Researchers are now implementingthe reactive control algorithms in laboratory experiments using MEMS technology and are expecting asignificant amount of skin-friction reduction comparabl
Trang 1microactuators for boundary-layer flow have been developed using MEMS technology [Ho and Tai, 1998].Spatial distribution of the skin friction has been measured using microsensors, and changes in boundary-layer flow due to microactuators have been thoroughly investigated Researchers are now implementingthe reactive control algorithms in laboratory experiments using MEMS technology and are expecting asignificant amount of skin-friction reduction comparable to that obtained from numerical studies.
In most numerical studies, sensors and actuators are collocated and distributed all over the tional domain In reality, sensors and actuators cannot be collocated, and actuators should be locateddownstream of the sensors Therefore, how to efficiently distribute the sensors and actuators is an impor-tant issue A second issue is how to develop a new control method that is well suited for experimentalapproach (note that most control methods investigated so far started from numerical studies) The thirdissue is the Reynolds number effect Relaminarization of turbulent flow due to control, as has beenobserved in numerical studies, might happen because of low Reynolds numbers At low Reynolds number,the dynamics of boundary-layer flow is mostly governed by near-wall phenomena, whereas it is affected notonly by the near-wall behavior but also by the fluid motion in the buffer or outer layer Therefore, variouscontrol algorithms may have to be developed depending on the Reynolds number
Bewley, T.R., Moin, P., and Temam, R (2001) “DNS-based Predictive Control of Turbulence: An Optimal
Benchmark for Feedback Algorithms,” J Fluid Mech., 447, pp 179–226.
Bushnell, D.M., and McGinley, C.B (1989) “Turbulence Control in Wall Flows,” Annu Rev Fluid Mech.,
21, pp 1–20.
Cantwell, B.J (1981) “Organized Motion in Turbulent Flow,” Annu Rev Fluid Mech., 13, pp 457–515.
Carlson, A., and Lumley, J.L (1996) “Active Control in the Turbulent Wall Layer of a Minimal Flow Unit,”
J Fluid Mech., 329, pp 341–371.
Choi, H., Moin, P., and Kim, J (1993a) “Direct Numerical Simulation of Turbulent Flow Over Riblets,”
J Fluid Mech., 255, pp 503–539.
Choi, H., Temam, R., Moin, P., and Kim, J (1993b) “Feedback Control for Unsteady Flow and Its
Application to the Stochastic Burgers Equation,” J Fluid Mech., 245, pp 509–543.
Choi, H., Moin, P., and Kim, J (1994) “Active Turbulence Control for Drag Reduction in Wall-Bounded
Flows,” J Fluid Mech., 262, pp 75–110.
Corino, E.R., and Brodkey, R.S (1969) “A Visual Investigation of the Wall Region in Turbulent Flow,”
J Fluid Mech., 37, pp 1–30.
Endo, T., Kasagi, N., and Suzuki, Y (2000) “Feedback Control of Wall Turbulence with Wall
Deformation,” Int J Heat Fluid Flow, 21, pp 568–575.
Gad-el-Hak, M (1994) “Interactive Control of Turbulent Boundary Layers: A Futuristic Overview,” AIAA
J., 32, pp 1753–1765.
Gad-el-Hak, M (1996) “Modern Developments in Flow Control,” Appl Mech Rev., 49, pp 365–380.
Gad-el-Hak, M., and Blackwelder, R.F (1989) “Selective Suction for Controlling Bursting Events in a
Boundary Layer,” AIAA J., 27, pp 308–314.
Hamilton, J.M., Kim, J., and Waleffe, F (1995) “Regeneration Mechanisms of Near-Wall Turbulence
Structures,” J Fluid Mech., 287, pp 317–348.
Trang 2Ho, C.-H., and Tai, Y.-C (1996) “Review: MEMS and Its Applications for Flow Control,” J Fluids Eng.,
118, pp 437–446.
Ho, C.-M., and Tai, Y.-C (1998) “Micro-Electro-Mechanical Systems (MEMS) and Fluid Flows,” Annu.
Rev Fluid Mech., 30, pp 579–612.
Jacobson, S.A., and Reynolds, W.C (1998) “Active Control of Streamwise Vortices and Streaks in
Boundary Layers,” J Fluid Mech., 360, pp 179–212.
Kang, S., and Choi, H (2000) “Active Wall Motions for Skin-Friction Drag Reduction,” Phys Fluids, 12,
pp 3301–3304
Kim, H.T., Kline, S.T., and Reynolds, W.C (1971) “The Production of Turbulence Near a Smooth Wall in
a Turbulent Boundary Layer,” J Fluid Mech., 50, pp 133–160.
Kim, J., Moin, P., and Moser, R (1987) “Turbulence Statistics in Fully Developed Channel Flow at Low
Reynolds Number,” J Fluid Mech., 177, pp 133–166.
Kline, S.J., Reynolds, W.C., Schraub, F.A., and Runstadler, P.W (1967) “The Structure of Turbulent
Boundary Layers,” J Fluid Mech., 30, pp 741–774.
Koumoutsakos, P (1999) “Vorticity Flux Control for a Turbulent Channel Flow,” Phys Fluids, 11, pp 248–250.
Kravchenko, A.G., Choi, H., and Moin, P (1993) “On the Relation of Near-Wall Streamwise Vortices to
Wall Skin Friction in Turbulent Boundary Layer,” Phys Fluids A, 5, pp 3307–3309.
Lee, C., Kim, J., Babcock, D., and Goodman, R (1997) “Application of Neural Networks to Turbulence
Control for Drag Reduction,” Phys Fluids, 9, pp 1740–1747.
Lee, C., Kim, J., and Choi, H (1998) “Suboptimal Control of Turbulent Channel Flow for Drag
Reduction,” J Fluid Mech., 358, pp 245–258.
Moin, P., and Bewley, T.R (1994) “Feedback Control of Turbulence,” Appl Mech Rev., 47, pp S3–S13.
Rathnasingham, R., and Breuer, K.S (1997) “System Identification and Control of a Turbulent Boundary
Layer,” Phys Fluids, 9, pp 1867–1869.
Robinson, S.K (1991) “Coherent Motions in the Turbulent Boundary Layer,” Annu Rev Fluid Mech., 23,
pp 601–639
Walsh, M.J (1982) “Turbulent Boundary Layer Drag Reduction Using Riblets,” AIAA Paper No 82-0169,
AIAA, Washington, DC
Trang 3Toward MEMS Autonomous Control
of Free-Shear Flows
15.1 Introduction .15-1 15.2 Free-Shear Flows: A MEMS Control Perspective .15-2
15.3 Shear-Layer MEMS Control System Components and
Issues .15-3
Sensors • Actuators • Closing the Loop: The Control Law
15.4 Control of the Roll Moment on a Delta Wing .15-9
Sensing • Actuation • Flow Control • System Integration
15.5 Control of Supersonic Jet Screech .15-17
Sensing • Actuation • Flow Control • System Integration
15.6 Control of Separation over Low-Reynolds-Number
Wings 15-29
Sensing • Flow Control
15.7 Reflections on the Future .15-32
Interest in the application of microelectromechanicalsystems (MEMS) technology to flow control and nostics started around the early 1990s During this relatively short time, there have been a handful attemptsaimed at using the new technology to develop and implement reactive control of various flow phenom-ena The ultimate goal of these attempts has been to capitalize on the unique ability of MEMS to inte-
diag-grate sensors, actuators, driving circuitry, and control hardware in order to attain autonomous active flow
management To date, however, there remains to be a demonstration of a fully functioning MEMS flowcontrol system whereby the multitude of information gathered from a distributed MEMS sensor array is suc-cessfully processed in real time using on-chip electronics to produce an effective response by a distributedMEMS actuator array
Notwithstanding the inability of the research efforts in the 1990s to realize autonomous MEMS controlsystems, the lessons learned so far are valuable in understanding the strengths and limitations of MEMS
in flow applications In this chapter, the research work aimed at MEMS-based autonomous control offree-shear flows over the past decade is reviewed The main intent is to use the outcome of these efforts
as a telescope to peek through and project a vision of future MEMS systems for free-shear flow control
15-1
Ahmed Naguib
Michigan State University
Trang 4The presentation of the material is organized as follows: first, important classifications of free-shear flowsare introduced in order to facilitate subsequent discussions; second, a fundamental analysis concerningthe usability of MEMS in different categories of free-shear flows is provided This is followed by an out-line of autonomous flow control system components, with a focus on MEMS in free-shear flow applica-tions and related issues Finally, the bulk of the chapter reviews prominent research efforts in the 1990s,leading to a vision of future systems.
Free-shear flows refer to the class of flows that develop without the influence imposed by direct contactwith solid boundaries However, in the absence of thermal gradients, nonuniform body-force fields, orsimilar effects, the vorticity in free-shear flows is actually acquired through contact with a solid boundary
at one point in the history of development of the flow The “free-shear state” of the flow is attained when
it separates from the solid wall, carrying with it whatever vorticity was contained in the boundary layer atthe point of separation The mean velocity profile of the shear layer at the point of separation is inflectional,and hence is inviscidly unstable; that is, extremely sensitive to small perturbations if excited at the appropri-ate frequencies This point is of fundamental importance to MEMS-based control
Since MEMS devices are micron in scale they are only capable of delivering proportionally small energieswhen used as actuators Therefore, if it were not for the high sensitivity of free-shear flows to disturbances
at the point of separation, there would be no point in attempting to use MEMS actuators for shear-layercontrol Moreover, active control of free-shear flows using MEMS as a disturbance source must be applied
at or extremely close to the point of separation The same statement can be extended to the more powerfulconventional actuators, such as glow-discharge, large-scale piezoelectric devices, large-scale flaps, etc., if high-gain or efficient control is desired
From a control point of view, it is useful to classify free-shear flows according to whether the separationline is stationary or moving Stationary separation line (SSL) flows include jets, single- or two-stream shearlayers, and backward facing step flows In these flows, separation takes place at the sharply defined trailingedge of a solid boundary at the origin of the free-shear flow On the other hand, moving separation line(MSL) flows include dynamic separation over pitching airfoils and wings, periodic flows through compres-sors, and forward facing step flows The separation point in the latter, although steady on average, jitters due
to the lack of a sharp definition of the geometry at separation (such as the nozzle lip for the shear layer rounding a jet)
sur-Whether the free-shear flow to be controlled is of the SSL or MSL type is extremely important, not onlyfrom the point of view of control feasibility and ease but also in terms of the need for MEMS versus conven-tional technology In particular, in SSL flows, actuators — whether MEMS or conventional — can be locateddirectly at the known point of separation where they would be effective As the operating conditions change(for example, through a change in the speed of a jet) the same set of actuators can be used to affect the desiredcontrol In contrast, in an MSL flow, the instantaneous location of the separation line has to be known andonly actuators located along this line should be used for control
To explain further, consider controlling dynamic separation over a pitching two-dimensional airfoil Asthe airfoil is pitched from, say, zero to a sufficiently large angle of attack, a separated shear layer is formed.The separation line of this shear layer moves with the increasing angle of attack To control this separat-ing flow using MEMS or other conventional actuators, it is necessary to track the location of the sepa-ration point in real time in order to activate only those actuators that are located along the separationline This would require distributed, or array, measurements on the surface of the airfoil Furthermore, sincethe separation-line-locating algorithm is likely to employ spatial derivatives of the surface measurements,the surface sensors must have high spatial resolution and be packed densely Therefore, in MSL problems, itseems inevitable to use MEMS sensor arrays if autonomous control is to be successful This is believed to
be true regardless of whether MEMS or conventional technology is used for actuation
When considering MEMS control systems, another useful classification is that associated with thecharacteristic size of the flow With the advent of MEMS, it is now common to observe flows confined to
Trang 5domains that are no larger than a few hundred microns in characteristic size Such flows may be found
in different microdevices including pumps, channels, nozzles, turbines, and others Therefore, it is tant to distinguish between microscale (MIS) flows and their macro counterparts (MAS) In particular,there should be no ambiguity that within a microdevice, only MEMS, or perhaps even NEMS (nano electro mechanical systems) are the only feasible method of control
impor-A good example of microdevices where free-shear flows may be encountered is the MIT microengineproject [Epstein et al., 1997] In the intricate, yet complex devices engineered in this project, shear layers exist
in the microflows over the stator and rotor of the compressor and turbine and in the sudden expansionleading to the combustion chamber Whether those shear layers and their susceptibility to excitation withinthe confines of the microdevice mimic that of macroscale shear layers is yet to be established However, asindicated above, if control is to be exercised, the limitation imposed by the scale of the device dictates that anyactuators and or sensors have to be as small as, if not smaller than, the device itself
Active control of MIS flows is an area that is yet to be explored This is in part because of the fairly shorttime since the interest in such flows started Therefore, further discussion of the subject would be appro-priately left until sufficient literature is available
Issues
To facilitate subsequent discussion of the research pertaining to MEMS-based control of free-shear flows, thedifferent components of the control system will be discussed and analyzed here Of particular importance arethe issues specific to utilization of MEMS technology to realize the control system components.Figure 15.1
displays a general functional block diagram for feedback control systems in SSL and MSL flows As seenfrom Figure 15.1, for both types of flows the information obtained from surface-mounted sensor arrays isfed to a flow-field estimator in order to predict the state of the flow field being controlled If the desiredflow field and deviations from it can be defined in terms of a signature measurable at the surface, the flowestimator may be bypassed altogether Any difference between the measured and desired flow states isused to drive surface mounted actuators in a manner that would force the flow toward the desired state
In the case of the MSL flow, the current location of the separation line must also be identified and fed tothe controller in order to operate the actuator set nearest to or at the position of separation (see Figure15.1, bottom)
mal component The former has one component in the streamwise (x) direction and the other in the spanwise (z) direction.
In addition to surface stresses, near-wall measurements of flow velocity may be achieved nonintrusivelyvia optical means An example of such system is currently being developed by Gharib et al (1999), who areutilizing miniature diode lasers integrated with optics in a small package to develop mini-LDAs (laserdoppler anemometers) for measurements of the wall shear and near-wall velocity With the continued
Trang 6miniaturization of components, it is not difficult to envision an array of MEMS-based LDA sensors deployedover surfaces of aero- and hydro-dynamic devices Of course, with LDA sensors flow seeding is necessary.Hence, such optical techniques may be practical only for flows where natural contamination is present orwhen practical means for seeding the flow locally in the vicinity of the measurement volume can be devised.Whereas surface sensors for MSL flows should cover the surface surrounding the flow to be controlled,those for SSL flows typically would be placed at the trailing edge However, in certain instances, placement
of the sensors downstream of the trailing edge may be feasible Examples include sudden expansion andbackward facing step flows (Figure 15.1, top) and jet flows, where a sting may be extended at the center
of the jet for sensor mounting
15.3.1.2 Sensor Characteristics
Properly designed and fabricated MEMS sensors should possess very high spatial and temporal tion because of their extremely small size Therefore, when considering the response of MEMS sensors forsurface measurements, only the sensitivity and signal-to-noise ratio (SNR) are of concern The need forhigh sensitivity and SNR is particularly important in detecting separation because the wall shear values arenear zero in the vicinity of the separation line (τw⫽ 0 at separation for steady flows) Also, hydrodynamicsurface pressure fluctuations are typically small for low-speed flows and require microphone-like sensitivitiesfor measurements
resolu-Flow field estimator
Flow field estimator
FIGURE 15.1 Conceptual block diagram of autonomous control systems for SSL (top) and MSL (bottom) flows.
Trang 7The concern regarding MEMS sensor sensitivity does not include indirect measurements of surface shear,such as that conducted using thermal anemometry, which can be achieved with higher sensitivity usingMEMS sensors than using conventional sensors (e.g., [Liu et al., 1994; and Cain et al., 2000]) On the otherhand, both direct measurements of the surface shear (using a floating element) and pressure measurements(through a deflecting diaphragm) rely on the force produced by those stresses acting on the area of the sen-sor Since typical MEMS sensor dimensions are less than 500 µm on the side, the resulting force is extremelysmall Because of this, no MEMS pressure sensor currently is known to have a sensitivity comparable to that
of 1/8⬙ capacitive microphones [Naguib et al., 1999a] Also, notwithstanding the creativity involved in oping a number of floating element designs and detection schemes [Padmanabhan et al., 1996; and Reshotko
devel-et al., 1996], the signal-to-noise ratio of such sensors remains significantly below that achievable with mal sensors Because of the direction reversing nature of separated flows, however, thermal shear sensors gen-erally are not too useful in MSL flows Direct or other-direction-sensitive sensors are required for conductingappropriate measurements in such flows
ther-15.3.1.3 Nature of the Measurements
Fundamentally, the information inferred from measurements of the wall-shear stress and near-wall velocitydiffers from that obtained from the surface pressure The latter is known to be a global quantity that is influ-enced by both near-surface as well as remote flow structures On the other hand, measurements of the shearstress and flow velocity provide information concerning flow structures that are in direct contact with thesensors
Because of the global nature of pressure measurements, pressure sensors are probably the most suitabletype of sensors for conducting measurements at the trailing edge in SSL flows More specifically, in most
if not all instances involving SSL flows, the control objective is aimed at controlling the flow downstream ofthe trailing edge Therefore, local measurements of the wall shear and velocity would not be of great use inpredicting the flow structure downstream of the trailing edge A possible exception is when hydrodynamicfeedback mechanisms are present, as is the case in a backward facing step where some structural influencesdownstream are naturally fed back to the trailing edge In such flows, however, the instability of the shear
layer may be absolute rather than convective [Huerre and Monkewitz, 1990] Fiedler and Fernholz (1990)
suggested that absolutely unstable flows are less susceptible to local periodic excitation such as that cussed here Control provisions aimed at blocking the hydrodynamic feedback loop seem to be moreeffective in such cases
dis-Another important issue concerning the nature of surface pressure measurements is that they inevitablycontain contributions from hydrodynamic as well as acoustic sources The latter could be either a conse-quence of the flow itself (such as jet noise) or of the environment, emanating from other surroundingsources that are not related to the flow field Typically, when the sound is produced by the flow, knowledge
of the general characteristics of the acoustic field (direction of propagation, special symmetries, etc.) enablesseparation of the hydrodynamic and acoustic contributions to pressure measurements In general, attentionmust be paid to ensure that the appropriate component of the pressure measurements is extracted
15.3.1.4 Robustness and Packaging
Perhaps the primary concern for the use of MEMS in practical systems is whether the minute, fairly fragiledevices can withstand the operating environment One solution is to package the devices in isolation fromtheir environment This solution has been adopted, for example, in the commercially available MEMSaccelerometers from Analog Devices, Inc Such a solution, although possibly feasible for the mini-LDA systems discussed above, in general is not useful for flow applications Most wall-shear and wall-pressuresensors and all actuators must interact directly with the flow Therefore, the ability of the minute devices
to withstand harsh, high-temperature, chemically reacting environments is of concern Also, the possibilityfor mechanical failure during routine operation and maintenance must be accounted for
15.3.1.5 Ability to Integrate with Actuators and Electronics
Although many types of MEMS sensors have been fabricated and characterized for use in flow tions, most of these sensors have been designed for isolated operation individually or in arrays If autonomous
Trang 8applica-control is to be achieved using the unique capability of MEMS for integration of components, the sensor rication technology must be compatible with that of the actuators and the circuitry More specifically, if aparticular fabrication sequence is successful in constructing a pressure sensor with certain desired charac-teristics, the same sequence may be unusable when the sensors are to be integrated with actuators on the samechip Examples of the few integrated MEMS systems that were developed for autonomous control include thefully integrated system used by Tsao et al (1997) for controlling the drag force in turbulent boundary layersand the actuator-sensor systems from Huang et al (1996) aimed at controlling supersonic jet screech (dis-cussed in detail later in this chapter).
fab-Tsao et al (1997) utilized a CMOS compatible technology to fabricate three magnetic flap-type actuatorsintegrated with 18 wall-shear stress sensors and control electronics In contrast, Huang et al (1996 and1998) developed hot-wire and pressure sensor arrays integrated with resonant electrostatic actuators Typi-cally two actuators were integrated with three to four sensors in the immediate vicinity of the actuators
15.3.2 Actuators
15.3.2.1 Actuator Types and Receptivity
There seems to be a multitude of possible of ways to excite a shear layer that differ in their nature of tion (mechanical, fluidic, acoustic, thermal, etc.), relative orientation to the flow (e.g., tangential versus lateral actuation), domain of influence (local versus global), and specific positioning with respect to the shearlayer Given the wide range of actuator types, it is confusing to choose the most appropriate type of actua-tion for a particular flow control application Because of the micron-level disturbances introduced by MEMSactuators, however, it is not ambiguous that the actuation scheme providing “the most bang for the money”should be utilized; that is, the type of actuation that is most efficient or to which the flow has the largest
excita-receptivity The ambiguity in picking and choosing from the list of actuators is for the most part due to
our lack of understanding of the receptivity of flows to the different types of actuation
When selecting a MEMS actuator, or a conventional actuator for that matter, for flow control, one is facedwith a few fundamental questions What type of actuator will achieve the control objective with minimuminput energy? That is, what type of actuator produces the largest receptivity? Is it a mechanical, fluidic,acoustic, or other type of actuator? If mechanical, should it oscillate in the normal, streamwise, or spanwisedirection? What actuator amplitude is needed to generate a certain flow-velocity disturbance magnitude? Inthe vicinity of the point of separation, is there an optimal location for the chosen type of actuator? All ofthese questions are fundamental not only to the design of the actuation system but also to the assessment
of the feasibility of using MEMS actuators to accomplish the control goal
15.3.2.2 Forcing Parameters
Perhaps one of the most limiting factors in the applicability of MEMS actuators in flow control is theirmicron-size forcing amplitude Typical mechanical MEMS actuators are capable of delivering oscillationamplitudes ranging from a few microns to tens of microns In shear flows evolving from the separation of
a thin laminar boundary layer, such small amplitudes can be comparable to the momentum thickness ofthe separating boundary layer, resulting in significant flow disturbance This is particularly true in SSL flows,where the actuator location can be maintained in the immediate vicinity of the separation point
To appreciate the susceptibility of shear-layers to very low-level actuation, it is instructive to consider anorder of magnitude analysis For example, consider a situation where it is desired to attenuate or eliminatenaturally existing two-dimensional disturbances in a laminar incompressible single-stream shear layer.One approach to attain this goal is through “cancellation” of the disturbance during its initial linear growth
phase, where the streamwise development of the instability velocity amplitude, or rms, is given by:
where, u⬘ is an integral measure of the instability amplitude at streamwise location x from the point of separation of the shear layer (trailing edge in this example), u⬘ o is the initial instability amplitude, x ois the
Trang 9virtual origin of the shear layer, and α iis the imaginary wavenumber component, or spatial growth rate,
of the instability wave
To cancel the flow instability, a periodic disturbance at the same frequency but 180° out of phase may
be introduced at some location x ⫽ x* near the point of separation of the shear layer The amplitude of
the disturbance introduced by the control should be of the same order of magnitude as that of the ral flow instability at the location of the actuator This may be estimated from Equation (15.1) as follows:
The location of the actuator should be less than one to two wavelengths (λ) of the flow instability stream of the trailing edge to be within the linear region This allows superposition of the control andnatural instabilities, leading to cancellation of the latter At the end of the linear range (approximately2λ ), the natural instability amplitude may be calculated from Equation (15.1) as:
Dividing Equation (15.2) by Equation (15.3), one may obtain an estimate for the required actuator-induceddisturbance amplitude in terms of the flow instability amplitude at the end of the linear growth zone:
The instability amplitude at the end of the linear region may be estimated from typical amplitude
satu-ration levels (about 10% of the freestream velocity, U) Also, if the natural instability corresponds to the
most unstable mode, its frequency would be given by (e.g., [Ho and Huerre, 1984]):
where St is the Strouhal number and θ is the local shear layer thickness For the most unstable mode, the instability convection speed is equal to U/2 and ⫺α iθ⫽ 0.1 (see [Ho and Huerre, 1984], for instance).Using this information and writing the frequency in terms of the wavelength and convection velocity,Equation (15.5) reduces to:
for the most unstable mode
Now, for the sake of the argument, assuming the actuator to be located at the shear layer separation point
(x* ⫽ 0), estimating u⬘(2λ ) as 0.1U, and substituting from Equation (15.6) in Equation (15.4), one obtains:
where R is the receptivity coefficient and u⬘ act. is the actuation amplitude Thus, if R is a number of order
one, the required actuator velocity amplitude for exciting flow structures of comparable strength to the
natural coherent structures in a high-speed shear layer (say, U ⫽ 100 m/s) is equal to 1 cm/s (i.e., of the
order of a few cm/s) If the actuator can oscillate at a frequency of 10 kHz (easily achievable with MEMS),the corresponding actuation amplitude is only about a few microns!
In MSL flows, the ability of the sensor array and associated search algorithm to locate the instantaneousseparation location of the shear layer and the actuator-to-actuator spacing may not permit flow excitation
Trang 10as close to the separation point as desired In this case, even for extremely thin laminar shear layers, strongeractuators may be needed to compensate for the possible suboptimal actuation location MEMS actuatorsthat are capable of delivering hundreds of microns up to order of 1 mm excitation amplitude have beendevised These include the work of Miller et al (1996) and Yang et al (1997).
Although the more powerful, large-amplitude actuators provide much needed “muscle” to the miniaturedevices, they nullify one of the main advantages of MEMS technology: the ability to fabricate actuators thatcan oscillate mechanically at frequencies of hundreds of kHz Traditionally, high-frequency (few to tens ofkHz) excitation of flows was possible only via acoustic means With the ability of MEMS to fabricatedevices with “microinertia,” it is now possible to excite high-speed flows using mechanical devices (e.g., see[Naguib et al., 1997])
To estimate the order of magnitude of the required excitation frequency, the frequency of the linearlymost unstable mode in two-dimensional shear layers may be used This frequency may be estimated fromEquation (15.5) Using such an estimate, it is straightforward to show that the most unstable mode fre-quency for typical high-speed MAS shear layers, such as that in transonic and supersonic jets, is of theorder of tens to hundreds of kHz On the other hand, if one considers a microscale (MIS) shear layer with
a momentum thickness in the range of one to 10 microns and a modest speed of 1 m/s, the ding most unstable frequency is in the range of 1.6–16 kHz Of course, this estimate assumes the MISshear layer instability characteristics are similar to those of the MAS shear layer, an assumption that awaitsverification
correspon-An inherent characteristic of high-frequency MEMS actuators with tens of microns oscillation amplitudes
is that they tend to be of the resonant type Moreover, the Q factor of such large-amplitude onators tends to be large Therefore, these high-frequency actuators are typically useful only at or very close
microres-to the resonant frequency of the actuamicrores-tor This is a limiting facmicrores-tor not only from the perspective of shearlayer control under different flow conditions, but also when multimode forcing is desired For instance,Corke and Kusek (1993) have shown that resonant subharmonic forcing of the shear layer surrounding anaxi-symmetric jet leads to a substantial enhancement in the growth of the shear layer To implement thisforcing technique it was necessary to force the flow at two different frequencies simultaneously using anarray of miniature speakers Although the modal shapes of the forcing employed by Corke and Kusek (1993)can be implemented easily using a MEMS actuator array distributed around the jet exit, only one frequency
of forcing can be targeted with a single actuator design as discussed above A possible remedy would usethe MEMS capability of fabricating densely packed structures to develop an interleaved array of two different actuators with two different resonant frequencies
15.3.2.3 Robustness, Packaging and Ability to Integrate With Sensors and Electronics
Similar to sensors, robustness of the MEMS actuators is essential if they are to be useful in practice In fact,actuators tend to be more vulnerable to adverse effects of the flow environment than sensors are, as theytypically protrude farther into the flow, exposing themselves to higher flow velocities, temperatures, forces,etc Additionally, as discussed earlier, the fabrication processes of the actuators must be compatible withthose of the sensors and circuitry if they are to be packaged together into an autonomous control system
15.3.3 Closing the Loop: The Control Law
Perhaps one of the most challenging aspects of realizing MEMS autonomous systems in practice is one that is not related to the microfabrication technology itself Given an integrated array of MEMS sensors andactuators that meets the characteristics described above, a fundamental question arises How should theinformation from the sensors be processed to decide where, when, and how much actuation should beexercised to maintain a desired flow state? That is, what is the appropriate control law?
Of course, to arrive at such a control law, one needs to know the response of the flow to the range of sible actuation This is far from being a straightforward task, however, given the nonlinearity of the systembeing controlled: the Navier–Stokes Equations Moin and Bewley (1994) provide a good summary of variousapproaches that have been attempted to develop control laws for flow applications Detailed discussion of
Trang 11pos-the topic is not part of this chapter, and it is mentioned in passing here only to underline its significance forimplementation of autonomous MEMS control systems.
This section discusses a MEMS system aimed at realizing autonomous control of the roll moment acting
on a delta wing It is based on the work of fluid dynamics researchers from the University of CaliforniaLos Angeles (UCLA) in collaboration with microfabrication investigators from the California Institute ofTechnology (Caltech) The premise of the control pursued by the UCLA/Caltech group is based on the dom-inating influence of the suction-side vortices of a delta wing on the lift force In particular, when a delta wing
is placed at a high angle of attack, the shear layer separating around the side edge of the wing rolls up into apersistent vortical structure A pair of these vortices (one from each side of the wing) is known to be respon-sible for generating about 40% of the lift force Plan- and end-view flow visualization images of such vor-tices are shown in Figure 15.2 Since the vortical structures evolve from a separating shear layer, theircharacteristics (e.g., location above the wing and strength) may be manipulated indirectly through alter-ation of the shear layer at or near the point of separation Such a manipulation may be used to change thecharacteristics of only one of the vortices, thus breaking the symmetry of the flow structure and leading
to a net rolling moment
The shear layer separating from the edge of the delta wing is thin (order of 1 mm for the UCLA/Caltechwork) and very sensitive to minute changes in the geometry Therefore, as discussed earlier in this chapter, theuse of microactuators to alter the characteristics of the shear layer and ultimately the vortical structure has agood potential for success Furthermore, when the edge of the wing is rounded rather than sharp, the specificseparation point location will vary with the distance from the wing apex, the flow velocity, and the position
of the wing relative to the flow Therefore, a distributed sensor–actuator array is needed to cover the areaaround the edge of the delta wing for detection of the separation line and actuation in its immediate vicinity
15.4.1 Sensing
To detect the location of the separation line around the edge of the delta wing, the UCLA/Caltech group lized an array of MEMS hot-wire shear sensors The sensors, which are described in detail by Liu et al (1994),consisted of 2 µm wide ⫻ 80 µm long polysilicon resistors that were micromachined on top of an evacu-ated cavity (a SEM view of one of the sensors is provided in Figure 15.3) The vacuum cavity provided ther-mal isolation against heat conduction to the substrate in order to maximize sensor cooling by the flow Theresulting sensitivity was about 15 mV/Pa, and the frequency response of the sensors was 10 kHz
uti-FIGURE 15.2 End (left) and plan (right) visualization of the flow around a delta wing at high angle of attack.
Trang 12Because of directional ambiguity of hot-wire measurements and the three-dimensionality of the aration line, it was not possible to identify the location of separation from the instantaneous shear-stressvalues measured by the MEMS sensors Instead, Lee et al (1996) defined the location of the separationline as that separating the pressure- and suction-side flows in the vicinity of the edge of the wing The dis-
sep-tinction between the pressure and suction sides was based on the rms level of the wall-shear signal This
was possible because the unsteady separating flow on the suction side produced a highly fluctuating shear signature in comparison to the more steady attached flow on the pressure side
wall-A typical variation in the rms value of the wall-shear sensor is shown as a function of the position
around the leading edge of the wing in Figure 15.4 Note that the position around the edge is expressed interms of the angle from the bottom side of the edge, as demonstrated by the inset in Figure 15.4 Because
the rms is a time-integrated quantity, the detection criterion was primarily useful in identifying the average
location of separation In a more dynamic situation where, for example, the wing is undergoing a pitchingmotion, different criteria or sensor types should be used to track the instantaneous location of separation
To map the separation position along the edge of the wing, Ho et al (1998) utilized 64 hot-wire sensorsthat were integrated during the fabrication process on a flexible shear stress skin This 80 µm thick skin,shown in Figure 15.5, covered a 1 ⫻ 3 cm2area and was wrapped around the curved leading edge of the delta
wing Using the rms criterion discussed in the previous paragraph, the separation line was identified for
different flow speeds A plot demonstrating the results is given in Figure 15.6, where the angle at whichseparation occurs (see inset in Figure 15.4 for definition of the angle) is plotted as a function of the distancefrom the wing apex As seen from the figure, the separation line is curved, demonstrating the three-dimensional nature of the separation Additionally, the separation point seems to move closer to the pressureside of the wing with increasing flow velocity (more details may be found in [Jiang et al., 2000])
15.4.2 Actuation
Two different types of actuators were fabricated for use in the MEMS rolling-moment control system: (1)magnetic flap actuators, and (2) bubble actuators The electromagnetic driving mechanism of the formerwas selected because of its ability to provide larger forces and displacements than the more common elec-trostatic type Passive as well as active versions of the magnetic actuator were conceived In the passivedesign, a 1 ⫻ 1 mm2ploysilicon flap (Figure 15.7) was supported by two flexible straight beams on a sub-strate A 5 µm thick permalloy (80% Ni and 20% Fe) was electroplated on the surface of the 1.8 µm thick flap
FIGURE 15.3 SEM image of the UCLA/Caltech shear stress sensor.
Trang 130 2 4 6 8 10
MEMS shear sensor
Wing edge
°
FIGURE 15.4 Distribution of surface-shear stress rms values around the edge of the delta wing.
FIGURE 15.5 UCLA/Caltech flexible shear-stress sensor array skin.
0 20 40 60 80 100
Trang 14The permalloy layer caused the flap to align itself with the magnetic field lines of a permanent magnet.Hence, it was possible to move the flap up or down by rotating a magnet embedded inside the edge of thewing as seen in Figure 15.8 The actuator is described in more detail in Liu et al (1995).
A photograph of the active flap actuator is shown in Figure 15.9 The construction of this device is erally the same as that of the passive one, except for a copper coil deposited on the silicon nitride flap Atime varying current can be passed through the coil to modulate the flap motion around an average posi-tion (determined by the permalloy layer electroplated on the flap and the magnetic field imposed by anexternal permanent magnet) The actuator response was characterized by Tsao et al (1994) The resultsdemonstrated the ability of the actuator to produce tip displacements of more than 100 µm at frequen-cies of more than 1 kHz
gen-To develop more robust actuators that are not only useable in wind tunnels but also in practice,Grosjean et al (1998) fabricated “balloon,” or bubble, actuators The basic principle of these actuators is based
on inflating flush-mounted flexible silicon membranes using pressurized gas As seen in Figure 15.10, thegas can be supplied through ports that are integrated under the membrane during the microfabricationprocess When, inflated, the bubbles can extend to heights close to 1 mm.Figure 15.11 demonstrates bubbleinflation with increasing pressure
Figures 15.12a through 15.12d display measurements of the change in the rolling moment as a function
of the location of the actuator around the leading edge of the wing Each of the four plots in Figure 15.12
FIGURE 15.7 SEM view of passive-type flap actuators.
Trang 15represents data acquired at a different angle of attack: α ⫽ 20°, 25°, 30°, and 35° for plots a through d tively Also, different lines represent different Reynolds numbers.
respec-The results shown in Figure 15.12were obtained for the magnetic flap actuators However, similar resultswere also produced using bubble actuators [Ho et al., 1998] In all cases, the actuators were deflected by
2 mm only on one side of the wing The results demonstrate a significant change in the rolling moment (up
to 40% for α ⫽ 25) for all angles of attack The largest positive roll moment (rolling toward the actuationside) change is observed around an actuator location of approximately 50° from the pressure side of thewing The location of maximum influence was found to be slightly upstream of the separation line (iden-tified earlier using the MEMS sensor array; seeFigure 15.4) as anticipated Another important feature inFigure 15.12 is the apparent collapse of the different lines, which suggests that the actuation impact isaffected very little, if any, by the changing Reynolds number
In addition to producing a net positive roll moment, the miniature actuators are also capable of ducing net negative moment at the lower angles of attack as implied by the negative peak depicted in
pro-Actuator on Actuator off
Permanent magnet
Wing edge
Actuator
FIGURE 15.8 Permanent magnet actuation mechanism inside delta wing.
FIGURE 15.9 SEM view of active-type flap actuator.
1.4 mm
2.0 mm
0.3 mm
8.6 mm 9.5 mm
Silicon rubber, 120 µm Parylene C, 1.6 µm
FIGURE 15.10 Schematic of bubble actuator details.