compact, high-power boundary layer separation control actuation ...
Proceedings of ASME FEDSM’01: 2001 ASME Fluids Engineering Division Summer Meeting New Orleans, Louisiana, May29-June1, 2001
18279 COMPACT, HIGH-POWER BOUNDARY LAYER SEPARATION CONTROL ACTUATION DEVELOPMENT Duane C. McCormick
1
Steven A. Lozyniak Douglas G. MacMartin United Technologies Research Center, East Hartford Connecticut 1 currently at California Institute of Technology, Pasadena California
ABSTRACT The development of a self-contained, synthetic jet actuator for stall control of a pitching airfoil is described. The test simulates the full-scale (chord, Reynolds and Mach numbers) conditions of rotorcraft retreating blade stall. The required slot velocity was 65 m/s amplitude at 250 Hz (Cµ = 0.1% and F+=1). The packaging of a selfcontained actuator inside the confined leading edge of the airfoil represented a significant challenge. The approach taken was a voicecoil linear motor driving a piston/spring assembly at resonance (first mass-spring mode). This paper describes the mechanical design aspects of the spring, actuator loss mechanisms, and the electroacoustic modeling of the device. The high stiffness requirement (dictated by the design frequency and moving mass) combined with the large displacement requirement (due to confined space of the leading edge limiting piston size) made the spring design the most difficult challenge. This challenge was overcome, resulting in a successful bench top actuator test that met performance targets and agreed well with the model. The final embodiment designed for use in the wind tunnel blade section leading edge is described.
Peter F. Lorber
limits rotor load, flight speed, and maneuverability, in addition to creating very large implusive blade pitching moments which can cause mechanical damage. For the full scale application, the relative external flow is Mach number is M=0.3 - 0.5, while the peak local Mach number is typically M=1.1 - 1.3. Full scale Reynolds number is Re~4x106. Though numerous studies have investigated separation control on leading edge stall, these full scale conditions are typically not addressed, particularly for a self-contained actuation system. As described in [4], based on previous dynamic separation control investigations, a forcing level of Cµ = 0.1-0.2% and forcing frequency of F+=1 was selected as the design specification. This level is expected to generate at least a 5O increase in stall angle and 10% Time-Averaged Mass and Momentum Flow
lo
-m wx
om
high x-mom
INTRODUCTION
With Control
In the area of flow control, a synthetic jet is generally considered the acoustic streaming of flow from an orifice or slot being driven by a pressure oscillation (with zero mean pressure difference) in an adjacent cavity. The pressure oscillation is usually generated by a moving diaphragm inside the cavity. Synthetic jets have been applied extensively in flow control studies to generate virtual shapes on solid walls (e.g., Glezer, et al. [1]) and to control boundary layer separation (e.g., McCormick [2]). Regarding boundary layer control, the synthetic jet can act to efficiently provide periodic forcing for dynamic separation control as described by Wygnanski [3], and at higher levels completely suppress the separation by sufficient axial momentum injection into at the wall, so-called directed synthetic jet [2], as shown in Fig. 1.
increase in maximum lift coefficient. If such blade performance improvements are realized, significant improvement in rotorcraft lift, speed, and maneuverability are possible. However, achieving the required actuation levels represents the most difficult barrier that must be overcome to make blade stall control practical for full scale rotorcraft.
This paper addresses the application of dynamic separation control to retreating blade stall control on rotorcraft blades with synthetic jets. This effort is a continuation of work described by Lorber, et al. [4], whose goal is to demonstrate control at realistic conditions and to develop practical actuation. Retreating blade stall is due to boundary layer separation near the leading edge of the rotor blade during rapid motion to high blade angle of attack. This stall
Two key aspects of the actuator design are first discussed; the design of the spring and minimization of actuator losses. These design modifications were validated in a benchtop prototype, enabling design of the actuation for integration into the rotor blade for testing. Previous work [2,4] developed an electro-acoustic model for the design of the synthetic jet, and the final section of the paper describes the updated model incorporating the design changes.
Acoustically driven cavity
Figure 1. Directed synthetic jet concept
1
Copyright © 2001 by ASME
NOMENCLATURE AN
neck or slot area (m2)
BL
voice coil force constant = (magnetic flux) x (voice coil length) (N/amp)
c
speed of sound (m/s) or airfoil chord (m)
CV
acoustic compliance on front side of piston = V/(ρc2) (m5/N)
CVB
acoustic compliance on backside of piston (m5/N)
CVT
acoustic compliance in blade trailing section (m5/N)
Cµ
momentum coefficient = ρhuN2/(ρxREFUREF2)
f
frequency (Hz)
As described in [4], an actuator screening study was performed early in this effort. Candidates included piezoelectric ceramics, fluidics, linear and rotary electromechanical motors, periodic bleed flow modulations, and others. Of these options, the linear electromechanical motor approach was found to have the highest metrics (efficiency, size, weight, complexity). By subsequent analysis and modeling, the moving voice coil linear motor was selected over a variable air-gap and variable reluctance motors due to lightest moving mass, more manageable tolerances, and commercial availability. The basic actuator consists of the voice coil motor directly driving a piston inside the cavity to which the slot is connected. The 0.20 0.18 High amplitude forcing (peak un = 80m/s)
0.16 0.14
nondimensional forcing frequency = f xREF./UREF
h
width of neck, or slot (m)
KD
neck jet dump loss coefficient
LN
effective acoustic length of neck or slot (m)
Mt
total moving actuator mass Mp+Mc+Mspring (kg)
0.02
MN
acoustic mass in neck, or slot = ρLN/AN (kg/m )
0.00 200
Efficiency
F+
0.12 0.10 0.08
Low amplitude forcing (peak un = 1 m/s)
0.06 0.04
4
250
300
MS
acoustic mass in spar slots (kg/m )
Re
voice coil resistance (ohms)
Rm
actuator mechanical losses (N-s/m5)
RN
neck acoustic resistance = 0.5KDρuN/AN + viscous neck loss + acoustic radiation loss (N-s/m5)
RE
acoustic resistance between blade edge and wind tunnel wall (N-s/m)
RS
acoustic resistance of spar slot (N-s/m)
RL
acoustic resistance of piston leakage (N-s/m)
SD
effective area of piston (m2)
uN
amplitude of neck or slot oscillation (m/s)
Up
piston volume velocity (m3/s)
UREF
freestream, reference velocity (m/s)
350
400
450
500
Frequency (Hz)
4
Figure 2. Typical actuator efficiency versus frequency
piston is attached to the stationary frame by a restoring spring. The main purpose of the spring is to balance the force required to oscillate the large moving interia (ωMtUp) by operating at resonance. The total
Spring
Coil
Field assembly
MN RN
3
V
volume of cavity (m )
VSPK
voltage across voice coil terminals (volts)
xREF
characteristic length of uncontrolled separation (e.g., for airfoil with leading edge separation, xREF = chord) (m)
Piston Spring
CV
RL CVB
Coil & field assembly
Figure 3. Initial bench top actuator [4]
BASIC ACTUATOR CONCEPT
moving mass, Mt, consists of the motor’s moving mass, the piston and any attachment features, and the moving part of the spring. Hence, there is a need to minimize moving mass in order to minimize the strain energy the spring must store. In addition, great attention must
2
Copyright © 2001 by ASME
be paid to the actuator losses so as not to dominate the force requirements (since the motor is nominally force limited). The limitation in operational bandwidth is illustrated in Figure 2, which gives a typical actuator efficiency versus frequency for a low and high level of input coil voltage level. At low excitation level, there are two peaks, the first related to the mass/spring resonance and the second due to the cavity/slot Helmholtz mode. At
The initial spring design with high stiffness and strain rate (previously described in [4] for a prototype actuator) is shown in Fig. 3. The spring was a circular plate of spring steel, held fixed at the OD and attached to the voice coil at the ID with spiral cuts creating long (relative to plate radius) cantilevered curved beams. Though this spring provided the required strain, the stiffness was low by a factor of four. It also was found to be prone to side-to-side instabilities. Commercially available wire-coil springs were found to provide adequate stiffness and strain capability. However, due to the difficulty in attaching to the spring ends, coil springs need to be deployed on both sides of the moving mass (i.e., keeping the springs in compression, or tension throughout the cycle). The extra springs and frames to support them were considered too difficult to achieve in the available space of the leading edge. A breakthrough in spring design was achieved by applying machined coil spring technology from the HELI-CAL Products Company (see Boehm [5]). By machining the coil spring, high
Field assembly (permanent magnet)
0.0 0.1 0.2
Contoured Flow Slot
Voice coil 0.5 0.0
0.3
Piston
0.4
Spring
0.5
Voice coil 5 cm piston
Field assembly
Figure 4. Low loss slot design and CFD solution [7] during peak in-stroke Spring
high excitation levels, the efficiency peak only exists for the mass/spring resonance at a much higher value of efficiency (since this is where the actuator is impedance matched). The Helmholtz mode is significantly damped such that it is imperceptible due to the increased power dissipation at the slot.
SPRING DESIGN Given the resonant-point operation, the initial spring stiffness is determined from the design forcing frequency (F+=1 criterion used here) and the moving mass of the voice coil and piston. After designing the spring and determining its moving mass (which is then added to the voice coil and piston mass), the calculation is repeated to find the new spring stiffness. This process is continued until a converged stiffness is found. In addition to the high stiffness that results from this process, the limited piston size drives the design to large spring deflection, ~2.5 mm amplitude. The combination of limited space, high amplitude displacement and high stiffness leads to very high spring strain rates (13-15%).
Figure 5. Bench top prototype actuator with machined coil spring strength attachments to the spring ends are possible, enabling operation in both compression and tension. This feature is important since it greatly simplifies the design. In addition, the HELI-CAL design employs multiple coil starts which eliminate moments about the spring axis and provide side-to-side stability.
LOSS MECHANISMS A second area critical to successful design is attention to losses; not only is higher efficiency preferable, but the actuator may not even be capable of achieving sufficient authority if the force output of the voice coil is dominated by overcoming losses. Loss mechanisms are broken down into two categories: external (e.g. slot) and internal losses to the mechanical part of the actuator. The external losses were
3
Copyright © 2001 by ASME
modeled individually, whereas, the internal losses are lumped into a single parameter (mechanical quality factor). Minimizing both is important to a successful actuator design. The external losses consist mainly of flow losses in the slot and piston leakage (through the unsealed gap at edge of piston), Neck Velocity Amplitude (m/s)
60
tangential injection momentum and increased base area drag (on the rounded edge). To mitigate the jet dump loss into the cavity, the contraction on the cavity side is designed to also act as a diffuser. Figure 4 shows the resulting slot design and corresponding CFD solution during in-stroke (see [7] for details) which illustrates the effect of in-stroke contouring on the diffuser flowfield. The in-stroke diffuser has been designed by steady diffuser criterion [8] to recover ~60% of the in-stroke dynamic head. The contour lines in Fig. 4 are the pressure coefficient referenced to the inlet slot conditions, indicating a 50% recovery of dynamic head.
Data
40
Model 20
0 0
10
20
30
40
50
As in loudspeaker design (e.g. Beranek [6]), internal actuator losses (other than the electrical resistance of the coil) can be modeled by the mechanical quality factor, Qm = ωoMt/Rm (ratio of inertial to loss impedance), though the scaling is far from perfect as described later. For this study, the “mechanical” resistance, Rm, represents the net sum of the structural losses in the spring, air viscous losses on the coil inside the field assembly, and the eddy currents in the metal former tube (spool on which the coil is wrapped) and piston. In addition, any non-linearity of the spring stiffness with displacement will look like an amplitude dependent quality factor. As discussed in
Input Power (W)
Spring
Figure 6. Bench top actuator output Voice coil
though as discussed below, other significant fluid losses are possible. The piston leakage can in principal be eliminated with a seal (e.g., speaker-like surround or bellows) but for the blade configuration a seal results in large percentage loss in piston area and increases the complexity substantially. In order to mitigate leakage loss, the key is to minimize the acoustic loading on the piston backside. Any loading results in increased driving pressure potential across the gap and substantially increases the jet dump losses associated with the leakage.
Spar
Slot flow losses are in part due to separation that can occur inside the slot on both the in-stroke and outstroke, reducing the effective flow area of the slot (vena contracta effect). Also, jet dump loss in the cavity during the in-stroke can cause significant loss (e.g.
Figure 8. Compact spring-piston-motor installation [4], achieving a quality factor ~ 50 (or more) is crucial to a successful actuator that can deliver the desired output without damaging the coil. Hence, significant attention needs to be given to the internal loss mechanisms.
60
Coil Temp (C)
Magnet
40
20
0 0
10
20
30
40
50
The mechanical quality factor can be quantified directly by measuring the frequency dependence of the voice-coil electrical impedance when (near) zero acoustic load is applied. Under this condition, the only significant losses present are the internal losses, so using the relation Qm= ∆f/fo (where ∆f is the bandwidth where the impedance is above the 3 dB down limit) gives a direct measurement of quality factor [6]. Alternatively, an accelerometer frequency response from a hammer test can be used to determine quality factor in the same manner (at least at low amplitude).
Input Power (W)
Figure 7. Bench top actuator thermal heating see Fig 19 of [4] illustrates the wasted kinetic energy into turbulence of the in-stroke flow). The viscous and acoustic radiation losses in the slot are negligible at high forcing level [2]. To eliminate the vena contracta effect on the outstroke, a simple contraction on the cavity side of the slot will do. On the in-stroke, this separation can be mitigated by rounding the upstream edge of the slot with some loss of
For a reasonable design which has a rigid connection on the stationary end of the spring, structural losses are small, ζ ~ 0.25% damping ratio (Qm=200) as determined from hammer tests without the magnet field assembly. Fluid viscous losses due to the voice coil motion were found to be no greater, and possibly less than the structural losses, as determined by impedance testing under vacuum conditions using the actuator described in [4]. In contrast to the structural and fluid viscous losses, eddy current loss was found to be
4
Copyright © 2001 by ASME
the most significant internal loss. This loss is due to any metallic material moving through a magnetic field, which causes electrical currents to flow near the material surface and create resistive losses. For the actuators built in this investigation, the primary eddy currents occurred in the coil former tube. Detailed measurements of the 50 Hz actuator [4] shown in Fig. 2 gave the following breakdown of losses: structural = 0.25%, fluid viscous ~0%, eddy current losses in
wall was 0.13 mm. The backside of the piston was vented to minimize piston leakage penalty. In addition to the adjustable ring, a metal bellows section was tested in its place, providing a sealed piston arrangement. Though this improved efficiency (as expected), the packaging of bellows in the Piston
5 module packs independently controlled across span
Module frame Module pack without frame
2.54 cm
Coil
Magnet
Motor/spring sub-assembly
Figure 10. Motor/spring subassembly and module pack leading edge would result in significantly reduced piston area which would increase the required stroke. For these reasons, the metal bellows arrangement was not pursued further. Figure 9. Blade section actuator arrangement former tube 4%, eddy current losses in spring/piston assembly 1%. The breakdown between the eddy current loss locations was found by testing with a former tube fabricated from plastic (thus eliminating eddy currents in the former tube). The operational quality factor is Qm~10, much lower than needed. These eddy currents can be mitigated significantly (at some expense) by either slotting the former tubes before winding or by construction with a non-metallic material. However, because the eddy currents do not depend on frequency, the effect on quality factor is greatly reduced (but not eliminated) by operation at the higher frequency of the retreating blade stall application. In other words, the quality factor scaling is only invariant for losses which scale linearly with frequency and/or mass (like, e.g., structural losses).
The moving mass had the following percentage breakdown: voice coil = 27%, piston = 12%, and moving part of spring = 61%. The moving mass of the spring is clearly very significant and has a major influence on the required stiffness. The actuator resonance occurred as 270 Hz. Based on electrical impedance measurements, the actuator was found to have a reasonable quality factor, Qm ~ 40.
Spar
BENCHTOP ACTUATION Using the machined coil spring concept, a new bench top actuator was designed and fabricated based on a commercially available voice-coil linear motor. The design frequency was nominally 250 Hz. Figure 5 shows the assembly and a close up photograph of the spring-motor-piston assembly. In order to minimize moving mass (i.e. no flanges with screws), a threaded connection was made between the free end of the spring and the voice coil and an EB weld connection was made with the piston. The cavity wall in the area of the piston was an adjustable ring to provide trim centering with the piston. Above the adjustable ring, the remainder of the cavity was formed by an aluminum housing which also suspended the motor field assembly (permanent magnet) over the voice coil and spring. Above the housing was a contoured flow slot (as discussed above) with a 2.5 mm gap and 5 cm span. The nominal gap between piston and cavity
Module pack frame Figure 11. Module pack installation on blade spar Separating out the eddy current losses via hammer tests with and without the field assembly was not possible with this arrangement, however, the higher quality factor is consistent with the expected scaling of mechanical and eddy current losses with frequency. Operation at 2.5 mm amplitude displacement was routinely obtained without failure. Ensemble and spatially averaged hot wire measurements in the slot were obtained to quantify performance. Figure 6 shows the
5
Copyright © 2001 by ASME
neck velocity amplitude versus input electrical power, which shows that near 60 m/s amplitude was achieved. An electro-acoustic circuit
In summary, the second bench top actuation prototype successfully demonstrated the machined coil spring concept, reasonable efficiency with piston leakage, and low-loss slot design.
BLADE ACTUATION
Load (N)
In order to adapt the bench top actuator into a blade design, a rigid brace was needed for both the fixed end of the spring and the field assembly. For the bench top actuator, these locations were on opposite sides of the piston. Due to the limited space of the blade section leading edge, such an arrangement was not possible. To solve this problem, the voice coil motor was located inside the spring, hence the fixed end of the spring and the field assembly were co-located, significantly simplifying the installation and improving the compactness of the actuator. Figure 8 shows a cross section of the MN RN Deflection (mm)
CV CVB Spar
Figure 12. Typical stiffness of spring
ZACT
The voice coil temperature was monitored due to concerns of overheating causing failure in the voice-coil adhesion material. Figure 7 plots the temperature versus input power (power consumed by coil resistive heating is about half of the total input value). The results showed that the coil heating is safely below the manufacturer’s
Acceleration/Force (volt/volt)
6000 4000
Magnet QM= 81 (ζ = 0.6%)
2000 0 250 255 260 265 270 275 280 285 290 295 300 Freq - Hz
Figure 13. Hammer test results for motor/spring subassembly stated thermal limit (155 oC) despite being significantly above the continuous operating power (9 W versus 25 W). This favorable result is due to the heat transfer to the magnet, which is connected to a large heat sink, maintaining the magnet at a near constant cool temperature. A heat transfer analysis of solely the conduction from the coil across the air gap into the magnet is sufficient to avoid overheating.
UN UP
ZF
MN CV
RN
US
RL ZB
ZB
CVB CVT
MS RS RE
Figure 14. Electro-acoustic model of actuator in blade section
12000
No magnet QM= 222 (ζ =0.2%)
CVT
ZF
V BL R CS2D
PAmb
10000
RE
RL
model (similar to that shown in Fig. 10 of [4] without the backside loading) was fit to this data mainly by adjusting the discharge coefficient of the slot. A value of Kd = 1.4 was found to fit the data well as shown in the figure, confirming the low-loss nature of the slot design (Kd = 1 is near ideal) and represents the lowest value achieved to date for the current effort. Based on the output flow power, the actuator efficiency was determined to be about 20%.
8000
MS RS
motor/spring sub-assembly in the blade leading edge. Three subassemblies were grouped together in a module pack, driving a rectangular piston as shown in Fig. 9. Five module packs cover the blade section span, isolated from each via a partition. The partitions enable a spatial course variation in phase of the forcing. The intent is to investigate, in addition to uniform phase forcing, opposite phase forcing for adjacent slots since such forcing can be achieved with several other actuator options (e.g., using both sides of the piston to drive adjacent slots). The backside of the pistons are vented through slots on the top and bottom of the spar in order to minimize the acoustic loading and hence the unsealed piston leakage. Figure 10 gives a photograph of the actuator subassembly and module pack components and Fig. 11 shows a photograph of the module packs (one complete assembly on far right) installed on the blade spar. The springs were fabricated from corrosion resistance steel, heat treated for strength and, after machining, shot peened for improved surface finish/durability. The fifteen springs were individually statically tested for spring rate. The variation from the mean spring rate was plus 4% and minus 3%. Trim weighs were attached to the voice coils to account for the spring rate variation. Figure 12 gives a typical force-deformation plot which illustrates the linearity and small hysterisis of the spring (the nonlinear behavior near zero load are due to the initial sitting of the actuator top and bottom into the test rig and should be ignored).
6
Copyright © 2001 by ASME
Figure 13 shows the results of a hammer test on an individual sub-assembly with and without the motor field assembly. Without the field assembly, the structural losses were identified to be about 0.2% damping, consistent with the original bench top actuator from [4]. With the field assembly, an additional 0.4% damping is incurred, primarily due to eddy currents in the former tube. As earlier
Current efforts as the manuscript deadline approached are aimed at the final assembly of the blade section in the wind tunnel with complete actuation and instrumentation. Hot wire measurements are planned in-situ as a final check out and verification of model predictions.
Efficiency
0.20 0.15 0.10
CONCLUSIONS A compact, high-power synthetic jet actuator for a full-scale retreating blade stall wind tunnel test has been developed. The key breakthrough was the application of machined coil spring technology to the spring design. In addition, detailed attention to loss mechanisms and mitigation of these losses are crucial to the successful design. Particularly noteworthy is the new, low-loss slot design that minimizes slot separation effects and recovers dynamic head on the instroke. A bench top prototype actuator has been successfully built and tested. The design has been adapted to the leading edge of the blade section.
Design point
0.05 0 300 Input power (W)
plot of the expected performance of the actuation system based on the bench top results and the model in Fig. 14. The model indicates an efficiency of about 15%, requiring 260 Watts of electrical power to drive each module pack. Under more ideal design constraints (e.g., larger piston, less acoustic back side load or sealed piston), this efficiency could be at least doubled.
200
ACKNOWLEDGMENTS 100
0 0
20
40
60
80
uN (m/s) Figure 15. Predicted blade actuator performance
hypothesized that the eddy current does not scale with frequency, the relative contribution is significantly less for the higher frequency actuator (0.4% versus 5%). For the assembled module pack, hammer tests and electrical impedance test indicate a quality factor of Qm~5055 which is somewhat lower than the individual subassembly but still very reasonable.
This work is supported by the DARPA Microadaptive Flow Control Program (Richard Wlezian) under U.S. Army Research Office (Thomas Doligalski) Contract DAAG55-98-C-0066. Gary Boehm from HELI-CAL Products Company is acknowledged for the machined coil spring design.
REFERENCES 1.
Glezer, A., Allen, M.G., Coe, D.J., Barton, S.L., Trautman, M.A., and Wiltse, J.W., “Synthetic Jet Actuator and Applications Thereof”, U.S. Patent 5,758,823, June 2, 1998.
2.
McCormick, D.C., “Boundary Layer Separation Control with Directed Synthetic Jets”, AIAA 2000-0519, January 2000.
3.
Wygnanski, I, “Method and Apparatus for Delaying Separation of Flow from a Solid Surface”, U.S. Patent 5,209,438, May 11, 1993.
4.
Lorber, P.F., McCormick, D.C., Anderson, T.J., Wake, B.E., MacMartin, D.G., Pollack, M., Corke, T., and Breuer, K., “Rotorcraft Retreating Blade Stall Control”, AIAA 2000-2475, June 2000.
5.
“Engineering Solutions for Design & Manufacturing, Machined Springs: a Very Good Solution”, NASA Tech Briefs, Vol. 23, No. 4, April 1999.
6.
Beranek, L.L., “Acoustics”, Acoustical Society of America, Cambridge, MA, 1993.
7.
Wake, B.E., and Lurie, E.A., “Computational Evaluation of Directed Synthetic Jets for Dynamic Stall Control”, 57th American Helicopter Society 57th Annual Forum, May 2001.
7
Copyright © 2001 by ASME
ELECTRO-ACOUSTIC MODEL Due to the cavities and spar slots on the backside of the piston, the electro-acoustic model of the synthetic jet is a bit more complicated than models previously presented [2], [4]. Figure 14 shows the front and backside loading are modeled as two impedances in series (ZF, ZB) and the piston leakage as a resistor in parallel (acoustic mass of leakage gap neglected). As before, the actuator impedance (ZACT) is a second order system consisting of the moving mass, spring compliance, internal losses (Qm), and the coil resistance (mapped into the acoustic side of the circuit, see [2] for details). The expanded piston loading impedances are also shown in the figure. Critical to a successful design is to ensure that the resonant mode of the backside is out of band relative to the operational frequency. Otherwise, a large velocity oscillation will occur in the spar slots and large pressure oscillation across the piston, causing excessive losses. For the current arrangement, this mode is ~600 Hz. Figure 15 gives a
8.
Cocanower, A.B., Kline, S.J., and Johnston, J.P., “A Unified Method for Predicting the Performance of Subsonic Diffusers of Several Geometries”, Stanford University, Department of Mechanical Engineering Report PD-10, May 1965.
8
Copyright © 2001 by ASME
18279 COMPACT, HIGH-POWER BOUNDARY LAYER SEPARATION CONTROL ACTUATION DEVELOPMENT Duane C. McCormick
1
Steven A. Lozyniak Douglas G. MacMartin United Technologies Research Center, East Hartford Connecticut 1 currently at California Institute of Technology, Pasadena California
ABSTRACT The development of a self-contained, synthetic jet actuator for stall control of a pitching airfoil is described. The test simulates the full-scale (chord, Reynolds and Mach numbers) conditions of rotorcraft retreating blade stall. The required slot velocity was 65 m/s amplitude at 250 Hz (Cµ = 0.1% and F+=1). The packaging of a selfcontained actuator inside the confined leading edge of the airfoil represented a significant challenge. The approach taken was a voicecoil linear motor driving a piston/spring assembly at resonance (first mass-spring mode). This paper describes the mechanical design aspects of the spring, actuator loss mechanisms, and the electroacoustic modeling of the device. The high stiffness requirement (dictated by the design frequency and moving mass) combined with the large displacement requirement (due to confined space of the leading edge limiting piston size) made the spring design the most difficult challenge. This challenge was overcome, resulting in a successful bench top actuator test that met performance targets and agreed well with the model. The final embodiment designed for use in the wind tunnel blade section leading edge is described.
Peter F. Lorber
limits rotor load, flight speed, and maneuverability, in addition to creating very large implusive blade pitching moments which can cause mechanical damage. For the full scale application, the relative external flow is Mach number is M=0.3 - 0.5, while the peak local Mach number is typically M=1.1 - 1.3. Full scale Reynolds number is Re~4x106. Though numerous studies have investigated separation control on leading edge stall, these full scale conditions are typically not addressed, particularly for a self-contained actuation system. As described in [4], based on previous dynamic separation control investigations, a forcing level of Cµ = 0.1-0.2% and forcing frequency of F+=1 was selected as the design specification. This level is expected to generate at least a 5O increase in stall angle and 10% Time-Averaged Mass and Momentum Flow
lo
-m wx
om
high x-mom
INTRODUCTION
With Control
In the area of flow control, a synthetic jet is generally considered the acoustic streaming of flow from an orifice or slot being driven by a pressure oscillation (with zero mean pressure difference) in an adjacent cavity. The pressure oscillation is usually generated by a moving diaphragm inside the cavity. Synthetic jets have been applied extensively in flow control studies to generate virtual shapes on solid walls (e.g., Glezer, et al. [1]) and to control boundary layer separation (e.g., McCormick [2]). Regarding boundary layer control, the synthetic jet can act to efficiently provide periodic forcing for dynamic separation control as described by Wygnanski [3], and at higher levels completely suppress the separation by sufficient axial momentum injection into at the wall, so-called directed synthetic jet [2], as shown in Fig. 1.
increase in maximum lift coefficient. If such blade performance improvements are realized, significant improvement in rotorcraft lift, speed, and maneuverability are possible. However, achieving the required actuation levels represents the most difficult barrier that must be overcome to make blade stall control practical for full scale rotorcraft.
This paper addresses the application of dynamic separation control to retreating blade stall control on rotorcraft blades with synthetic jets. This effort is a continuation of work described by Lorber, et al. [4], whose goal is to demonstrate control at realistic conditions and to develop practical actuation. Retreating blade stall is due to boundary layer separation near the leading edge of the rotor blade during rapid motion to high blade angle of attack. This stall
Two key aspects of the actuator design are first discussed; the design of the spring and minimization of actuator losses. These design modifications were validated in a benchtop prototype, enabling design of the actuation for integration into the rotor blade for testing. Previous work [2,4] developed an electro-acoustic model for the design of the synthetic jet, and the final section of the paper describes the updated model incorporating the design changes.
Acoustically driven cavity
Figure 1. Directed synthetic jet concept
1
Copyright © 2001 by ASME
NOMENCLATURE AN
neck or slot area (m2)
BL
voice coil force constant = (magnetic flux) x (voice coil length) (N/amp)
c
speed of sound (m/s) or airfoil chord (m)
CV
acoustic compliance on front side of piston = V/(ρc2) (m5/N)
CVB
acoustic compliance on backside of piston (m5/N)
CVT
acoustic compliance in blade trailing section (m5/N)
Cµ
momentum coefficient = ρhuN2/(ρxREFUREF2)
f
frequency (Hz)
As described in [4], an actuator screening study was performed early in this effort. Candidates included piezoelectric ceramics, fluidics, linear and rotary electromechanical motors, periodic bleed flow modulations, and others. Of these options, the linear electromechanical motor approach was found to have the highest metrics (efficiency, size, weight, complexity). By subsequent analysis and modeling, the moving voice coil linear motor was selected over a variable air-gap and variable reluctance motors due to lightest moving mass, more manageable tolerances, and commercial availability. The basic actuator consists of the voice coil motor directly driving a piston inside the cavity to which the slot is connected. The 0.20 0.18 High amplitude forcing (peak un = 80m/s)
0.16 0.14
nondimensional forcing frequency = f xREF./UREF
h
width of neck, or slot (m)
KD
neck jet dump loss coefficient
LN
effective acoustic length of neck or slot (m)
Mt
total moving actuator mass Mp+Mc+Mspring (kg)
0.02
MN
acoustic mass in neck, or slot = ρLN/AN (kg/m )
0.00 200
Efficiency
F+
0.12 0.10 0.08
Low amplitude forcing (peak un = 1 m/s)
0.06 0.04
4
250
300
MS
acoustic mass in spar slots (kg/m )
Re
voice coil resistance (ohms)
Rm
actuator mechanical losses (N-s/m5)
RN
neck acoustic resistance = 0.5KDρuN/AN + viscous neck loss + acoustic radiation loss (N-s/m5)
RE
acoustic resistance between blade edge and wind tunnel wall (N-s/m)
RS
acoustic resistance of spar slot (N-s/m)
RL
acoustic resistance of piston leakage (N-s/m)
SD
effective area of piston (m2)
uN
amplitude of neck or slot oscillation (m/s)
Up
piston volume velocity (m3/s)
UREF
freestream, reference velocity (m/s)
350
400
450
500
Frequency (Hz)
4
Figure 2. Typical actuator efficiency versus frequency
piston is attached to the stationary frame by a restoring spring. The main purpose of the spring is to balance the force required to oscillate the large moving interia (ωMtUp) by operating at resonance. The total
Spring
Coil
Field assembly
MN RN
3
V
volume of cavity (m )
VSPK
voltage across voice coil terminals (volts)
xREF
characteristic length of uncontrolled separation (e.g., for airfoil with leading edge separation, xREF = chord) (m)
Piston Spring
CV
RL CVB
Coil & field assembly
Figure 3. Initial bench top actuator [4]
BASIC ACTUATOR CONCEPT
moving mass, Mt, consists of the motor’s moving mass, the piston and any attachment features, and the moving part of the spring. Hence, there is a need to minimize moving mass in order to minimize the strain energy the spring must store. In addition, great attention must
2
Copyright © 2001 by ASME
be paid to the actuator losses so as not to dominate the force requirements (since the motor is nominally force limited). The limitation in operational bandwidth is illustrated in Figure 2, which gives a typical actuator efficiency versus frequency for a low and high level of input coil voltage level. At low excitation level, there are two peaks, the first related to the mass/spring resonance and the second due to the cavity/slot Helmholtz mode. At
The initial spring design with high stiffness and strain rate (previously described in [4] for a prototype actuator) is shown in Fig. 3. The spring was a circular plate of spring steel, held fixed at the OD and attached to the voice coil at the ID with spiral cuts creating long (relative to plate radius) cantilevered curved beams. Though this spring provided the required strain, the stiffness was low by a factor of four. It also was found to be prone to side-to-side instabilities. Commercially available wire-coil springs were found to provide adequate stiffness and strain capability. However, due to the difficulty in attaching to the spring ends, coil springs need to be deployed on both sides of the moving mass (i.e., keeping the springs in compression, or tension throughout the cycle). The extra springs and frames to support them were considered too difficult to achieve in the available space of the leading edge. A breakthrough in spring design was achieved by applying machined coil spring technology from the HELI-CAL Products Company (see Boehm [5]). By machining the coil spring, high
Field assembly (permanent magnet)
0.0 0.1 0.2
Contoured Flow Slot
Voice coil 0.5 0.0
0.3
Piston
0.4
Spring
0.5
Voice coil 5 cm piston
Field assembly
Figure 4. Low loss slot design and CFD solution [7] during peak in-stroke Spring
high excitation levels, the efficiency peak only exists for the mass/spring resonance at a much higher value of efficiency (since this is where the actuator is impedance matched). The Helmholtz mode is significantly damped such that it is imperceptible due to the increased power dissipation at the slot.
SPRING DESIGN Given the resonant-point operation, the initial spring stiffness is determined from the design forcing frequency (F+=1 criterion used here) and the moving mass of the voice coil and piston. After designing the spring and determining its moving mass (which is then added to the voice coil and piston mass), the calculation is repeated to find the new spring stiffness. This process is continued until a converged stiffness is found. In addition to the high stiffness that results from this process, the limited piston size drives the design to large spring deflection, ~2.5 mm amplitude. The combination of limited space, high amplitude displacement and high stiffness leads to very high spring strain rates (13-15%).
Figure 5. Bench top prototype actuator with machined coil spring strength attachments to the spring ends are possible, enabling operation in both compression and tension. This feature is important since it greatly simplifies the design. In addition, the HELI-CAL design employs multiple coil starts which eliminate moments about the spring axis and provide side-to-side stability.
LOSS MECHANISMS A second area critical to successful design is attention to losses; not only is higher efficiency preferable, but the actuator may not even be capable of achieving sufficient authority if the force output of the voice coil is dominated by overcoming losses. Loss mechanisms are broken down into two categories: external (e.g. slot) and internal losses to the mechanical part of the actuator. The external losses were
3
Copyright © 2001 by ASME
modeled individually, whereas, the internal losses are lumped into a single parameter (mechanical quality factor). Minimizing both is important to a successful actuator design. The external losses consist mainly of flow losses in the slot and piston leakage (through the unsealed gap at edge of piston), Neck Velocity Amplitude (m/s)
60
tangential injection momentum and increased base area drag (on the rounded edge). To mitigate the jet dump loss into the cavity, the contraction on the cavity side is designed to also act as a diffuser. Figure 4 shows the resulting slot design and corresponding CFD solution during in-stroke (see [7] for details) which illustrates the effect of in-stroke contouring on the diffuser flowfield. The in-stroke diffuser has been designed by steady diffuser criterion [8] to recover ~60% of the in-stroke dynamic head. The contour lines in Fig. 4 are the pressure coefficient referenced to the inlet slot conditions, indicating a 50% recovery of dynamic head.
Data
40
Model 20
0 0
10
20
30
40
50
As in loudspeaker design (e.g. Beranek [6]), internal actuator losses (other than the electrical resistance of the coil) can be modeled by the mechanical quality factor, Qm = ωoMt/Rm (ratio of inertial to loss impedance), though the scaling is far from perfect as described later. For this study, the “mechanical” resistance, Rm, represents the net sum of the structural losses in the spring, air viscous losses on the coil inside the field assembly, and the eddy currents in the metal former tube (spool on which the coil is wrapped) and piston. In addition, any non-linearity of the spring stiffness with displacement will look like an amplitude dependent quality factor. As discussed in
Input Power (W)
Spring
Figure 6. Bench top actuator output Voice coil
though as discussed below, other significant fluid losses are possible. The piston leakage can in principal be eliminated with a seal (e.g., speaker-like surround or bellows) but for the blade configuration a seal results in large percentage loss in piston area and increases the complexity substantially. In order to mitigate leakage loss, the key is to minimize the acoustic loading on the piston backside. Any loading results in increased driving pressure potential across the gap and substantially increases the jet dump losses associated with the leakage.
Spar
Slot flow losses are in part due to separation that can occur inside the slot on both the in-stroke and outstroke, reducing the effective flow area of the slot (vena contracta effect). Also, jet dump loss in the cavity during the in-stroke can cause significant loss (e.g.
Figure 8. Compact spring-piston-motor installation [4], achieving a quality factor ~ 50 (or more) is crucial to a successful actuator that can deliver the desired output without damaging the coil. Hence, significant attention needs to be given to the internal loss mechanisms.
60
Coil Temp (C)
Magnet
40
20
0 0
10
20
30
40
50
The mechanical quality factor can be quantified directly by measuring the frequency dependence of the voice-coil electrical impedance when (near) zero acoustic load is applied. Under this condition, the only significant losses present are the internal losses, so using the relation Qm= ∆f/fo (where ∆f is the bandwidth where the impedance is above the 3 dB down limit) gives a direct measurement of quality factor [6]. Alternatively, an accelerometer frequency response from a hammer test can be used to determine quality factor in the same manner (at least at low amplitude).
Input Power (W)
Figure 7. Bench top actuator thermal heating see Fig 19 of [4] illustrates the wasted kinetic energy into turbulence of the in-stroke flow). The viscous and acoustic radiation losses in the slot are negligible at high forcing level [2]. To eliminate the vena contracta effect on the outstroke, a simple contraction on the cavity side of the slot will do. On the in-stroke, this separation can be mitigated by rounding the upstream edge of the slot with some loss of
For a reasonable design which has a rigid connection on the stationary end of the spring, structural losses are small, ζ ~ 0.25% damping ratio (Qm=200) as determined from hammer tests without the magnet field assembly. Fluid viscous losses due to the voice coil motion were found to be no greater, and possibly less than the structural losses, as determined by impedance testing under vacuum conditions using the actuator described in [4]. In contrast to the structural and fluid viscous losses, eddy current loss was found to be
4
Copyright © 2001 by ASME
the most significant internal loss. This loss is due to any metallic material moving through a magnetic field, which causes electrical currents to flow near the material surface and create resistive losses. For the actuators built in this investigation, the primary eddy currents occurred in the coil former tube. Detailed measurements of the 50 Hz actuator [4] shown in Fig. 2 gave the following breakdown of losses: structural = 0.25%, fluid viscous ~0%, eddy current losses in
wall was 0.13 mm. The backside of the piston was vented to minimize piston leakage penalty. In addition to the adjustable ring, a metal bellows section was tested in its place, providing a sealed piston arrangement. Though this improved efficiency (as expected), the packaging of bellows in the Piston
5 module packs independently controlled across span
Module frame Module pack without frame
2.54 cm
Coil
Magnet
Motor/spring sub-assembly
Figure 10. Motor/spring subassembly and module pack leading edge would result in significantly reduced piston area which would increase the required stroke. For these reasons, the metal bellows arrangement was not pursued further. Figure 9. Blade section actuator arrangement former tube 4%, eddy current losses in spring/piston assembly 1%. The breakdown between the eddy current loss locations was found by testing with a former tube fabricated from plastic (thus eliminating eddy currents in the former tube). The operational quality factor is Qm~10, much lower than needed. These eddy currents can be mitigated significantly (at some expense) by either slotting the former tubes before winding or by construction with a non-metallic material. However, because the eddy currents do not depend on frequency, the effect on quality factor is greatly reduced (but not eliminated) by operation at the higher frequency of the retreating blade stall application. In other words, the quality factor scaling is only invariant for losses which scale linearly with frequency and/or mass (like, e.g., structural losses).
The moving mass had the following percentage breakdown: voice coil = 27%, piston = 12%, and moving part of spring = 61%. The moving mass of the spring is clearly very significant and has a major influence on the required stiffness. The actuator resonance occurred as 270 Hz. Based on electrical impedance measurements, the actuator was found to have a reasonable quality factor, Qm ~ 40.
Spar
BENCHTOP ACTUATION Using the machined coil spring concept, a new bench top actuator was designed and fabricated based on a commercially available voice-coil linear motor. The design frequency was nominally 250 Hz. Figure 5 shows the assembly and a close up photograph of the spring-motor-piston assembly. In order to minimize moving mass (i.e. no flanges with screws), a threaded connection was made between the free end of the spring and the voice coil and an EB weld connection was made with the piston. The cavity wall in the area of the piston was an adjustable ring to provide trim centering with the piston. Above the adjustable ring, the remainder of the cavity was formed by an aluminum housing which also suspended the motor field assembly (permanent magnet) over the voice coil and spring. Above the housing was a contoured flow slot (as discussed above) with a 2.5 mm gap and 5 cm span. The nominal gap between piston and cavity
Module pack frame Figure 11. Module pack installation on blade spar Separating out the eddy current losses via hammer tests with and without the field assembly was not possible with this arrangement, however, the higher quality factor is consistent with the expected scaling of mechanical and eddy current losses with frequency. Operation at 2.5 mm amplitude displacement was routinely obtained without failure. Ensemble and spatially averaged hot wire measurements in the slot were obtained to quantify performance. Figure 6 shows the
5
Copyright © 2001 by ASME
neck velocity amplitude versus input electrical power, which shows that near 60 m/s amplitude was achieved. An electro-acoustic circuit
In summary, the second bench top actuation prototype successfully demonstrated the machined coil spring concept, reasonable efficiency with piston leakage, and low-loss slot design.
BLADE ACTUATION
Load (N)
In order to adapt the bench top actuator into a blade design, a rigid brace was needed for both the fixed end of the spring and the field assembly. For the bench top actuator, these locations were on opposite sides of the piston. Due to the limited space of the blade section leading edge, such an arrangement was not possible. To solve this problem, the voice coil motor was located inside the spring, hence the fixed end of the spring and the field assembly were co-located, significantly simplifying the installation and improving the compactness of the actuator. Figure 8 shows a cross section of the MN RN Deflection (mm)
CV CVB Spar
Figure 12. Typical stiffness of spring
ZACT
The voice coil temperature was monitored due to concerns of overheating causing failure in the voice-coil adhesion material. Figure 7 plots the temperature versus input power (power consumed by coil resistive heating is about half of the total input value). The results showed that the coil heating is safely below the manufacturer’s
Acceleration/Force (volt/volt)
6000 4000
Magnet QM= 81 (ζ = 0.6%)
2000 0 250 255 260 265 270 275 280 285 290 295 300 Freq - Hz
Figure 13. Hammer test results for motor/spring subassembly stated thermal limit (155 oC) despite being significantly above the continuous operating power (9 W versus 25 W). This favorable result is due to the heat transfer to the magnet, which is connected to a large heat sink, maintaining the magnet at a near constant cool temperature. A heat transfer analysis of solely the conduction from the coil across the air gap into the magnet is sufficient to avoid overheating.
UN UP
ZF
MN CV
RN
US
RL ZB
ZB
CVB CVT
MS RS RE
Figure 14. Electro-acoustic model of actuator in blade section
12000
No magnet QM= 222 (ζ =0.2%)
CVT
ZF
V BL R CS2D
PAmb
10000
RE
RL
model (similar to that shown in Fig. 10 of [4] without the backside loading) was fit to this data mainly by adjusting the discharge coefficient of the slot. A value of Kd = 1.4 was found to fit the data well as shown in the figure, confirming the low-loss nature of the slot design (Kd = 1 is near ideal) and represents the lowest value achieved to date for the current effort. Based on the output flow power, the actuator efficiency was determined to be about 20%.
8000
MS RS
motor/spring sub-assembly in the blade leading edge. Three subassemblies were grouped together in a module pack, driving a rectangular piston as shown in Fig. 9. Five module packs cover the blade section span, isolated from each via a partition. The partitions enable a spatial course variation in phase of the forcing. The intent is to investigate, in addition to uniform phase forcing, opposite phase forcing for adjacent slots since such forcing can be achieved with several other actuator options (e.g., using both sides of the piston to drive adjacent slots). The backside of the pistons are vented through slots on the top and bottom of the spar in order to minimize the acoustic loading and hence the unsealed piston leakage. Figure 10 gives a photograph of the actuator subassembly and module pack components and Fig. 11 shows a photograph of the module packs (one complete assembly on far right) installed on the blade spar. The springs were fabricated from corrosion resistance steel, heat treated for strength and, after machining, shot peened for improved surface finish/durability. The fifteen springs were individually statically tested for spring rate. The variation from the mean spring rate was plus 4% and minus 3%. Trim weighs were attached to the voice coils to account for the spring rate variation. Figure 12 gives a typical force-deformation plot which illustrates the linearity and small hysterisis of the spring (the nonlinear behavior near zero load are due to the initial sitting of the actuator top and bottom into the test rig and should be ignored).
6
Copyright © 2001 by ASME
Figure 13 shows the results of a hammer test on an individual sub-assembly with and without the motor field assembly. Without the field assembly, the structural losses were identified to be about 0.2% damping, consistent with the original bench top actuator from [4]. With the field assembly, an additional 0.4% damping is incurred, primarily due to eddy currents in the former tube. As earlier
Current efforts as the manuscript deadline approached are aimed at the final assembly of the blade section in the wind tunnel with complete actuation and instrumentation. Hot wire measurements are planned in-situ as a final check out and verification of model predictions.
Efficiency
0.20 0.15 0.10
CONCLUSIONS A compact, high-power synthetic jet actuator for a full-scale retreating blade stall wind tunnel test has been developed. The key breakthrough was the application of machined coil spring technology to the spring design. In addition, detailed attention to loss mechanisms and mitigation of these losses are crucial to the successful design. Particularly noteworthy is the new, low-loss slot design that minimizes slot separation effects and recovers dynamic head on the instroke. A bench top prototype actuator has been successfully built and tested. The design has been adapted to the leading edge of the blade section.
Design point
0.05 0 300 Input power (W)
plot of the expected performance of the actuation system based on the bench top results and the model in Fig. 14. The model indicates an efficiency of about 15%, requiring 260 Watts of electrical power to drive each module pack. Under more ideal design constraints (e.g., larger piston, less acoustic back side load or sealed piston), this efficiency could be at least doubled.
200
ACKNOWLEDGMENTS 100
0 0
20
40
60
80
uN (m/s) Figure 15. Predicted blade actuator performance
hypothesized that the eddy current does not scale with frequency, the relative contribution is significantly less for the higher frequency actuator (0.4% versus 5%). For the assembled module pack, hammer tests and electrical impedance test indicate a quality factor of Qm~5055 which is somewhat lower than the individual subassembly but still very reasonable.
This work is supported by the DARPA Microadaptive Flow Control Program (Richard Wlezian) under U.S. Army Research Office (Thomas Doligalski) Contract DAAG55-98-C-0066. Gary Boehm from HELI-CAL Products Company is acknowledged for the machined coil spring design.
REFERENCES 1.
Glezer, A., Allen, M.G., Coe, D.J., Barton, S.L., Trautman, M.A., and Wiltse, J.W., “Synthetic Jet Actuator and Applications Thereof”, U.S. Patent 5,758,823, June 2, 1998.
2.
McCormick, D.C., “Boundary Layer Separation Control with Directed Synthetic Jets”, AIAA 2000-0519, January 2000.
3.
Wygnanski, I, “Method and Apparatus for Delaying Separation of Flow from a Solid Surface”, U.S. Patent 5,209,438, May 11, 1993.
4.
Lorber, P.F., McCormick, D.C., Anderson, T.J., Wake, B.E., MacMartin, D.G., Pollack, M., Corke, T., and Breuer, K., “Rotorcraft Retreating Blade Stall Control”, AIAA 2000-2475, June 2000.
5.
“Engineering Solutions for Design & Manufacturing, Machined Springs: a Very Good Solution”, NASA Tech Briefs, Vol. 23, No. 4, April 1999.
6.
Beranek, L.L., “Acoustics”, Acoustical Society of America, Cambridge, MA, 1993.
7.
Wake, B.E., and Lurie, E.A., “Computational Evaluation of Directed Synthetic Jets for Dynamic Stall Control”, 57th American Helicopter Society 57th Annual Forum, May 2001.
7
Copyright © 2001 by ASME
ELECTRO-ACOUSTIC MODEL Due to the cavities and spar slots on the backside of the piston, the electro-acoustic model of the synthetic jet is a bit more complicated than models previously presented [2], [4]. Figure 14 shows the front and backside loading are modeled as two impedances in series (ZF, ZB) and the piston leakage as a resistor in parallel (acoustic mass of leakage gap neglected). As before, the actuator impedance (ZACT) is a second order system consisting of the moving mass, spring compliance, internal losses (Qm), and the coil resistance (mapped into the acoustic side of the circuit, see [2] for details). The expanded piston loading impedances are also shown in the figure. Critical to a successful design is to ensure that the resonant mode of the backside is out of band relative to the operational frequency. Otherwise, a large velocity oscillation will occur in the spar slots and large pressure oscillation across the piston, causing excessive losses. For the current arrangement, this mode is ~600 Hz. Figure 15 gives a
8.
Cocanower, A.B., Kline, S.J., and Johnston, J.P., “A Unified Method for Predicting the Performance of Subsonic Diffusers of Several Geometries”, Stanford University, Department of Mechanical Engineering Report PD-10, May 1965.
8
Copyright © 2001 by ASME
