (Melton et al. 2005). Such PFC devices
are also quite common on airfoils with deflectable TE flaps since these systems alone
serve to increase the effective camber of the model by increasing CL at a given α, but also
introduce stall at lower α.
Figure 5.1: Simplified high-lift airfoil with LE droop and TE flap.
104
The OSU version of the NASA EET is designed without a LE droop for
simplicity and the consequences of this decision are visible in Figure 5.2, which shows
CL versus α for selected Re and δf. The OSU airfoil reaches CL,max at relatively low α (6-
10o). Deflecting the flap increases CL over all α, but introduces stall at lower α as
expected from a change in effective camber. In addition, some Re effects are apparent
especially between the low (240k) and high Re (750k) test cases in that the generated CL
increases with Re at a given α. Sample CP curves in Figure 5.3 give more detail of the LE
stall behavior. The flow abruptly separates from the LE rather than gradually moving
upstream from the flap shoulder as α is increased, beginning with a reduced and
broadened LE suction peak at α=6o. This is also accompanied by a closed separation
bubble slightly downstream of the LE near x/c=0.1 in the α=6o and α=8o cases before
deep stall, recognized by the near-zero pressure gradient on the suction surface is
established at α=10o.
105
Figure 5.2: Baseline CL vs. α for sample Re and δf.
Figure 5.3: Baseline CP behavior for Re=410k and δf=20o at various α.
106
The flow field over the deflected flap is characterized by vortex shedding at Fc+
between 1-3 depending on δf where larger flap deflections create lower frequency
shedding (Figure 5.4). The dynamic signature of vortices shed from the flap shoulder is
independent of α until the separation point moves upstream after which a turbulent
spectrum with no recognizable frequency peak is established (Figure 5.5). The frequency
of vortex shedding obeys F+ scaling as expected (Figure 5.6). The flat spectrum
beginning around Fc+=10 in Figure 5.6 is created by an overly noisy measurement system
and is unphysical. The behavior of flow separating from the deflected flap shoulder as
described in Figures 5.2-5.6 is consistent across the parameter space surveyed. Separation
from the LE is much more chaotic, with no recognizable dynamic signature (Figure 5.7).
Figure 5.4: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.95 for Re=410k and α=0o at various δf.
107
Figure 5.5: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.95 for Re=410k and δf=20o at various α.
Figure 5.6: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.90 for α=0o and δf=30o at various Re.
108
Figure 5.7: Baseline PSD of fluctuating pressure, cp, measured at x/c=0.40 for Re=750k, δf=0 at various α.
A direct comparison of OSU and NASA results without LE droop in Figure 5.8
shows three significant differences: 1) The pre-stall dCL/dα for the two airfoils is
considerably different when δf=0; 2) the difference in stall angle between the two airfoils
ranges from ~2o for δf=0 to ~4o for δf =30o and 3) even when the pre-stall dCL/dα are
similar, the OSU results are shifted by CL~-0.1. The latter discrepancy is most easily
explained. A comparison of the CP curves for the two models at α=8o with δf=0 shows
good LE suction peak agreement, however NASA EET creates a slightly fuller profile
over the main element (Figure 5.9). This is partially due to the joint between the flap and
airfoil main element in the OSU case. This CP agreement persists for moderate flap
deflections (< 20o).
109
Figure 5.8: Baseline comparison of OSU and NASA (Melton et al. 2006) results for tripped and untripped behavior of CL vs. α for Re=750k and various δf.
Figure 5.9: OSU (a) and NASA (b) CP curves at Re=750k, α=0 and δf=0 (Melton et al. 2003).
a) b)
110
Figure 5.10 compares the stall characteristics of the OSU and NASA airfoils at
Re=410k with δf=12o. The only obvious discrepancy between the two data sets occurs at
α=8o where the NASA model appears to enter a deeper stall than OSU which retains a
broad, but visible suction peak at the LE with secondary peak downstream near x/c=0.15.
This dual peak signifies a closed separation bubble on the LE and is prevalent at many of
the post-stall conditions. This is the cause for the OSU flat post-stall CL behavior in
Figure 5.8 compared to the drastic CL decrease seen in the NASA curves. Note also that
the unpublished NASA data from Figure 5.10b shows that some Re effects were
prevalent at the low Re cases since the NASA CL curves at α=8o show no sign of stall at
Re=750k even for δf=30o.
Figure 5.10: OSU (a) and NASA (b) stall characteristics at Re=410k and δf =12o (Melton 2006).
a) b)
x/c
Cp
0 0.2 0.4 0.6 0.8 1
-6
-4
-2
0
P=3166, cL=1.21, a=2.0, ds=0.1, df=12.2P=3167, cL=1.33, a=4.0, ds=0.0, df=12.2P=3168, cL=1.47, a=6.0, ds=-0.3, df=12.3P=3169, cL=1.59, a=8.0, ds=-0.3, df=12.2
111
The first two NASA/OSU discrepancies mentioned above can be addressed
together, but first note that the OSU version is 5/8 scale of NASA and thus has a more
limited array of static pressure taps due to size constraints. This is especially true near the
flap shoulder as the NASA model is instrumented with taps that expose upon flap rotation
whereas OSU model does not. Nonzero δf, alone can rectify the dCL/dα discrepancy at
OSU pre-stall (Figure 5.8). The CP curves in Figure 5.11 highlight this behavior. At
δf=10o, Re=240k shows separated flow on the flap (Figure 5.11). Increasing the
freestream Re to 410k is sufficient transition the boundary layer to turbulent and prevent
separation at the flap shoulder indicated by the large suction peak now visible. This effect
can be seen over the entire suction surface as an increase in circulation is apparent. In this
case, sufficient resolution is available to capture this TE suction peak, but this is not the
case for δf=20o in Figure 5.11. Despite this lack of resolution, it is clear that δf=20o
rectifies the Re discrepancy and shows CP agreement at all cases surveyed. This suggests
that the increased circulation caused by the flap deflection encourages turbulent transition
farther upstream on the airfoil. However, this alone is not sufficient to cause immediate
transition at the LE since the stall angle is still quite different.
To further explore this behavior in cruise configuration (δf=0o), various tripping
mechanisms in form of grit roughness, low profile vortex generators and
backward/forward facing steps created by spanwise tape were applied to the airfoil LE.
Tripping devices are common on many airfoils and are used to accelerate the laminar to
turbulent transition in an effort to delay separation in regions of high adverse pressure
gradient. A sample of these results is also shown in Figure 5.8. Note that the trip strip
112
increases the stall angle by approximately two degrees in both cases and the δf=0 dCL/dα
now agrees well with NASA results. The presence of a passive plasma actuator at the
airfoil LE can also serve as a trip. This effect will be considered in section 7.1 which
describes studies of LE airfoil separation control. Additional tripping experiments are
able to more closely match NASA results, but this effect is limited to Re=750k and
similar improvements in performance are not repeatable at lower Re. Consequently, no
LE trip has been used in the TE separation control results that follow. Because of these
issues, results are limited to α≤6o for δf =20-40o due to boundary layer separation
upstream of the flap shoulder. Smaller values of δf are not analyzed because separation
occurs much further downstream of the flap shoulder (Figure 5.11). As a final check of
the validity of comparison of OSU baseline results to NASA, the pressure spectrum at
x/c=0.95 is shown in Figure 5.12. Despite the deflected droop in the NASA work, the
dynamics of the TE flow field remain quite similar.
113
Figure 5.11: Re and δf effects on OSU airfoil CP distributions.
a)
b)
114
Figure 5.12: OSU (a) and NASA (b) PSD of fluctuating pressure, cp, at Re=240k, α =0, δf =45, δs =-25 in (Melton et al. 2006).
Measurements depicting the three-dimensionality of the baseline flow are shown
in Figure 5.13 for the three primary flap deflections surveyed in this work at Re=410k
and α=0. Measurements are shown at ¼, ½ and ¾ span. The agreement is quite good even
for cases in which large separated regions which are inherently 3D exist on the flap.
Since baseline agreement between OSU and NASA has been verified, no blockage or
wind tunnel corrections for integral parameters have been employed. These corrections
will change values of integral parameters (CL, CD), but the effects of control are the
primary focus of this work and are believed to be conservative irrespective of whether
such corrections are employed.
a) b)
115
Figure 5.13: 3D behavior of CP for Re=410k, α=0 and δf=20 (a), δf =30 (b) and δf =40 (c).
a)
b)
c)
116
The state of the boundary layer is an important parameter for separation control.
Since transition occurs somewhere on the main element downstream of the LE, it is
important to verify the boundary layer approaching the flap is turbulent so that results are
viable for practical large-scale applications. The boundary layer velocity profile has been
measured upstream of the flap shoulder (x/c=0.7) using PIV. A sample result is shown in
Figure 5.14. The results of these measurements are summarized in Table 5.1. In all cases,
the boundary layer follows a turbulent profile according to:
1/ nU yU δ∞
= 5.1
where U is the average velocity profile, U∞ is the freestream velocity, δ is the 99%
boundary layer thickness. The exponent, n, varies from 4 to 6 and δ decreases in
thicknesses with increasing Re as expected. The shape factor, H= δ*/θ, where δ* is the
displacement thickness and θ is the momentum thickness remains relatively constant
around 1.4 ensuring the boundary layer approaching the flap is similar in all cases. The
baseline results show that even though LE stall characteristics between the NASA and
OSU models are different, the results are consistent at OSU pre-stall α as the separating
boundary layer is turbulent and has similar dynamic features upon separation.
117
Figure 5.14: PIV-measured boundary layer velocity profile with power law fit for Re=240k, α=0, δf=30.
118
Re/1000 δf (deg) n δ (mm) δ* (mm) θ (mm) H
240 20 5.32 5.14 0.81 0.59 1.38
410 20 5.51 4.77 0.73 0.54 1.36
750 20 5.9 4.41 0.64 0.48 1.34
240 30 5 6.24 1.04 0.74 1.40
410 30 4.86 5.14 0.88 0.62 1.41
750 30 5.76 4.41 0.65 0.48 1.35
240 40 4.89 6.98 1.18 0.84 1.41
410 40 5.14 5.87 0.96 0.69 1.39
750 40 4.93 4.77 0.81 0.57 1.41
Table 5.1: Boundary layer properties at x/c=0.70 measured using PIV.
119
Chapter 6: Trailing Edge Separation Control
6.1 Single AC-DBD Plasma Actuator Control Results
A single DBD plasma actuator is applied to the flap shoulder of the airfoil as
shown in Figure 6.1. This is done by first deflecting the flap and then applying the
various layers of adhesive tape according to the description in Section 4.1. The resulting
actuator location is x/c~0.77 independent of δf. This location is an acceptable estimate of
the separation point where actuators are generally found to be effective for controlling
flow separation. Note that this is in contrast to the fixed actuator locations used by
NASA. Thus, the most relevant comparison of forthcoming results are made in reference
to NASA actuation locations at and between FWD and slot 3 which are located at
x/c=0.725 and 0.757 respectively in reference to the cruise configuration (δf=0) (Melton
et al. 2006) Despite the documented sensitivity of TE control authority to actuator
location, the following results are quite repeatable for DBD plasma actuators manually
applied to the airfoil surface.
120
Figure 6.1: Example of a DBD plasma actuator applied at the flap shoulder (x/c=0.77) of the simplified NASA EET airfoil, δf=30o.
It should be noted that each time an actuator is placed on the model the baseline
flow changes slightly due to the presence of the actuator alone, most notably from the
surface discontinuity created at the actuator leading and trailing edge. Figure 6.2a shows
the effect of the actuator on the CP distribution around the model. There is some change
most notably near the LE, but overall this effect is marginal. For practical applications,
this is not a concern since the actuator should be embedded in the airfoil substrate nearly
flush mounted to the surface. A more important result in Figure 6.2b shows that the
dynamics of the flow field are relatively unaffected by the passive presence of an actuator
on surface. These results are consistent for the conditions surveyed. For the remainder of
the work, any reference to the baseline configuration pertains to the case where an
actuator is present on the model surface without plasma (power off).
Exposed HV electrode
Dielectric extent
121
Figure 6.2: Effects of passive (power off) DBD plasma actuator on CP (a) and PSD of fluctuating pressure, cp, at x/c=0.90 (b).
a)
b)
122
Results of frequency sweeps for various Re, δf, actuator locations, input voltages,
modulation waveforms and carrier and modulation frequencies are presented in Figure
6.3a. The data are plotted against FL,m+. Only results outside of the ±0.02
uncertainty/repeatability band are relevant. The actuation cases that do not employ low
frequency modulation, referred to as quasi-steady actuation hereafter, have been plotted
at FL+=0 to reduce the abscissa range for clarity. In reality, FL
+ is in the range 2.8-8.5 for
the cases shown. Only a relatively narrow band of frequencies, FL,m+=0.3-0.5, shows
clear increases in CL. In reality, this effective band of frequencies is even narrower than it
appears. Portions of the data in Figure 6.3a are plotted using a different scale in Figure
6.3b. The modulation frequencies are normalized by fTE in this case. A discrete frequency
preference is now obvious in that excitation at F*=1 generates the most lift enhancement.
In some cases, the subharmonic of this frequency (F*=1/2) also has an effect, but there is
no experimental evidence to support that vortex shedding from the flap is shifted to this
lower frequency. Rather, the increased CL value seems to occur through a realizable, but
less efficient amplification of the natural shedding. In general, the control authority is
found to decrease with increasing Re and δf. The frequency preference in Figure 6.3 is
similar to presented NASA results of lift enhancement for a slightly different version of
the same airfoil using piezoelectric driven ZNMF actuators. However, the DBD plasma
actuator control authority for low frequency excitation above FL,m+=1 and quasi-steady
forcing is significantly reduced in comparison. This is due to the low momentum nature
of DBD plasma compared to the NASA actuation which limits control authority to very
narrow frequency bands corresponding to natural flow instabilities.
123
Figure 6.3: Effect of dimensionless actuation frequencies FL,m+ (a) and F* (b) on ΔCL at
α=0o for various Re and δf.
a)
b)
124
A sample set of CP curves for the results of Figure 6.3 are presented in Figure 6.4.
The difference between the baseline and both actuation cases is clearly visible on the
flap. The quasi-steady actuation case at FL+=12.7 generates less suction on the flap than
the baseline case. Forcing at F*=1 increases suction on the flap and more importantly
enhances circulation around the entire model as the area enclosed by the CP curve
expands especially on the suction surface. This circulation increase and corresponding
upstream effect create the majority of the lift enhancement. This behavior is typical of TE
airfoil separation control in that a significant portion of the lift enhancement is due to
upstream effects rather than full reattachment of flow to the flap (Kiedaisch et al. 2006;
Melton et al. 2006). It should also be noted that the model scale and obstruction of static
pressure taps by the actuator do not allow the resolution of any suction peak near the flap
shoulder (x/c=0.75) as exhibited in NASA results (Melton et al. 2006). However, a slight
favorable pressure gradient is visible around x/c=0.7 for both forcing cases suggesting it
does exist. The lack of CP resolution here at least partially accounts for the increased ΔCL
values in the NASA case over OSU.
125
Figure 6.4: Baseline and controlled CP behavior for Re=240k, α=0o and δf =20.
Figure 6.5 provides a more illustrative example of the behavior exhibited in the
CP curves. The streamlines calculated from two-component PIV are shown for the three
cases of Figure 6.4. The effect of actuation is readily apparent on the time-averaged flow
field. Quasi-steady forcing (Figure 6.5b) serves to energize the boundary layer and move
the separation point downstream. This delayed separation has the effect of lengthening
and flattening the recirculation region and confirms the suggestion that a slightly stronger
suction peak should exist in the CP curve near the flap shoulder in comparison to the
baseline. The persistence of the recirculation region on the flap explains the diminished
circulation increase in comparison to forcing at F*=1 (Figure 6.5c) which has a drastic
effect of the time-averaged streamlines. In this case, the separation location has again
126
moved downstream, but now the time-averaged streamlines resemble almost potential
flow with only a slight recirculation region visible. This also confirms the existence of a
stronger suction peak near the flap shoulder in comparison to both the baseline and quasi-
steady forcing cases.
127
Figure 6.5: Baseline and controlled (dc=50%) time-averaged streamlines for Re=240k, α=0o and δf =20o.
c) FL+=12.7
FL,m+=0.5
F*=1
b) FL+=12.7
a) Baseline
128
The time-averaged spanwise vorticity fields, *
/c U∞Ω = Ω , corresponding to the
streamlines of Figure 6.5 are shown in Figure 6.6. As expected, strong regions of
vorticity exist above and below the flap corresponding to the shear layer between the
high-speed freestream and recirculation region. Strong vorticity on the flap persists for all
cases emphasizing that even when time-averaged streamlines more closely follow the
flap surface as in
Figure 6.5c, there is still considerable interaction here from flow structures
excited by the plasma (Figure 6.6c). This is further discussed in the context of phase-
averaged measurements that follow. The vorticity fields additionally show the extent of
recirculation region and its effect on the time-averaged wake. Quasi-steady forcing serves
to delay separation, but lengthens the extent of the recirculation region in comparison to
both the baseline and F*=1 forcing cases. The latter case is found to reduce the size of the
recirculation region and pull the region of strong vorticity closer to the flap element.
These effects are visible in the vorticity on both the pressure and suction sides of the
airfoil downstream of the flap and are most evident from profiles at x/c=1.1 shown in
Figure 6.7.
129
Figure 6.6: Baseline and controlled (dc=50%) time-averaged normalized vorticity,
*Ω ,
for Re=240k, α=0o and δf =20o.
c) FL+=12.7
FL,m+=0.5
F*=1
b) FL+=12.7
a) Baseline
130
Figure 6.7: Baseline and controlled time-averaged vorticity magnitude, *Ω at x/c=1.1 for
Re=240k, α=0o and δf =20o.
The time-averaged results of Figures 6.5-6.7 do not truly represent the physical
behavior of the flow. Recall that the most striking benefits of control are realized by
amplifying the natural instability of the trailing edge flow field. In a temporal sense, this
slightly delays boundary layer separation, but a more obvious result is the amplification
and organization of the natural vortex shedding downstream. This increases entrainment
of freestream momentum into the shear layer and recirculating region even beyond the
point of separation from the flap surface. While this does not serve to fully reattach the
flow, the reduced size of the recirculating region has a profound effect on the circulation
around the model and subsequently on the sectional CL as seen in Figure 6.3.
131
This organized nature of shedding structures is most apparent from measurements
of phase-locked vorticity (Figure 6.8) and pressure fluctuations on the TE flap where both
the broadband and the coherent portions become amplified considerably (Figure 6.9). The
latter case has been verified using the techniques from Section 3.4.2. This highlights the
detrimental aspect of forcing at F*=1 in that the dynamic loads on the flap due to vortex
shedding are substantially increased. Unfortunately, transducer and hot wire
measurements in the presence of plasma actuation are quite difficult due to EMI and
reliable results have not been obtained for the quasi-steady forcing cases. Instead, the
spatial correlation of PIV data is used to investigate the organization of the flow field in
the wake in Figure 6.10. The normalized spatial correlation of the fluctuating vertical
component of velocity, Rvv, is computed along the y/c coordinate associated with the
trailing edge of the deflected flap (y/c=yT) using
( )( ) ( )
( ) 2
, ,,
,
T T
vv T
T
v x y v x x y dxR x y
v x y dx
∞
−∞∞
−∞
+ ∆∆ =
∫
∫ 6.1
This location is a reasonable estimate of the symmetry plane associated with the
spatial dynamics of the vertical fluctuating velocity component in the turbulent wake
(Figure 6.11). A correlation function is computed for each instantaneous fluctuating
vertical velocity field and the average correlation waveform is computed using the 1000
image sample set. The baseline and F*=1 forcing cases behave as expected from the
132
results in Figure 6.9. Both exhibit a similar wavelength with the latter being amplified
considerably due to forcing near the natural vortex shedding frequency. The quasi-steady
forcing case has similar correlation levels as the baseline, but with a slightly shorter
wavelength consistent with oscillations at a higher frequency. While the shift in the
correlation peak toward higher frequencies is weak, the trend is in general agreement
with pressure spectra of NASA using piezoelectric ZNMF actuation at high frequency
(Melton et al. 2006).
Figure 6.8: Phase-averaged normalized vorticity fields (ΔΦ=π), *
Ω , for Re=240k, α=0o and δf =30o forced at FL
+=8.5, FL,m+=0.4 (F*=1).
133
Figure 6.9: Baseline and controlled PSD of fluctuating pressure, cp, measured at x/c=0.90 Re=240k, α=0o and δf=30o.
Figure 6.10: Spatial correlation of the v component of velocity, Rvv, at y/c=yT for Re=240k, α=0o and δf=30o for baseline and controlled (dc=50%) flows.
134
The POD method is used to gain additional insight into the dynamics of the flow
field (Holmes et al. 1996). The method decomposes an uncorrelated set of velocity field
snapshots into a set of spatial orthonormal modes or POD modes, ( )xiϕ , and modal
amplitude coefficients for these modes, ai(t). The decomposition is proper in the sense
that the modes are organized in order of decreasing energy content to produce an optimal
set of basis functions that capture the dominant characteristics of the fluctuating velocity
field. The POD expansion (6.2) expresses the fluctuations of the two-dimensional vector
field ( ) ( ) ( ), , ,,t t tw u v = x x x in terms of the modes that contain the major portion of
the kinetic energy in the flow.
( ) ( ) ( )1
, N
i ii
tw a t=
≅ ∑x xϕ 6.2
where x in this case represents two-dimensional space (x,y). Each snapshot contains u and
v components of instantaneous velocity on the x-y plane near the tunnel centerline. For
the purposes of this work 1000 snapshots are used to derive the POD modes and
statistical quantities.
The first four normalized POD modes for the v component of velocity,
( ) ( )1 4 1 4 /y y U− − ∞∗ϕ = ϕ , for the baseline and controlled flows using quasi-steady
(F+L=8.5), F*=1 and F*=2 forcing at Re=240k, α=0 and δf=30 are shown in Figure 6.11.
In all cases, a recognizable oscillation exists in the trailing edge wake. This is expected
for the baseline and F*=1 forcing case and further supports the suggestion that
135
oscillations persist near this low frequency even for quasi-steady forcing. Forcing at
F*=1 shows more amplified spatial modes compared to the other cases and a
recognizable signature of structures emanating from the flap shoulder near the actuator
location. At F*=2, a clear oscillation exists in modes 3 and 4 at a wavelength consistent
with higher frequency forcing. These shorter wavelength cases have little effect on the
overall flow field and most importantly CL.
Figure 6.12 presents corresponding energy results for the data of Figure 6.11 and
further emphasizes the low dimensional organized nature of the low frequency forced
flow field (F*=1) as nearly 60% of the total energy is contained in modes 1 and 2. The
quasi-steady forcing removes energy from the first 2 modes and redistributes it to mode 4
in particular. The F*=2 case has lower energy content in modes 1 and 2, but slightly
higher values for modes 3-7. The low energy levels of modes 3 and 4 explain why the
interesting high frequency dynamic content in Figure 6.11 at F*=2 has little effect on the
overall flow field. Figure 6.12b shows cumulative energy of the data of Figure 6.12a
where it is seen that the increased energy in modes 1 and 2 for the F*=1 forcing case
accelerates convergence, while the quasi-steady does the opposite. The F*=2 case also
has weaker modes 1 and 2, but the higher energy levels in modes 3-7 are sufficient to the
make summation of this energy quite similar to the baseline. The data in Figures 6.11-
6.12 are representative examples for the parameter space surveyed. In general, the trends
outlined above tend to increase and decrease with increasing and decreasing control
authority. For example, at higher Re such as 410k, the quasi-steady forcing has little
effect on the POD energy content.
136
Figure 6.11: First four normalized baseline and controlled (AM) POD modes of the v component of velocity, ( )1 4 y−
∗ϕ , for Re=240k, α=0o and δf=30o.
Baseline
FL+=12.7
Mode 1
FL+=12.7, F*=1
FL+=12.7, F*=2
Mode 2 Mode 3 Mode 4
137
Figure 6.12: Modal energy (a) and cumulative energy (b) for baseline and controlled (AM) POD modes for Re=240k, α=0o and δf=30o.
a)
b)
138
NASA results showed modulating the high frequency waveform using a square
wave (BM) rather than a sine wave (AM) could increase the efficiency of actuation since
the duty cycle can now be lowered thereby decreasing power consumption (Melton et al.
2004). A similar study for three δf at Re=240k is shown in Figure 6.13 using DBD
plasma. Recall from Section 4.2 that the power consumption of AM is approximately
equal to BM at 40% duty cycle. Duty cycles of at least 20% create a significant
improvement over the lower cases and each flap configuration has a preferred range of
duty cycle percentage for control that increases with δf. These results show that BM can
be both more efficient and more effective for enhancing lift. This is due to the response
of the natural instability to on/off nature of the BM perturbation in comparison to the
more gradual sinusoidal envelope associated with the AM case although quantification of
this effect requires future study. While this conclusion generally agrees with NASA’s
results, the functional relationship between ΔCL and duty cycle is quite different (Melton
et al. 2004). In their work, ΔCL was shown to saturate at 45% duty cycle for a δf=20o at
α=6o for excitation near the flap shoulder. Further increases in duty cycle required more
power input, but showed none of the detrimental effects on CL exhibited here. This
discrepancy is likely due to the pulsed blowing nature of the DBD plasma compared to
pulsed blowing/suction in the NASA case although this has not been show in the
literature to the author’s knowledge. Note that studies of LE separation control with
plasma actuators report significant authority and even complete reattachment for duty
cycles as low as 6% (Benard et al. 2009b). The results of Figure 6.13 highlight the more
139
challenging problem of separation control over a TE flap which requires greater periodic
momentum input in comparison (Melton et al. 2006).
Figure 6.13: Effect of modulation waveform on ΔCL for Re=240k, α=0o and variable δf when forcing at FL
+=8.5, F*=1.
An attempt to capture the pulsing effect of the AC-DBD plasma actuator in still
air is shown in Figure 6.14 using the momentum definitions outlined in Section 4.2.
These measurements have been acquired for similar forcing conditions as those shown in
the δf=30 case. The intent of this calculation was to show some correlation between the
140
oscillatory momentum coefficient and ΔCL. There does appear to be some relationship
here as both functions peak around 70% dc, but the momentum curve does not drop off as
significantly at higher dc. This highlights the difficulty in comparing actuator still air
measurements to behavior in a flow control environment and shows that considerably
more work is required to quantify these amplitude effects (Seifert and Tillman 2009).
Figure 6.14: ΔCL and <J/ρ> as a function of BM dc for Re=240k, α=0, δf=30, FL+=8.5,
FL,m+=0.4, F*=1.
With this knowledge, the effect of modulation and carrier frequency can be
further discussed. It is well known that AC-DBD plasmas create most of the momentum
transfer in the forward stroke half cycle of the carrier frequency (Forte et al. 2007; Enloe
141
et al. 2008b). For example, 100 Hz modulation of a 2 kHz carrier frequency with 50%
duty cycle corresponds to 10 high frequency cycles per modulation cycle. A 4 kHz carrier
frequency with the same modulation frequency (100 Hz) and duty cycle (50%) would
contain 20 high frequency cycles per modulation cycle. This implies that for a given
carrier and modulation frequency, there may exist an optimum number of high frequency
cycles and subsequent relaxation time for creating the strongest perturbations. This
optimum would obviously be governed by the system in question, the time response of
the induced flow to the pulsed signal and how receptive a particular flow field is to
excitation. This also implies that for a given dielectric with optimum carrier frequency,
there exists an upper limit for low frequency modulation. As before, this is governed by
both a minimum number of high frequency cycles to affect the flow and necessary
relaxation time to create the perturbation effect.
These findings indicate that dielectrics with high optimized frequencies give the
most flexibility for exciting high bandwidth low frequency modulations. In practical
applications, this point may be moot since scaling by reduced frequency (F+) dictates that
as length scales increase frequency scales decrease thus easing requirements for high
frequency excitation. However, for the sake of argument, a future AC DBD plasma
actuator design criteria could center around selecting a suitably robust dielectric that has
a dimensionless optimum carrier frequency on the order of F+~10 to allow the high
frequency forcing benefits (Glezer et al. 2005) while retaining significant bandwidth in
the low frequency regime more traditionally investigated. The success of such a device
142
would hinge on the selection or design of dielectric materials specifically optimized for
DBD plasma actuation.
The efficacy of DBD plasma actuators for TE separation control at nonzero α is
shown in Figure 6.15. Characterization of the baseline flow indicates that the dynamics of
the TE flow field remain relatively consistent as long as the separation location does not
change (α<6o) (Figure 5.5). This is also apparent in control results as forcing at F*=1 for
a given δf creates a near constant increase in CL for α<6o. The diminished control
authority at α=6o is due to the movement of the separation point upstream of the actuation
location. To retain the benefits of AFC at α≥6o, some PFC or AFC actuator is required at
the LE to maintain attached flow along the main element up to the flap shoulder. As
expected, the control authority decreases with increasing δf and Re, respectively. In the
former, the progressively more severe adverse pressure gradient and centripetal
acceleration imposed along flap shoulder requires greater momentum input to delay
separation and enhance circulation around the model. The reduction in actuator control
authority as Re is increased is well-established in the literature for both DBD plasma and
more traditional ZNMF devices operating at a fixed amplitude (Melton et al. 2006;
Greenblatt et al. 2008).
143
Figure 6.15: Effect of Re, α, and δf on ΔCL for forcing at F*=1 using dc=50%.
Re=240k
Re=410k
Re=750k
144
The effect of control on CD has yet to be considered. NASA results show that
forcing can have a significant effect on Cdp that is strongly dependent on both Cμ and
actuation location and does not necessarily correlate with CL (Melton et al. 2004; Melton
et al. 2006). Due to the inaccuracy of Cdp for quantifying total drag values and the
obstruction of static pressure taps by the plasma actuator, these effects are not considered.
Instead, calculations of CD from PIV wake surveys that include contributions of turbulent
stresses are employed (Lu and Bragg 2003).
.1 .1
20.3 0.3
2 1DU U y vv uu yC d dU U c U c∞ ∞ ∞− −
− = − +
∫ ∫ 6.3
where U and U∞ are the time-averaged local and freestream dynamic pressure
respectively. The turbulent stresses must be included since measurements of static
pressure in highly turbulent regions are difficult to obtain.
The effect of control is apparent in the near wake velocity profiles, but this
phenomena disappears downstream (Figure 6.16a). Note that the profiles are offset for
clarity and the spanwise Re normal stress is assumed negligible. The uu stress behaves
similarly to the mean velocity in that the effects of control are seen in the near wake, but
quickly dissipate downstream (Figure 6.16b). Only the vv stress shows an effect of
control significantly downstream (Figure 6.16c). This profile with equation 6.3 suggests
that CD should increase for F*=1 forcing. While this effect seems apparent in the wake
profiles, the contribution of this term to the CD calculation is quite small, less than 10%
145
on average across the parameter space, such that overall calculations of CD at x/c=1.5
show no significant change between the baseline and control cases at α=0o due to slight
variations in the U and uu profiles. The uncertainty surrounding CD calculations from
wake surveys for these highly separated flows is well established (Zaman and Culley
2008). Thus, significantly more work is required to quantify the effects of DBD plasma
on CD in this flow field, but at present the effect of increasing CL via instability
amplification does not appear to be prohibitive from this perspective.
146
Figure 6.16: Mean U velocity profiles (a) and Reynolds stresses uu (b) and vv (c) used
in CD calculations for Re=240k, α=0 and δf=20 at x/c=1.1, 1.2, 1.3, 1.4 and 1.5 (left to right).
a)
b)
c)
147
6.2 Perspective on OSU Single AC-DBD Control Results
As the NASA and ADVINT programs have shown, increasing the amplitude of
forcing generally serves to improve control authority until some saturation level is
reached. The scaling laws for this behavior are not well defined and an open question is
whether to cast this in terms of Cμ, velocity ratio, etc (Seifert and Tillman 2009). In this
work, the amplitude effect has been cast as Cμ and values are generally one order of
magnitude less than those used by NASA. Yet control results at low Re are comparable
when forcing at F*=1 seemingly rendering DBD plasma actuators quite efficient from a
Cμ standpoint when used primarily to amplify the natural instability. This analysis is
complicated by the vague definition of Cμ calculations for DBD plasmas as well as the
uncertainty on both the effects of direction and surface area/volume over which control is
applied. It should be noted that a complete optimization for DBD plasma induced flow
has not been employed in this work and significant improvements have been
demonstrated in the literature. Recent studies that show plasma induced thrust can be
increased by an order of magnitude suggest that application of these devices at more
practical takeoff and landing velocities speeds is realizable (Thomas et al. 2009). The
obvious question is whether additional increases in Cμ from DBD plasma follow similar
trends as those shown for Cμ increases in the ZNMF NASA work (Melton et al. 2006).
An investigation of this possibility requires more robust dielectrics and higher voltage
inputs which are beyond our capabilities at this time.
148
6.3 Attempts to Extend AC-DBD Plasma Control Authority
Literature has shown that the simultaneous use of multiple actuators distributed
along the airfoil chord has been more successful than the contribution from each actuator
alone (Greenblatt 2007). NASA studies of specifically TE separation control showed that
additional CL gains could be realized using actuators both upstream and downstream of
the flap shoulder. The relative phase between the two modulation waveforms was found
to be an important parameter. Operating the two devices in phase created additional lift
gains, but operating the two with a pi phase shift had a detrimental effect (Melton et al.
2004). This was studied using two AC-DBD plasma actuators straddling the flap shoulder
operating as a parallel circuit (Figure 6.17). Figure 6.18 shows that using both actuators
can create additional increases in control authority, but this has not yet been superior to a
single actuator placed in an optimized location. Future work is required to examine the
effect actuator phase on this system, but these studies are limited by the size of the airfoil
model and actuators. The latter cannot be significantly reduced in size without reducing
the induced flow generated by the device. Thus, there is a limitation on spacing these
devices around the airfoil. Quantifying the effect of two AC-DBD plasma actuators for
TE separation control will likely require a larger airfoil to allow more options for spacing
the devices.
149
Figure 6.17: DBD plasma actuators straddling the flap shoulder.
Figure 6.18: Normalized time-averaged vorticity magnitude profiles, *Ω at x/c=1.1 for
Re=240k, α=0, δf=20 for baseline and AC-DBD actuators operating in phase at FL+ =12.7
and F*=1.
150
The effect of forcing direction on control authority of AC-DBD plasma for TE
separation control has also been investigated. The majority of published studies use
devices that produce force dominant body forces in the streamwise direction. Figure 6.19
shows the result of producing the body force upstream. In this case, the actuator has been
rotated about the electrode interface which is believed to be the region where force
density created by the plasma should be near maximum (Enloe et al. 2004b; Corke et al.
2007). The flow is forced using both quasi-steady and F*=1 forcing. The quasi-steady
forcing has a slightly detrimental effect on the CP distribution. Forcing at F*=1 still
generates some lift enhancement although the effect is decreased in comparison to the
normal orientation. This may seem like a surprising result, but given the control
mechanism, it is explainable. Recall that the effect of F*=1 forcing is to amplify the
natural instability rather than simply energize the boundary layer. Thus, the introduction
of any perturbation should be able to partially accomplish this. This is clear from Figure
6.19 although it is apparent that the downstream orientation is more effective. This
reverse orientation even creates similar dynamic features in POD results (Figure 6.20).
While in this case, the downstream orientation is most effective, the influence of the
direction of DBD induced forcing is an open question. Arrangements that generate
momentum over 180 degrees can easily be constructed allowing detailed studies of this
effect not easily obtainable with typical ZNMF actuators (Porter et al. 2009).
151
Figure 6.19: Effect of reversing AC-DBD momentum direction on CP.
152
Figure 6.20: First four normalized baseline and controlled (AM) POD modes of the v component of velocity, ( )1 4 y−
∗ϕ , for Re=240k, α=0o and δf=30o with reverse actuator at x/c=0.77.
Baseline
FL+=12.7
Mode 1
FL+=12.7, F*=1
FL+=12.7, F*=2
Mode 2 Mode 3 Mode 4
153
6.4 Single NS-DBD Separation Control Results
In an attempt to extend control authority to higher Re, the effect of DBD plasma
actuators driven by repetitive nanosecond pulses is investigated. These devices have
demonstrated control authority up to Mach 0.74 in isolated testing for airfoil LE
separation control (Roupassov et al. 2009). An in-house developed pulser is used to test
the efficacy of this type of plasma actuator for controlling TE airfoil separation. In this
work, the construction of this device is identical to AC driven plasma actuators which
offer a good benchmark comparison especially in this flow field where effective control
parameters like frequency and actuator location have already been established.
Amplification of natural flow instabilities over a deflected flap has been identified
as an effective control strategy for the airfoil TE system. Control using NS-DBD plasma
has been investigated similarly. The characterization results shown previously suggest
that velocities generated by the device are too weak to have an effect especially when
compared to AC-DBD plasma. Instead, manipulation of the natural instability is proposed
to be accomplished using thermal perturbations. Results of multiple attempts at TE
separation control using this device are shown in Figure 6.21. The plasma is generated
using a 2 kHz carrier similar to the AC-DBD case and burst modulated using dc=50% at
F*=1. Pulses at fTE alone (FL+=0.4) are also tested. This is a representative example of
multiple experiments, none of which have yet to demonstrate that the NS-DBD plasma
possesses control authority at the TE. This is an unexpected result given the efficacy of
the NS-DBD plasma for LE separation control at high speeds in the literature (Roupassov
et al. 2009) and in-house (Chapter 7). Increasing the burst frequency in an effort to
154
increase the amount of energy deposited into the flow at the optimal forcing frequency is
also not successful (Figure 6.22). This is likely due to the reasons outlined in Section 4.4
where it is postulated that bursts of pulses should be applied at frequencies greater than
800 kHz to have a cumulative effect on the thermal mechanism. These results will be
further discussed in comparison to contrasting results for LE separation control in
Chapter 9.
Figure 6.21: NS-DBD effect on CP for Re=240k, α=0 and δf=30.
155
Figure 6.22: NS-DBD effect on CP for Re=240k, α=0 and δf=30 with increased energy
input.
156
Chapter 7: Leading Edge Separation Control
ZNMF periodic excitation is widely established as an effective forcing
mechanism for separation control in many aerodynamic systems. This has been
extensively explored for LE stall control on airfoils due to drastic lift and drag
performance gains realized by reattaching these massively separated flows. Initial studies
of AC-DBD plasma actuators also focused on this system (Roth et al. 2000; Post and
Corke 2004) and it has remained widely popular for demonstrating AC-DBD plasma
control efficacy. The authority of AC-DBD plasma for controlling LE separation has
been mostly limited to freestream velocities below ~30 m/sec although some claims of
control at higher speeds have been published and more are certainly to follow given the
new optimization techniques being suggested (Thomas et al. 2009).
The control of LE stall with AC-DBD plasma at relatively low speed is a mature
research area and further work is not compelling unless control authority is extended to
higher speeds. There is nothing radically new or different about the AC-DBD plasma
used in this work and similar performance as seen in the literature is expected when
applied to the LE of the OSU version of the simplified NASA EET airfoil. Consequently,
a detailed study of this type of actuation is not employed on the LE. Rather, the AC-DBD
plasma actuator is used as a benchmark for the performance of the NS-DBD plasma
actuator. The latter has shown control authority on an airfoil LE up to Mach 0.74 in
157
isolated testing (Roupassov et al. 2009), but both flow physics and control using this
actuator are not well-understood.
7.1 Single DBD Plasma Actuator Control Results: Downstream Orientation
The NS-DBD plasma actuator as described in section 4.3 is mounted at the LE of
the airfoil near x/c=0 similar to Figure 7.1. The OSU airfoil requires a boundary layer trip
to perform similarly to the simplified NASA EET. When the NS-DBD plasma actuator is
mounted at the LE near x/c=0, the passive (plasma off) presence of the device can
function in this fashion. An example of this behavior is shown in Figure 7.2. The NASA
airfoil reaches CLmax at 12o with a sharp drop-off at higher α indicating stall. The OSU
airfoil with no LE passive actuator has a different pre-stall dCL/dα. The stall
characteristics of the OSU airfoil are also quite different with a gradual saturation of CL
and no indication of a sharp drop like the NASA case. The passive plasma actuator
curves are acquired by creating holes in the actuator materials such that CP and
subsequent CL measurements can be made. When these holes exist, the actuator cannot be
used for plasma generation because arcs form between the high voltage and ground
electrodes. The addition of the actuator to the LE rectifies the dCL/dα discrepancy and a
more uniform CL offset of the OSU model is now apparent. The additional curve in this
case shows the result of overshooting the target Re then gradually reducing the freestream
velocity back to the target value to take advantage of hysteresis effects often encountered
in separation control (Greenblatt and Wygnanski 2000). This overshoot allows extension
of the stall angle to more closely match NASA results by forcing the boundary layer at
the LE to undergo turbulent transition farther upstream. In doing so, the flow is now able
158
to negotiate the adverse pressure gradient encountered over the model surface. Note that
the post-stall characteristics of the OSU airfoil differ from NASA until α=16o where all
cases now agree. The majority of testing with actuators mounted near x/c=0 is performed
here.
Figure 7.1: DBD plasma actuator mounted at the LE of the OSU simplified NASA EET airfoil.
Exposed HV electrode
Dielectric extent
159
Figure 7.2: CL curves for the OSU airfoil with and without passive LE actuator and Re overshoot compared to NASA (Melton et al. 2006).
The stall characteristics of the OSU version of the NASA EET are very sensitive
to LE geometry due to a small radius of curvature in this region. Slight changes to the LE
boundary conditions caused by the application of actuators can change the stall
characteristics considerably. This is particularly frustrating when using DBD plasma
actuators composed of adhesive tape that must be applied to the airfoil surface by hand.
The actuator inevitably fails by arcing between the HV and ground electrodes requiring
replacement. Application of a new actuator to the LE can result in a different post-stall
CP distribution. These changes in the CP distribution can also occur within a given run
due to slight variations in the freestream velocity, turbulence intensity and LE boundary
160
conditions. Figure 7.3 gives an example of the various types of baseline behavior at post
stall with an actuator at the LE for Re=750k and α=16o. As in Chapter 6, baseline refers
to the condition when an actuator is applied to the LE without plasma (power off). Note
that static taps at x/c<0.05 are covered by the actuator and preclude measurements at
these points. In some cases, the post-stall baseline can exhibit a flat, zero pressure
gradient on the suction surface, while other cases can exhibit closed separation bubbles of
various size at various locations in the range x/c=0 to x/c=0.4. The time scale of this
changing baseline state can be as long as five minutes. The imprecise nature of applying
adhesive layers of tape to the leading edge by hand makes this a very time consuming
problem. In addition, hysteresis and Re effects are common, most notably in the α=10-12o
range.
161
Figure 7.3: Various examples of baseline (plasma off) behavior observed when an actuator is applied to the LE.
The difficulty in reproducing baseline data has a strong effect on control
authority. When post-stall CP characteristics exhibit a near-zero pressure gradient on the
suction surface, NS-DBD plasma control authority is quite effective. However, when
closed separation bubbles of various size and location exist, the flow is much more
difficult to affect. This is due to variations of the separation characteristics near the LE.
The zero pressure gradient case indicates stall with separation near the LE which is near-
optimal for control with this actuator location. The nonzero pressure gradient on the
suction surface is created by post-stall closed separation bubbles and variable separation
location. This introduces a complicating factor in the analysis. To remove this variable,
the focus of analysis for actuators located near x/c=0 is on post-stall baselines that exhibit
162
a near zero pressure gradient along the suction surface. For completeness, the lower α
cases are discussed briefly.
At pre-stall α (10-12o), both AC and NS-DBD plasma actuation at 1 kHz can
substantially improve the CP distribution at Re=750k and 1000k as seen in Figures 7.4
and 7.5. The Re=750k cases also include CP profiles with (Passive Actuator) and without
(No Actuator) a passive actuator with holes for comparison (Figure 7.4a and 7.5a). The
No Actuator cases do not exhibit a zero pressure gradient on the suction surface. The
passive actuator alone can substantially improve the CP distribution in the α=12o case as
suction increases and the closed separation bubble near x/c=0.1 is weakened
considerably. In all cases, the DBD plasma actuator seems to function as an active trip
that is a substantial improvement over passive presence of the device alone. In these
cases, the flow remains attached after turning off both plasma discharges (not shown).
This indicates incipient separation at these nominally pre-stall angles of attack where a
well-designed boundary layer trip can and should result in attached flow for this airfoil
shape.
163
Figure 7.4: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at α=10o.
a)
b)
164
Figure 7.5: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at α=12o.
a)
b)
165
At α=14o in Figure 7.6, the NS-DBD plasma actuator consistently outperforms the
AC-DBD discharge by attaching the normally separated flow at both Re although some
isolated cases do not follow this trend. The CP curves in Figure 7.6 contain data from
various tests using different actuators applied to the same location on the airfoil LE.
Thus, there are repeated data series in some cases. This is done to highlight the variable
performance and complication that arises from applying the DBD very near the LE. For
example, Figure 7.6a shows two data series for NS-DBD forcing at 2 kHz which show
considerably different levels of control authority. Similarly, in Figure 7.6b, AC-DBD
forcing at 2 kHz has variable performance. This phenomenon will be discussed
throughout this section. However, the data in Figure 7.6 begin to show more reliable
performance for the NS-DBD plasma actuator.
166
Figure 7.6: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at α=14o.
a)
b)
167
Figure 7.7 shows additional examples of the variable performance encountered
for both the NS and AC-DBD plasma at the airfoil LE at α=16o for Re=750k and 1000k.
As in Figure 7.6, the plots are created from multiple data sets with different actuators
placed at x/c=0. In the low Re case, both AC and NS-DBD plasma can reattach the
normally separated flow as evidenced by the suction peak created near the LE, but some
isolated cases do not follow this trend. In the Re=750k case, the 2 kHz NS-DBD performs
poorly during one test and the AC-DBD at 1 kHz performs well. At Re=1000k, the NS-
DBD at 1 kHz has some variable performance and little control authority is demonstrated
by the AC-DBD plasma. Excitation using 2 kHz AC-DBD plasma is not effective for
Re=1000k in any of the surveyed cases. This result is not presented in Figure 7.7. The
lack of control authority for AC-DBD plasma at higher Re and freestream velocity is
expected.
168
Figure 7.7: Effect of DBD plasma at x/c=0 on CP for Re=750k (a) and Re=1000k (b) at α=16o.
a)
b)
169
The preceding results have been shown in the interest of full disclosure of the
variable performance associated with using both AC and NS-DBD plasma actuators near
x/c=0. Because of the sensitivity of the developing boundary layer to LE boundary
conditions, only results that have been repeated at least twice are used for analysis.
Figure 7.8 shows CP profiles developed on the airfoil for forcing at various
frequencies. In this case, higher frequencies appear to generate the greatest effect with
control authority progressively increasing up to 1 kHz where saturation ensues. This
result is consistent with measurements of NS-DBD applied to a NACA 0015 airfoil
(Roupassov et al. 2009), but is somewhat contrary to the majority of literature that shows
a preferred frequency should exist for controlling airfoil LE separation at post-stall
(Seifert and Tillman 2009).
170
Figure 7.8: Effect of NS-DBD plasma actuation at x/c=0 on CP for Re=750k at α=16o showing preference for higher frequency forcing.
Figure 7.9 shows another frequency investigation of both the AC and NS-DBD
plasma at the airfoil LE. The frequency sweep is actually more detailed than shown, but
has been down-sampled for clarity. In this case, a repeatable baseline condition with zero
pressure gradient stall on the suction surface is established and detailed frequency sweeps
are performed using both AC and NS-DBD plasma by merely changing the excitation
input to the actuator. As such, this figure represents a direct comparison of the two types
of discharges for controlling flow separation. Low frequency AC signals are created
using a 2 kHz carrier frequency and modulated by the low frequency indicated in the
figure legend using 25% dc. Both the NS and AC DBD plasma produce the greatest
effect at frequencies in the range 600-800 Hz. Note that 2 kHz forcing does not produce a
171
significant change in the Cp distribution in these results. This is in contrast to Figure 7.9
and again highlights the sensitivity of the system to LE boundary conditions for actuators
placed at x/c=0.
172
Figure 7.9: Comparison of the effects of NS (a) and AC (b) DBD plasma on CP for Re=750k.
a)
b)
173
Clearly, the NS-DBD outperforms the AC-DBD in this case as a significantly
stronger suction peak is visible Figure 7.9. The results of Figure 7.9 are more succinctly
summarized in Figure 7.10 where the full frequency data set is plotted. The ordinate is
created by plotting the change in CP value on the suction side closest to the LE
(x/c=0.05). The lack of CP resolution at the LE prevents the accurate calculation of CL,
but the minimum CP value should correlate quite well. Thus, these results can be
interpreted as a scaled measure of CL. The results are reported in dimensionless form
(Fc+) based on the airfoil chord (25.4 cm (10 in)). The results clearly show the NS-DBD
outperforms the AC discharge in this case and both have a preferred frequency for control
around Fc+=3-4. Also note that the performance of this particular actuator is not as
impressive as some previously shown as CP nearest the LE is 2.5 compared to what
would be near 3.0 if a similar metric is employed in Figure 7.8. Despite this, the NS-DBD
appears superior to the AC-DBD for actuators operating at similar energy and power
levels as the discharge power for the peak minimum CP values is created using between
10-12 W in each case corresponding to CE=~0.08%.
174
Figure 7.10: Effect of Fc+ of AC and NS-DBD plasma at x/c=0 on suction side CP at
x/c=0.05.
7.2 Single DBD Plasma Actuator Control Results: Upstream Orientation
The previous results have shown the capability and at times superiority of NS
over AC-DBD plasma for controlling LE separation with actuators mounted near x/c=0.
However, the variable CP response with respect to actuation frequency is troubling and
also exists in archived literature (Roupassov et al. 2009). To understand this behavior
more fully, a baseline repeatability study was performed. In the course of this work,
results showed that placement of a surface discontinuity due to tape layers near x/c=0
created variable baseline and controlled flow behavior. The sensitivity of the flow to this
discontinuity can be explained by considering the boundary layer developing on the
175
airfoil LE. The boundary layer begins to develop at the stagnation point which for airfoils
at high angle of attack is on the pressure side of the model indicated by CP=1. The
boundary layer always begins in a laminar state and over some distance transitions to
turbulent depending on Re, surface roughness, freestream turbulence and the pressure
gradient developed by flow over the model. When this developing boundary layer is in a
laminar or transition state, it is very sensitive to small changes such as those encountered
by the discontinuity created on both sides of the HV electrode. To make the LE boundary
layer less sensitive to these effects, various tripping attempts both upstream and on the
HV electrode were attempted. If the transition to turbulence could be accelerated, it is
suspected that more repeatable performance could be established. These attempts were
met with little success due to the small scale of the model. Instead, a study on HV
electrode location was employed based on the assumption that discontinuities created by
this element caused the majority of repeatability issues. After significant effort,
repeatable baseline and control results were established by moving the interface of the
HV and ground electrodes downstream a distance of 6 mm in arc length, ds, and
reorienting the discharge such that plasma is formed on the upstream side of the high
voltage electrode. To further remove additional surface discontinuities near x/c=0, wide
sections of Kapton tape (5 cm (2 in)) were used as the dielectric and wrapped further
around the model LE as shown in Figure 7.11. Note that (10 cm (4 in)) wide dielectrics
were also tested, but these performed similarly to the more narrow variety. The reverse
actuator orientation is used to remove discontinuities, but also to keep the actuator
interface region which is believed to be most important for separation control as near as
176
possible to the LE in an effort to remain near or slightly upstream of the separation line.
This is another advantage of the NS-DBD. It is postulated that, unlike momentum
producing devices, the direction of actuator layout is not important for control authority
using thermal effects. This reverse arrangement has been previously explored (Roupassov
et al. 2009). Note that in this arrangement, a direct comparison of NS and AC-DBD
plasma is not as appropriate since the AC induced flow counters the freestream velocity
and has not been as widely examined in the literature. However, limited testing (Figure
7.12) shows, to some degree, similar AC-DBD performance as in the normal arrangement
(Figure 7.9b).
Figure 7.11: Photograph of the DBD plasma actuator mounted near the airfoil LE in reverse arrangement ds=6 mm from x/c=0.
Exposed HV electrode
Dielectric extent
177
Figure 7.12: Effect of AC and NS-DBD plasma for reverse actuator arrangement ds=6 mm from x/c=0.
With the newly developed baseline and control repeatability, detailed flow
diagnostics can now be employed. Minimum CP values developed on the model as a
function of Fc+ for various α are shown in Figure 7.13. At nominally pre-stall α (10-12o),
no frequency preference is visible and once a threshold level is reached, control authority
remains relatively constant. In the post-stall region beginning at α=14o, some optimal
frequencies become apparent around Fc+=4. As α is further increased, the preferred
frequency increases up to a maximum of around 5.5 at α =16o while the CP value tends to
increase. At α =18o, control authority is quite weak. From this information, it is apparent
that control authority at α>15o has been reduced by using the repeatable actuator
178
arrangement. This is an expected result since the actuator interface is now further
downstream (ds=6 mm). From this data, it appears this is near the optimal actuator
location for controlling flow over the airfoil at α < 14o. As α is progressively increased,
the separation line moves upstream and the actuator is now downstream of the optimal
location for producing control authority.
Figure 7.13: Effect of Fc+ of NS-DBD plasma for reverse actuator arrangement ds=6 mm
from x/c=0 on suction side CP at x/c=0.05 at Re=750k.
Despite this reduced control authority at high α, some significant flow physics are
revealed. Figure 7.14 shows a sample of the results of Figure 7.13, but now adds similar
curves for Re=1000k (62 m/sec). It is clear that control authority does not suffer due to
179
the increased freestream velocity and in the two lower α cases, it is actually augmented
considerably. This is even more promising when considering that q∞, which may be a
more reliable scaling parameter than velocity only, increases in proportion to U∞2. Note
also that Fc+ scaling holds quite well as the frequency preference for all three α is quite
similar. It is also clear that there is some effect on the flow for all Fc+ as the CP value has
been reduced by at least 0.5 in all cases. With this in mind, forcing in the various
frequency regions is explored using PIV.
Figure 7.14: Effect of Fc+ of NS-DBD plasma at for reverse actuator arrangement ds=6
mm from x/c=0 on suction side CP at x/c=0.05 at Re=750k and Re=1000k.
180
Figure 7.15a shows the NS-DBD plasma actuator effect on CP for α=12o. This
coincides with CL,max for the NASA airfoil and for the OSU model when a properly
designed and placed trip can effectively attach the flow (Figure 7.2). The forcing
frequency shown in Figure 7.15 corresponds to a condition where the threshold frequency
has been reached and further increases to do not result in significant CP improvement.
The baseline flow is characterized by a near zero pressure gradient along the suction
surface indicating separated flow. This is apparent in the time-averaged vorticity field in
Figure 7.15b as a strong region of vorticity signifies the boundary between the freestream
and separated region over the model. Control at Fc+=4 reattaches the normally separated
flow creating a strong suction peak at the LE. The time-averaged vorticity field of Figure
7.15c shows stronger levels of vorticity near the LE indicating a significant effect of the
actuator on the flow in this region. It can also be seen that the flow then separates from
the TE near x/c=0.8. Figure 7.16 shows phase-locked normalized velocity fluctuations,
* /v v U∞= , over the airfoil calculated according to:
v V V= − 7.1
where V and V denote phase and time-averaged quantities respectively.
The figure in question is one representative sample of four phases of the actuation
period. None of the phase-averaged images show any recognizable dynamic content.
Thus, reattachment of the nominally separated flow does not appear to occur through
excitation of large coherent structures. Instead, the flow has become reattached through
the use of the actuator as an active tripping device. This is supported by both NASA and
OSU results passive control results.
181
Figure 7.15: Time-averaged behavior of CP (a) and normalized vorticity,
*Ω , for baseline
(b) and NS-DBD forcing at Fc+=4 (c) at Re=750k, α=12o.
a)
b)
c)
182
Figure 7.16: Phase-averaged normalized velocity fluctuations, *v , for NS-DBD plasma forcing at Fc
+=4 (c) at Re=750k, α=12o.
Figure 7.17a shows the baseline and controlled CP distributions for α=14o. In this
case, the baseline is characterized by an even more zero pressure gradient stall along the
suction surface with control being quite effective for reattaching the separated flow along
nearly the entire suction surface. The time-averaged normalized vorticity field clearly
shows this effect as the region of strong vorticity bounding the freestream and separated
region is significantly more intense than the same figure for the α=12o case (Figure
7.15b). Control moves the region of strong vorticity near the LE further upstream. This
strong vorticity extends downstream to approximately x/c=0.2 after which a more diffuse
region is observed (Figure 7.17c). The controlled vorticity in this case differs
substantially from the α=12o case.
183
Figure 7.17: Time-averaged behavior of CP (a) and normalized vorticity,
*Ω , for baseline
(b) and NS-DBD forcing at Fc+=4 (c) at Re=750k, α=14o.
a)
b)
c)
184
The nature of flow structures in of Figure 7.18a gives the most striking
description of the controlled flow field. Clearly, coherent structures dominate the flow
and the number of these structures is consistent with forcing at Fc+=4. The effectiveness
of such dynamic structures for entraining freestream momentum and reattaching the
nominally separated flow of the airfoil is well established (Darabi and Wygnanski 2004).
Only one phase of the actuation signal is shown, but remaining phases confirm the
propagation of these organized regions along the airfoil chord. Further downstream, the
organized nature of these structures is broken. A pair of these positive and negative
regions corresponds to the outer edges of a clockwise rotating structure. However, from
Figure 7.18a one cannot discern which pair of these regions correspond to a vortex.
Normalized swirling strength, λci*=λcic/U∞, which separates regions of rotation
from pure shear is used to further identify this behavior. This technique is based on
critical point analysis of the local velocity gradient tensor and its eigenvalues (Adrian et
al. 2000). In two-dimensional form, the velocity gradient tensor is:
U Ux y
WV Vx y
∂ ∂ ∂ ∂ ∇ = ∂ ∂ ∂ ∂
7.2
where in this case the gradient has been applied to the phase-averaged vector field, W
.
The parameter of interest is the imaginary component of the eigenvalues of Equation 7.2
which is nonzero only if:
185
2 21 1 02 4
U V U V U Vy x x y x y
∂ ∂ ∂ ∂ ∂ ∂ − + + < ∂ ∂ ∂ ∂ ∂ ∂ 7.3
The imaginary component of the eigenvalues of Equation 7.2, λci, is calculated using a 2nd
order accurate central difference scheme and the normalized results are shown in Figure
7.18b. It is now apparent that the regions on either side of x/c=0.15 form a coherent
structure. The same can be said for x/c=0.35 and 0.6. The more commonly used vorticity
field is shown in Figure 7.18c. By definition this term also includes components of pure
shear, which in this flow masks the location of vortex cores. Vorticity is well-suited for
identifying vortex shedding from the airfoil trailing edge since vorticity of different signs
is alternatively shed form the flap similar to the Von Karmen vortex sheet observed for
cylinders in cross flow (Figure 6.8). Clearly, it is not as useful for identifying coherent
structures over the airfoil as all vorticity is of the same sign.
186
Figure 7.18: Phase-averaged normalized velocity fluctuations, *v (a), swirling strength,
*ciλ , (b) and vorticity,
*Ω , (c) for NS-DBD forcing at Fc
+=4 for Re=750k, α=14o.
a)
b)
c)
187
The frequency preference indicated in Figure 7.14 for α=16o shows three
relatively distinct forcing regimes. The flow response will be analyzed in order of
increasing frequency. Low frequency forcing (Fc+=0.6) has an effect on the naturally
stalled flow as seen in the CP curves, but in this case the minimum value is not observed
nearest LE (Figure 7.19a). Rather it appears further downstream near x/c=0.15. The time-
averaged vorticity certainly shows this change in behavior as the very apparent shear
layer is diffused considerably (Figure 7.19b,c). As before, the dynamic content of the
flow is most compelling (Figure 7.20). In this low frequency case, the phase averaged
velocity fluctuations are massively amplified as they move over the model surface. The
vortex cores (x/c=0.2 and 0.8) are again identified using swirling strength. Note that these
structures do not follow the airfoil surface as closely as those shown in Figure 7.18 and
instead are swept downstream by freestream momentum rather quickly once they move
away from the LE. Also, the magnitude of the phase averaged velocity fluctuations is
approximately twice as large are those seen in Figure 7.18a as noted by the change in
color scale. These massive coherent structures have been identified for all the low
frequency forcing cases in Figure 7.14. The vortex shedding associated with this low
frequency forcing has a rather strong audible signature that is quite obvious during wind
tunnel operation.
188
Figure 7.19: Time-averaged behavior of CP (a) and normalized vorticity,
*Ω , for baseline
(b) and NS-DBD forcing at Fc+=0.6 (c) at Re=750k, α=16o.
a)
b)
c)
189
Figure 7.20: Phase-averaged normalized velocity fluctuations, *v (a) and swirling strength, *
ciλ , (b) for NS-DBD forcing at Fc+=0.6 for Re=750k, α=16o.
Optimal forcing (Fc+=5.6) at α=16o produces reattachment at the LE and over
approximately half the model chord (Figure 7.21). The control authority in this case is
diminished compared to optimal forcing in the α=14o case (Figure 7.17a) as the suction
peak only reaches CP=-3. This is likely due to the movement of the separation line
upstream resulting in a less than optimized position for the plasma actuator which was
a)
b)
190
knowingly traded for repeatability. Forcing still has a significant effect on the time-
averaged vorticity field as the intense shear layer over the entire model is eliminated
(Figure 7.21b,c). As usual, notable effects of control can be seen in the dynamic features
of the separated flow (Figure 7.22). In this case, the coherent structures do not propagate
as far downstream as those at α=14o before dissipating. However, their organization near
the LE is quite apparent in both the phase averaged velocity fluctuations and the swirling
strength that shows a clear vortex core at x/c=0.15.
191
Figure 7.21: Time-averaged behavior of CP (a) and normalized vorticity,
*Ω , for baseline
(b) and NS-DBD forcing at Fc+=5.6 (c) at Re=750k, α=16o.
a)
b)
c)
192
Figure 7.22: Phase-averaged normalized velocity fluctuations, *v (a) and swirling strength, *
ciλ , (b) for NS-DBD forcing at Fc+=5.6 for Re=750k, α=16o.
Lastly the effect of high frequency forcing (Fc+=11.3) is investigated. Although
the LE CP value is similar to the low frequency forcing case (Figure 7.19), the location of
this minimum is further upstream signifying a very different flow field (Figure 7.23). As
in all cases, the time-averaged vorticity shows a significant effect, but merely breaking up
the intense layer of shear over the airfoil is not sufficient to substantially reduce
a)
b)
193
separation. The dynamic features are less apparent, but still exist near the LE (Figure
7.24). In this case, the color scale has been reduced considerably in an effort to bring out
the short wavelength low amplitude, but measureable phase averaged velocity
fluctuations generated by the high frequency forcing. This reduced color scale also
emphasizes the chaotic features of the dissipated region downstream. The robustness of
the swirling strength parameter is exhibited in Figure 7.24b. Despite the weak oscillation
observed in Figure 7.24a, as very distinct vortex core is identified at x/c=0.1.
194
Figure 7.23: Time-averaged behavior of CP (a) and normalized vorticity,
*Ω , for baseline
(b) and NS-DBD forcing at Fc+=11.3 (c) at Re=750k, α=16o.
a)
b)
c)
195
Figure 7.24: Phase-averaged normalized velocity fluctuations, *v (a) and swirling strength, *
ciλ , (b) for NS-DBD forcing at Fc+=11.3 for Re=750k, α=16o.
The preceding results for Re=750k have also been acquired at Re=1000k. As the
Fc+ in Figure 7.14 predicts, the flow fields are nearly identical in terms of control
authority, time-averaged and phase-averaged content. This is an encouraging result given
that this higher Re flow is in the upper range at which standard AC-DBD plasma
actuators lose control authority as documented in Figure 7.10. The efficiency at which the
a)
b)
196
NS-DBD plasma actuator excites coherent structures is particularly appealing. Although
optimal frequencies exist for control, it seems that the NS-DBD possesses sufficient high
amplitude characteristics to create structures at any frequencies in the range surveyed.
7.3 Comparison of Single NS-DBD Plasma Actuators to Passive Flow Control
The NS-DBD plasma controlled CL curve is shown in Figure 7.25 as a final
metric for demonstrating its capability. Recall that CP values are not measurable here due
to the actuator application. The CL values are estimated using the measured CP values and
linearly extrapolating the NS-DBD controlled data at the LE based on pre-stall values for
passive actuator with holes for CP measurements. This methodology is similar to other
simplified CL estimates (Janiszewska and Lee 2005), but also employs the measured data
away from the airfoil LE. The error bars on the figure are quite large due to the
uncertainty introduced by this extrapolation. As suggested by previous data, the NS-DBD
is capable of extending the stall angle to 16o and this effect is not diminished at
Re=1000k shown by the single data point. This figure is compared to a similar
experiment from NASA using a well-established PFC device in the form of a LE droop
(Figure 7.26) similar to the schematic in Figure 5.1. In terms of stall angle, the
performance rivals that of the NASA droop in these conditions. The actual CL values
between the two cases are considerably different as CL for NASA peaks near 2.0 for the
maximum droop angle (-30o) while the NS-DBD controlled CL is approximately 1.7.
These discrepancies can be explained. Baseline characterization showed that the OSU lift
curve is shifted by approximately -0.1 in comparison to NASA. Additionally, the
drastically different CP distribution created by the droop in the NASA case changes CL
197
considerably. For example, the NASA CL value at α=12o increases by approximately 0.2
when the maximum droop angle (-30o) is used. Thus, within the CL variation described
above the extrapolated values are comparable.
Figure 7.25: Effect of AFC with NS-DBD plasma actuation on CL for the OSU airfoil.
198
Figure 7.26: Effect of PFC with using LE droop on CL for the simplified NASA EET airfoil (Melton et al. 2005)
199
Chapter 8: Summary and Conclusions
Control of flow separation from a high-lift airfoil has been examined
experimentally using dielectric barrier discharge (DBD) plasma actuators driven by high
voltage AC and nanosecond pulse (NS) waveforms. Actuators are composed of copper
tape electrodes and Kapton tape dielectric arranged in an asymmetric fashion. Actuators
are adhered to the leading edge (LE) and trailing edge (TE) flap shoulder of a simplified
supercritical high-lift airfoil with chord length of 25.4 cm in an effort to reduce or
eliminate flow separation from these regions. Surveyed flow conditions include Reynolds
numbers in the range 240k-1000k (15-62 m/sec) for incidence angles between 0-18
degrees and flap deflections between 0-40 degrees.
Actuator characterization in quiescent air shows that very different behavior is
generated when creating plasma discharges with these two radically different waveforms.
The AC-DBD plasma actuator generates near wall jets with maximum mean velocities of
~3.5 m/sec approximately 1 mm from the surface when exciting using a 3 kHz waveform
at 20 kVpp. Modulating this high frequency carrier signal with a low frequency signal
(50-400 Hz) results in fuller mean velocity profiles with lower maximum values. The
unsteady character of the modulated signals creates a near surface vortex train in still air
whose strength varies depending on the character of the modulating waveform. The
electrohydrodynamic effects of the AC-DBD plasma actuator are quantified by
200
integrating the near-wall profiles and casting the result as a dimensionless momentum
coefficient commonly used to describe flow control actuator amplitude. This metric is
found to be only weakly correlated with the control authority of the device when operated
with different modulation waveforms highlighting the limitations of quiescent air
measurements for quantifying actuator performance.
The NS-DBD plasma actuator induced mean velocity is approximately five times
lower than the AC-DBD. The low momentum production of the AC-DBD and its limited
control authority coupled with this result suggest that velocity production of the NS-DBD
plasma actuator is not a control mechanism for the flow conditions surveyed in this work.
schlieren imaging of the discharge shows that a very complex compression wave
structure is generated by the NS-DBD thermal effects in quiescent conditions. A single
planar and multiple localized spherical shock waves are generated for each high voltage
pulse. The speed of these waves a few mm from the discharge is approximately sonic
(Mach 1), but compression wave strength near the surface may be substantially greater.
Simplified estimates predict near surface temperature rises of ~200 K and compression
wave strength of approximately Mach 1.1. A modified momentum coefficient is
suggested for thermal effects based on pressure difference across the near surface
compression wave. This new coefficient produces values similar to those expected for
more traditional fluidic actuation techniques that demonstrate similar airfoil LE control
authority. High spatial resolution near surface measurements of this behavior are required
to fully characterize the device as a flow control actuator.
201
The time-averaged power dissipated by the two DBD plasma actuators is
relatively small. The AC-DBD plasma actuator consumes ~26 W when operated at 1 kHz
and 18 kVpp on a 46 cm long DBD load while the NS-DBD consumes <17 W operating
at the same frequency and 12 kV. The instantaneous power of the NS-DBD can be as
high as 600 kW due to the very short pulse duration (~100 ns) which deposits ~10 mJ
into the load generating the strong thermal effect seen in the schlieren photography.
Control of turbulent boundary layer separation over a deflected TE flap is
demonstrated using a single AC-DBD plasma actuator placed near the flap shoulder.
Quasi-steady forcing at 3 kHz slightly delays separation and increases static pressure on
the flap due to a lengthening and flattening of the separated region, but does not have a
drastic effect on the measured lift. Amplifying the natural instability associated with
vortex shedding from the flap shoulder by forcing with low frequency modulated
versions of the high frequency waveform is most effective for increasing lift and reducing
the time-averaged recirculation region on the flap. This comes at the expense of increased
dynamic loading and more organized turbulent fluctuations in the wake. Amplification of
this natural instability creates more negative static pressure values on the flap, but the
majority of the lift increase is due to upstream effects from enhanced circulation around
the entire model. These results persist even when significant reattachment downstream of
the flap shoulder is not achieved. The lift increase due to amplification of the natural
instability is relatively independent of incidence angle as long as the separation location
and the underlying dynamics do not change. Calculations of sectional total drag from
wake surveys that include turbulent stresses suggest lift increase via instability
202
amplification is not prohibitive. The control authority of the actuator generally decreases
with increasing Reynolds number and flap deflection consistent with the limited nature of
AC-DBD plasma momentum addition.
Both the effectiveness and efficiency of the actuator for forcing at low frequencies
can be further enhanced using a square wave modulation signal (BM) instead of a
sinusoidal signal (AM). This is due to the on/off nature of the BM perturbation in
comparison to the more gradual sinusoidal envelope associated with the AM case. Each
flap deflection angle is found to have a preferred duty cycle range for BM that increases
with increasing deflection angle. Beyond this preferred duty cycle, control authority is
found to decrease considerably. The falloff at high dc is believed to be due to the zero net
mass flux (ZNMF) pulsed blowing nature of the AC-DBD plasma actuator and is in
contrast to existing data for a similar airfoil using more traditional piezoelectric ZNMF
excitation (Melton et al. 2004). The NS-DBD plasma actuator has shown no control
authority at the TE flap shoulder.
The findings, aside from the dc and NS-DBD results, are generally consistent with
a similar study by NASA using piezoelectric driven ZNMF actuators on larger scale
version of the same airfoil (Melton et al. 2004; Melton et al. 2006). However, actuator
characterization shows that values of momentum coefficient for DBD plasma on the
airfoil are generally an order of magnitude lower than those used at NASA yet control
authority, especially at low Reynolds number, is comparable. Two open questions then
remain: do further increases in momentum coefficient generate a similar trend as shown
by NASA and is DBD plasma more efficient from a momentum perspective in this flow
203
field as momentum coefficient appears to show? Definitive answers require future work,
but such studies are certainly warranted given recent advances in DBD plasma thrust
generation that appear to make their use in transport aircraft feasible (Thomas et al.
2009).
Control of separation from the LE of the same airfoil is investigated primarily
with NS-DBD plasma actuators. The AC-DBD plasma actuator is well-established for
controlling LE separation up to 30 m/sec and is used as a metric for evaluating the NS-
DBD performance (Moreau 2007). The actuator is mounted across the span of the airfoil
to produce nominally two-dimensional perturbations. Locating the actuator at the LE near
x/c=0 facing downstream is very effective for controlling separation, however
repeatability of the both baseline and control conditions is difficult due to the sensitivity
of the flow field to small changes in the LE boundary conditions caused by the passive
presence of the device. Despite this complexity, repeated tests indicate the NS-DBD
outperforms the AC-DBD plasma actuator used in this work for controlling LE separation
and demonstrates similar scaling with dimensionless frequency while increasing the stall
angle by six degrees at Reynolds numbers up to 1000k (62 m/sec). This rivals the
performance of a passive LE droop used on a similar airfoil by NASA.
To improve repeatability, the actuator is moved downstream by 6 mm in arc
length along the suction surface and reversed such that plasma is formed in the upstream
direction using the assumption that control via thermal effects is not strongly dependent
on forcing direction. The slight movement of the surface discontinuity created by the tape
layers is sufficient to produce both repeatable baseline and control results. The control
204
authority at high incidence (α=16o) in this configuration is slightly reduced in comparison
to the x/c=0 case. This is likely due to a less optimal location of the actuator. Detailed
diagnostics in this configuration reveal the NS-DBD plasma actuator can act as an active
trip for nominally pre-stall incidence angles for this airfoil (≤1 2o). At high incidence, the
NS-DBD is found to reattach the normally separated flow through excitation of coherent
spanwise vortices that entrain freestream momentum into the separated region. The
device is found to generate these structures across the frequency range surveyed, but the
optimal dimensionless frequency for controlling separation is in the range four to six
depending on the incidence angle. The control authority in this configuration is
independent of the Reynolds numbers surveyed and obeys dimensionless frequency
scaling.
The contrasting performance of the NS-DBD plasma actuator at the LE and TE in
comparison to the AC-DBD is the most compelling question left unanswered in this
work. It is postulated that the near surface behavior of DBD plasma actuators coupled
with the varying state of the boundary layer at LE and TE is a possible cause for this
inconsistency. High resolution near surface DBD measurements that include the effects
of cross-flow are required to make a substantive conclusion on this topic.
205
Chapter 9: Discussion of Future Work
The previous sections have demonstrated some contrasting effects that deserve
discussion. The AC-DBD plasma actuator is found to be superior to NS-DBD plasma for
controlling separation from a deflected TE flap in the Re range 240k-750k (15-45 m/s). In
direct contrast, the NS-DBD plasma actuator is found to be superior to AC-DBD plasma
for controlling separation from the airfoil LE in the Re range 750k-1000k (45-62 m/s).
The exact cause of this discrepancy is unknown at this time, but some possibilities can be
considered.
Let us first examine the systems under consideration. The boundary layer
approaching the deflected flap in TE separation control studies has been characterized as
turbulent with thickness of 4.5-7 mm depending on Re. Separation control at the LE is a
substantially easier problem in terms of momentum requirements due to the state of the
boundary layer here (Melton et al. 2006). For airfoils at high α, the boundary layer
develops from the stagnation point on the pressure side around the suction surface of the
airfoil. At high α when flow separates abruptly from the LE, it is difficult to define the
state of this boundary layer (laminar/transitional/turbulent) especially in small scale
laboratory applications. However, one can assume that it is substantially thinner than the
boundary layer measured in the TE case. Despite the different boundary layer
characteristics at the LE and TE flap shoulder, literature shows that similar dimensionless
206
frequency scaling applies to both if the separation length scales are properly identified
(Seifert et al. 1996).
Actuator characterization showed two fundamentally different mechanisms are
involved when generating plasma with these two radically different waveforms. The NS-
DBD showed relatively weak compression waves propagating away from the surface. It
is suspected that the compression wave strength is substantially greater very near the
surface but quickly weakens as it propagates away. This is complicated by the highly
three dimensional nature of the compression waves generated by the device. To quantify
these effects, high spatial resolution near surface measurements must be performed.
However, if this assumption is correct it may explain the superior performance of the NS-
DBD for controlling LE airfoil separation where a very thin boundary layer is present. If
this is the case, it would also suggest that the AC-DBD plasma, while not as strong near
the surface, may produce a slightly stronger effect than the NS-DBD away from the
surface. A more fundamental question is whether there is any difference in instability
response to thermal versus EHD generated perturbations.
Another possibility for the discrepancy in control authority between these two
devices is the streamwise distribution of the forcing effect. For example, it is assumed
that the asymmetric electrode interface is the most important region for control authority
in both cases. This assumption is based on modeling results for the AC-DBD plasma
actuator (Corke et al. 2007), but similar studies have not been employed for the NS-DBD.
The effects of freestream flow on the discharge characteristics have also not been fully
207
investigated. Thus, it is possible that the optimized actuator location may not be
equivalent for the two discharges.
Lastly, an obvious missing piece to this work is the combination of NS-DBD
forcing on the LE and AC-DBD forcing on the TE to even more fully enhance lift. While
time constraints did not permit the experiments, both existing literature and TE results
suggest that if flow can be attached up to the flap shoulder using NS-DBD, the additional
CL lift gains from TE should be conserved (Melton et al. 2006).
208
References
Abbot I, vanDoenhoff A (1959) Theory of Wing Sections. New York, NY: Dover
Abe T, Takizawa Y, Sato S, Kimura N (2008) Experimental study for momentum transfer in a dielectric barrier discharge plasma actuator. AIAA Journal 46[9]:2248-2256
Adelgren R, Yan H, Elliott G, Knight D, Zheltovodov A (2005) Control of Edney IV Interaction by Pulsed Laser Energy Deposition. AIAA Journal 43[2]:256-269
Adrian R, Christensen K, Liu Z (2000) Analysis and interpretation of instantaneous turbulent velocity fields. Experiments in Fluids 29:275-290
Amitay M, Glezer A (2002) Role of Actuation Frequency in Controlled Flow Reattachment over a Stalled Airfoil. AIAA Journal 40[2]:209-216
Balcon N, Benard N, Lagmich Y, Boeuf J, Touchard G, Moreau E (2009) Postive and Negative Sawtooth Signals Applied to a DBD Plasma Actuator - Influence on the Electric Wind. Journal of Electrostatics 67:140-145
Benard N, Balcon N, Moreau E (2009a) Electric Wind Produced by a Surface Dielectric Barrier Discharge Operating Over a Wide Range of Relative Humidity. AIAA Paper 2009-0488
Benard N, Jolibois J, Forte M, Touchard G, Moreau E (2007) Control of an Axisymmetric Subsonic Air Jet by Plasma Actuator. Experiments in Fluids 43[4]:603-616
Benard N, Jolibois J, Moreau E (2009b) Lift and Drag Performances of an Axisymmetric Airfoil Controlled by Plasma Actuator. Journal of Electrostatics 67:133-139
209
Benedict L, Gould R (1996) Towards better uncertainty estimates for tublence statistics. Experiments in Fluids 22:129-136
Boucinha V, Magnier P, Leroy-Chesneau A, Weber R, Joussot R, Dong B, Hong D (2008) Characterization of the Ionic Wind Induced by a Sine DBD Actuator used for Laminar-to-Turbulent Transition Delay. AIAA Paper 2008-4210
Cerchie D, Halfron E, Hammerich A, Han G, Taubert L, Trouve L, Varghese P, Wygnanski I (2006) Some Circulation and Separation Control Experiments. In: Applications of Circulation Control Technology. eds Joslin, R and Jones, G. Vol. 214, pp. 113-165, Reston, VA: AIAA Inc.
Chan S, Zhang X, Gabriel S (2007) Attenuation of Low-Speed Flow-Induced Cavity Tones Using Plasma Actuators. AIAA Journal 45[7]:1525-1538
Corke T, Enloe C, Wilkinson S (2010) Dielectric Barrier Discharge Actuators for Flow Control. Annual Review of Fluid Mechanics 42:505-529
Corke T, Post M, Orlov D (2007) SDBD Plasma Enhanced Aerodynamics: Concepts, Optimization and Applications. Progress in Aerospace Sciences 43:193-217
Corke T, Post M, Orlov D (2009) Single Dielectric Barrier Discharge Plasma Enhanced Aerodynamics: Physics, Modeling and Applications. Experiments in Fluids 46:1-26
Darabi A, Wygnanski I (2004) Active Management of Naturally Separated Flow over a Solid Surface. Part 1. The Forced Reattachment Process. Journal of Fluid Mechanics 510:105-129
DeSalvo M, Whalen E, Glezer A (2010) High-Lift Enhancement using Fluidic Actuation. AIAA Paper 2010-0863
ELD (2006) 24 inch High Speed Recirculating Wind Tunnel: Installation, Operation and Maintenance Instructions. Lake City, MN
ELD (2007) 10 inch Chord Two-dimensional Airfoil Model with Adjustable Flap: Installation, Operation and Maintenance Instructions. Lake City, MN
210
Enloe C, Font G, McLaughlin T, Orlov D (2008a) Surface Potential and Longitudinal Electric Field Measurements in the Aerodynamic Plasma Actuator. AIAA Journal 46[11]:2730-2740
Enloe C, McHarg M, McLaughlin T (2008b) Time-Correlated Force Production Measurements of the Dielectric Barrier Discharge Plasma Aerodynamic Actuator. Journal of Applied Physics 103[073302]:1-7
Enloe C, McLaughlin T, Font G, Baughn J (2006) Parameterization of Temporal Structure in the Single-Dielectric-Barrier Aerodynamic Plasma Actuator. AIAA Journal 44[6]:1127-1136
Enloe C, McLaughlin T, VanDyken R, Kachner K, Jumper E, Corke T (2004a) Mechanisms and Responses of a Single Dielectric Barrier Plasma Actuator: Plasma Morphology. AIAA Journal 42[3]:589-594
Enloe C, McLaughlin T, VanDyken R, Kachner K, Jumper E, Corke T, Post M, Haddad O (2004b) Mechanisms and Responses of a Single Dielectric Barrier Plasma Actuator: Geometric Effects. AIAA Journal 42[3]:595-604
Falkenstein Z, Coogan J (1997) Microdischarge Behaviour in the Silent Discharge of Nitrogen Oxygen and Water Air Mixtures. Journal of Physics D: Applied Physics 30:817-825
Font G, Enloe C, Newcomb J, Teague A, Vasso A, McLaughlin T (2010) Effects of Oxygen Content on the Behavior of the Dielectric Barrier Discharge Aerodynamic Plasma Actuator. AIAA Paper 2010-0545
Forte M, Jolibois J, Pons J, Moreau E, Touchard G, Cazalens M (2007) Optimization of a Dielectric Barrier Discharge Actuator by Stationary and Non-Stationary Measurements of the Induced Flow Velocity: Application to Airflow Control. Experiments in Fluids 43:917-928
Freeman A, Catrakis H (2008) Direct Reduction of Aero-Optical Aberrations by Large Structure Suppression Control in Turbulence. AIAA Journal 46[10]:2582-2590
211
Gad-el-Hak M, Bushnell D (1991) Separation Control: Review. Journal of Fluids Engineering 113:5-29
Gaitonde D (2009) Simulation of Supersonic Nozzle Flows with Plasma-based Control. AIAA Paper 2009-4187
Glezer A, Amitay M, Honohan A (2005) Aspects of Low- and High-Frequency Actuation for Aerodynamic Flow Control. AIAA Journal 43[7]:1501-1511
Goksel B, Greenblatt D, Rechenberg I, Nayeri C, Paschereit C (2006) Steady and Unsteady Plasma Wall Jets for Separation and Circulation Control. AIAA Paper 2006-3686
Golubovskii Y, Maiorov V, Behnke J, Behnke J (2002) Influence of Interaction between Charged Particles and Dielectric Surface over a Homogeneous Barrier Discharge in Nitrogen. Journal of Physcis D: Applied Physics 35:751-761
Greenblatt D (2007) Dual Location Separation Control on a Semispan Wing. AIAA Journal 45[8]:1848-1860
Greenblatt D, Goksel B, Rechenberg I, Schule C, Romann D, Paschereit C (2008) Dielectric Barrier Discharge Flow Control at Very Low Flight Reynolds Numbers. AIAA Journal 46[6]:1528-1541
Greenblatt D, Wygnanski I (2000) The Control of Flow Separation by Periodic Excitation. Progress in Aerospace Sciences 36:487-545
Greenblatt D, Wygnanski I (2003) Effect of Leading-Edge Curvature on Airfoil Separation Control. Journal of Aircraft 40[3]:473-481
He C, Corke T, Patel M (2009) Plasma Flaps and Slats: An Application of Weakly Ionized Plasma Actuators. Journal of Aircraft 46[3]:864-873
Hoener S, Borst H (1975) Fluid-Dynamic Lift. Brick Town, NJ: Hoerner Fluid Dynamics
212
Holmes P, Lumley J, Berkooz G (1996) Turbulence, Coherent Structures, Dynamical Systems and Symmetry. Cambridge, UK: Cambridge University Press
Hoskinson A, Hershkowitz N, Ashpis D (2008) Force measurements of single and double barrier DBD plasma actuators in quiescent air. Journal of Physics D: Applied Physics 41[245209]:1-9
Huang J, Corke T, Thomas F (2006) Plasma Actuators for Separation Control of Low-Pressure Turbine Blades. AIAA Journal 44[1]:51-57
Janiszewska J, Lee J (2005) A Simple Method for Determining Lift and Drag on a Wing. AIAA Paper 2005-1061
Jayaraman B, Shyy W (2008) Modeling of Dielectric Barrier Discharge-Induced Fluid Dynamics and Heat Transfer. Progress in Aerospace Sciences 44:139-191
Jolibois J, Forte M, Moreau E (2008) Application of an AC Barrier Discharge Actuator to Control Airflow Separation above a NACA 0015 Airfoil: Optimization of the Actuation Location Along the Chord. Journal of Electrostatics 66:496-503
Jukes T, Choi K, Johnson G, Scott S (2006) Characterization of Surface Plasma-Induced Wall Flows Through Velocity and Temperature Measurements. AIAA Journal 44[4]:764-771
Kiedaisch J, Demanett B, Reinhard P, Nagib H (2007) Active Flow Control for High Lift Airfoils: Dynamic Flap Actuation. AIAA Paper 2007-1120
Kiedaisch J, Nagib H, Demanett B (2006) Active Flow Control Applied to High-Lift Airfoils Utilizing Simple Flaps. AIAA Paper 2006-2856
Kim W, Do H, Mungal G, Cappelli M (2007) On the Role of Oxygen in Dielectric Barrier Discharge Actuation of Aerodynamic Flows. Applied Physics Letters 91[18]:181501
Kogelschatz U (2003) Dielectric-barrier Discharges: Their History, Discharge Physics, and Industrial Applications. Plasma Chemistry and Plasma Processing 23[1]:1-46
213
Lavision (2007) FlowMaster: Product Manual.
Likhanskii A, Semak V, Schneider M, Opaits D, Miles R, Macheret S (2009) The Role of the Photoionization in the Numerical Modeling of the DBD Plasma Actuator. AIAA Paper 2009-0841
Lin J, Dominik C (1997) Parametric Investigation of a High-Lift Airfoil at High Reynolds Numbers. Journal of Aircraft 34[4]:485-491
Lu B, Bragg M (2003) Airfoil Drag Measurement with Simulated Leading-Edge Ice using the Wake Survey Method. AIAA Paper 2003-1094
Mabe J, Calkins F, Wesley B, Woszidlo R, Taubert L, Wygnanski I (2009) Single Dielectric Barrier Discharge Plasma Actuators for Improved Airfoil Performance. Journal of Aircraft 46[3]:847-855
McLean J, Crouch J, Stoner R, Sakurai S, Seidel G, Feifel W, Rush H (1999) Study of the Application of Separation Control by Unsteady Excitation to Civil Transport Aircraft. NASA CR 209338
Melton L (2006) Private Communication
Melton L, Schaeffler N, Lin J (2007) High-Lift System for a Supercritical Airfoil: Simplified by Active Flow Control. AIAA Paper 2007-0707
Melton L, Schaeffler N, Yao C-S, Seifert A (2005) Active Control of Flow Separation from Supercritical Airfoil Leading-Edge Flap Shoulder. Journal of Aircraft 42[5]:1142-1149
Melton L, Yao C-S, Seifert A (2003) Active Control of Separation from the Flap of a Supercritical Airfoil. AIAA Paper 2003-4005
Melton L, Yao C-S, Seifert A (2004) Application of Excitation from Multiple Locations on a Simplified High-Lift System. AIAA Paper 2004-2324
214
Melton L, Yao C-S, Seifert A (2006) Active Control of Separation from the Flap of a Supercritical Airfoil. AIAA Journal 44[1]:34-41
Moreau E (2007) Airflow Control by Non-Thermal Plasma Actuators. Journal of Physics D: Applied Physics 40:605-636
Nagib H, Kiedaisch J, Reinhard P, Demanett B (2006) Control Techniques for Flows with Large Separated Regions: A New Look at Scaling Parameters. AIAA Paper 2006-2857
Nagib H, Kiedaisch J, Reinhard P, Demanett B (2007) Active Flow Control for High Lift Airfoils: Separation versus Circulation Control. AIAA Paper 2007-1119
Opaits D, Neretti G, Likhanskii A, Zaidi S, Shneider M, Miles R, Macheret S (2007) Experimental Investigation of DBD Plasma Actuators Driven by Repetitive High Voltage Nanosecond Pulses with DC or Low-Frequency Sinusoidal Bias. AIAA Paper 2007-4532
Opaits D, Neretti G, Zaidi S, Shneider M, Miles R (2008) DBD Plasma Actuators Driven by a Combination of Low Frequency Bias Voltage and Nanosecond Pulses. AIAA Paper 2008-1372
Opaits D, Shneider M, Miles R (2009a) Electrodynamic Effects in Nanosecond-Pulse-Sustained Long Dielectric-Barrier-Discharge Plasma Actuators. Applied Physics Letters 94[061503]:3
Opaits D, Zaidi S, Shneider M, Miles R (2009b) Improving Thrust by Suppressing Charge Build-up in Pulsed DBD Plasma Actuators. AIAA Paper 2009-0487
Pack LG, Schaeffler NW, Yao. CS, Seifert A (2002) Active Control of Separation from the Slat Shoulder of a Supercritical Airfoil. AIAA Paper 2002-3156
Palm P, Meyer R, Plonjes E, Rich J, Adamovich I (2003) Nonequilibrium Radio Frequency Discharge Plasma Effect on Conical Shock Wave: M=2.5 Flow. AIAA Journal 41[3]:465-469
215
Patel M, Ng T, Vasudevan S, Corke T, Post M, McLaughlin T, Suchomel C (2008) Scaling Effects of an Aerodynamic Plasma Actuator. Journal of Aircraft 45[1]:223-236
Patel M, Sowle Z, Corke T, He C (2007) Autonomous Sensing and Control of Wing Stall Using a Smart Plasma Slat. Journal of Aircraft 44[2]:516-527
Poggie J, Tillman C, Flick P, Silkey J, Osborne B, Ervin G, Maric D, Mangalam S, Mangalam A (2010) Closed-Loop Stall Control on a Morphing Airfoil using Hot-Film Sensors and DBD Actuators. AIAA Paper 2010-0547
Pons J, Moreau E, Touchard G (2005) Asymmetric surface dielectric barrier discharge in air at atmospheric pressure: electrical properties and induced airflow characteristics. Journal of Physics D: Applied Physics 38:3635-3642
Porter C, Abbas A, Cohen K, McLaughlin T, Enloe CL (2009) Spatially Distributed Forcing and Jet Vectoring with a Plasma Actuator. AIAA Journal 47[6]:1368-1378
Porter C, Baughn J, McLaughlin T, Enloe C, Font G (2007) Plasma Actuator Force Measurements. AIAA Journal 45[7]:1562-1570
Post M (2004) Plasma Actuators for Separation Control on Stationary and Oscillating Airfoils. Dissertation, University of Notre Dame
Post M, Corke T (2004) Separation Control on High Angle of Attack Airfoil Using Plasma Actuators. AIAA Journal 42[11]:2177-2184
Rae W, Pope A (1984) Low Speed Wind Tunnel Testing. 2nd edn New York: John Wiley & Sons, Inc.
Rizzetta D, Visbal M (2009) Large Eddy Simulation of Plasma-Based Control Strategies for Bluff Body Flow. AIAA Journal 47[3]:717-729
Roth J, Dai X (2006) Optimization of the Aerodynamic Plasma Actuator as an Electrohydrodynamic (EHD) Electrical Device. AIAA Paper 2006-1203
216
Roth J, Sherman D, Wilkinson S (2000) Electrohydrodynamic Flow Control with a Glow-Discharge Surface Plasma. AIAA Journal 38[7]:1166-1172
Roupassov D, Nikipelov A, Nudnova M, Starikovskii A (2009) Flow Separation Control by Plasma Actuator with Nanosecond Pulsed-Periodic Discharge. AIAA Journal 47[1]:168-185
Samimy M, Kim J-H, Kastner J, Adamovich I, Utkin Y (2007) Active Control of High-Speed and High-Reynolds-Number Jets using Plasma Actuators. Journal of Fluid Mechanics 578:305-330
Samimy M, Kim J, Kearney-Fischer M, Sinha A (2010) Acoustic and Flow Fields of an Excited High Reynolds Number Axisymmetric Perfectly-Expanded Mach 1.3 Jet. Journal of Fluid Mechanics accepted for publication
Seifert A, Darabi A, Wygnanski I (1996) Delay of Airfoil Stall by Periodic Excitation. Journal of Aircraft 33[4]:691-698
Seifert A, Pack L (2002) Active Flow Separation Control on Wall-Mounted Hump at High Reynolds Numbers. AIAA Journal 40[7]:1363-1372
Seifert A, Tillman C (2009) Fixed Wing Airfoil Applications. In: Fundamentals and Applications of Modern Flow Control. eds Joslin, R and Miller, D. Vol. 231, pp. 231-257, Reston, VA: AIAA Inc.
Sidorenko A, Budovsky A, Pushkarev A, Maslov A (2008) Flight Testing of DBD Plasma Separation Control System. AIAA Paper 2008-0373
Singh K, Roy S (2008) Force approximation for a plasma actuator operating in atmospheric air. Journal of Applied Physics 103[013305]:1-6
Smith D, Dickey E, Klein TV (2006) The ADVINT Program. AIAA Paper 2006-2854
Sosa R, Artana G, Moreau E, Touchard G (2007) Stall control at high angle of attack with plasma sheet actuators. Experiments in Fluids 42:143-167
217
Stanislas M, Okamoto K, Kähler C, Westerweel J, Scarano F (2008) Main results of the third international PIV Challenge. Experiments in Fluids 45[1]:27-71
Sung Y, Kim W, Mungal M, Cappelli M (2006) Aerodynamic Modification of Flow over Bluff Objects by Plasma Actuation. Experiments in Fluids Vol. 41:479-486
Takeuchi N, Yasuoka K, Ishii S (2007) Inducing Mechanisms of Electrohydrodynamic Flow by Surface Barrier Discharge. IEEE Transactions on Plasma Science 35[6]:1704-1709
Thomas F, Corke T, Igbal M, Kozlov A, Schatzman D (2009) Optimization of Dielectric Barrier Discharge Plasma Actuators for Active Aerodynamic Flow Control. AIAA Journal 47[9]:2169-2178
Thomas F, Kozlov A, Corke T (2008) Plasma Actuators for Cylinder Flow Control and Noise Reduction. AIAA Journal 46[9]:1921-1931
Utkin YG, Keshav S, Kim J-H, Kastner J, Adamovich IV, Samimy M (2007) Development and Use of Localized Arc Filament Plasma Actuators for High-Speed Flow Control. Journal of Physics D: Applied Physics 40[3]:685-694
Vorobiev A, Rennie RM, Jumper E, McLaughlin T (2008) Experimental Investigation of Lift Enhancement and Roll Control using Plasma Actuators. Journal of Aircraft 45[4]:1315-1321
Wygnanski I (2004) The Variables Affecting the Control of Separation by Periodic Excitation. AIAA Paper 2004-2505
Zaman K, Culley D (2008) Flow Separation Control over an Airfoil: Implication of Wake Velocity-Deficit Reduction. AIAA Paper 2008-3768