Sei sulla pagina 1di 6

ASK THE EXPERTS

Bernoulli, Bode, and Budgie

n this issue of IEEE Control Systems Magazine, Gilead Tadmor and Bernd Noack answer a query on active flow control.

Q. In my controls course, the professor mentioned active flow control, which he said means that feedback control is used to control a fluid. It sounds interesting, but it also seems impossiblehow can you control a fluid? Doesnt it just flow according to what the NavierStokes equations say? Also, I saw something on passive flow control. What does that mean? Thanks for your help. Gilead and Bernd. These are good questions, and you are not alone in asking them. The fact that flow control blends elements of fluid dynamics and control theory naturally leads to conflicts between the conceptual worlds, nomenclatures, and thinking traditions in the two disciplines. Those accustomed to standard systems and control nomenclature, vocabulary, and, indeed, the conceptual world that we take for granted, commonly find many flow control terms confusing. Newcomers are often baffled by lesser matters, such as the fact that the symbol u denotes the velocity field, that is, the state of a fluid flow system, and not the control function. Counterpart issues, large and small, mystify flow control novices whose background is in fluid dynamics. It might surprise us, but our attention to transient dynamics and the response to changing
Digital Object Identifier 10.1109/MCS.2010.939940 Date of publication: 16 March 2011

operating conditions, the very essence of feedback control, is often perceived as an inexplicable obsession by physicists accustomed to steady-state analysis. Yet what we have found through joint work is that it is precisely this dissonance that makes this area as rich and fresh and exciting as it is. In lieu of a full review of the motivations for fluid flow control, let us mention that this field is an outgrowth of acute needs and opportunities in areas where traditional engineering methods fall short of producing ample solutions. These needs include drastic reduction of turbulence drag in ground, air, and maritime transportation, the stabilization of clean and efficient, ultra-lean combustion, efficient harvesting of wind and water-current energy, and energy-efficient buildings. All of these applications are essential to our hopes for maintaining a healthy population and a vibrant economy, while meeting the reduction of CO2, NOx, and other emissions to sustainable levels. This list may continue with the need for flow control in microfluidic diagnostics, defense against biochemical terrorism, micro air vehicles, and more. The first chapter of [1], titled Brief History of Flow Control, offers interesting reading about the evolution of the field. Review articles in [1][6], as well as the articles mentioned below, provide a technical-level glimpse into a wide range of flow-control problems and methods used to address them. In addition, [1] reviews flow-control terminology and summarizes discussions by members of the flow-control community that were motivated by the perplexing mismatch of concepts and

notation in this cross-cultural field. In this short essay we attempt to demystify, at an intuitive level, some basic flow-control concepts as well as some key physical mechanisms by which flow control works.

HOW CAN WE CONTROL THE DYNAMICS OF FLUID FLOWS?


The generic way by which one controls fluid dynamics is by manipulating the interaction between the fluid motion and its boundaries. These interactions are at the root of the unsteady behavior of the flow over the wings of aircraft and birds, fins, buildings, mixers, turbine blades, cars, and walls. The effects of flow control and the intuitions used in its design are governed by standard fluid physics and the constitutive equations that represent it, including the Navier-Stokes and other pertinent equations. Thus, Bernoullis principle is used in places where the flow is forced to accelerate to generate low pressure, regulating the lift over an airfoil, or redirecting the flow around a car, in a desirable pattern. The essence of successful flow control is in finding receptive points and actuation patterns that enable a small effort to enforce the desired flow dynamics.

PASSIVE VERSUS ACTIVE FLOW CONTROL


Let us consider familiar examples to illustrate the basic meanings of these concepts. A sudden pressure drop, such as at the rear edge of a tractortrailer, above and behind the cab, along the entire undercarriage, and behind the wheels, causes flow separation and instability. This process is associated with drag losses due to
1066-033X/11/$26.002011IEEE

18 IEEE CONTROL SYSTEMS MAGAZINE

APRIL 2011

(a)

(b)

FIGURE 1 Schematic illustration of passive drag reduction. (a) Turbulence is caused by the sudden pressure drop and flow separation at the rear edge of a tractor-trailer, above and behind its cab, along the entire undercarriage, and behind the wheels. Drag forces can be explained by the net transfer of kinetic energy from the vehicle to flow turbulence. (b) Turbulence and drag forces are attenuated by passive add-ons, such as a front visor, cab-roof extension, cab and trailer skirts, and boat tails at the trailers rear.

the transfer of kinetic energy from the vehicle to turbulent flow structures in the surrounding air. Add-on devices, such as a rounded extension above the roof of the cab of a tractortrailer, cab and trailer skirts, a front visor, and rear-end boat-tails, schematically represented in Figure 1, attenuate turbulence by reducing areas of sharp pressure drop around the vehicle [3]. More sophisticated devices respond to boundary layer turbulence with resonance-generated traveling wave patterns over a flexible skin. These and similar devices are inspired by a drag-reduction strategy that evolved in marine mammals [7]. Spoilers, located at the rear top and under the front bumper of passenger cars, and the small vertical spoilers that are often attached to the top surface of the wings of an airplane, serve similar objectives. All these examples concern passive flow control, in the sense that they do not require powered actuators. The moving control surfaces of an airplane, namely, the rudder, elevator, and ailerons, are also used to shape flow patterns. This time the objective is to generate aerodynamic forces for lift, yaw, and pitch by modifying the timeaveraged flow patterns around the wings, as well as around the horizontal and vertical stabilizers. Control surfaces are different from passive flow-control devices in two ways, namely, control surfaces require input power, and they are manipulated by time-varying control commands. In this sense, the use of control surfaces to regulate aerody-

namic moments on an airplane is an example of active flow control.

FLOW CONTROL VERSUS FLIGHT CONTROL


Notwithstanding the argument for viewing control surfaces as flow-control devices, the term active flow control is commonly reserved for mechanisms that act at time constants that are commensurate with those of unsteady flow

are needed to fully utilize the potential flight envelopes of micro air vehicles. These scales cannot be reached with traditional control surfaces, whose bulk and energy demands become yet another serious drawback. Likewise, if we are to successfully use local control of skin turbulence as a means for reducing drag forces affecting commercial airliners, actuation must match the small length and time

The generic way by which one controls fluid dynamics is by manipulating the interaction between the fluid motion and its boundaries.
phenomena. To illustrate this point, we note that both the time constants associated with the mechanical response of the elevators, ailerons, and the rudder, as well as those governing the response of an airplane to changes in the angles of those control surface, are typically longer by orders of magnitude than the time constant of flow instabilities over wings and in their wake. Control surfaces and similar mechanical actuators are therefore generally viewed as belonging, rather, to the realm of flight control. As aircraft are miniaturized, the gap between time constants relevant for flight control and those governing flow instabilities shrinks. Control mechanisms that interact with the flow at these length and time scales scales that dominate leading boundary layer instabilities. The development of flow-control actuators that meet these needs has thus been an area of intense research for nearly two decades [8][10].

SHEAR-LAYER INSTABILITY
To understand the mechanisms exploited by active flow-control actuators, we need to first digress and say a few words about shear layers and their central role in fluid mechanics. A shear layer is an area characterized by a sharp transverse gradient in flow velocity. As illustrated by Figure 2(a), the no slip and no penetration conditions imply that the relative velocity of the fluid vanishes along a wall. Fluid velocity increases to ambient
APRIL 2011

IEEE CONTROL SYSTEMS MAGAZINE 19

Flow-control actuators typically act by injecting small, pulsating disturbances into shear layers.
levels as the distance from the wall grows. By their very nature, shear layers are prone to instability. The snapshot of a mixing layer in Figure 2(e) illustrates the generic evolution of an unstable shear layer. Here, shear is created at the confluence of two flows arriving at different velocities. The color code represents levels of vorticity, a quantity reflecting the orientation and strength of rotational motion in the flow. Instability is manifested by the emergence of small, rotating structures, that is, vortices, at the inception of the shear layer. These small structures are represented by the thin blue layer on the left (upstream) side of the plot. As the rotation field that forms one vortex affects neighboring vortices, small vortices rotate around each other and gradually aggregate into larger and larger structures, while they convect downstream (toward the right). In a parallel development, the momentum of vortices diffuses into disorganized turbulence in the ambient flow. Depleted of energy, large vortices eventually disappear. Figure 2(b) shows the two counterrotating vortices that form the recirculation bubble in the wake of a bluff body. These wake vortices are the product of a process that begins with the emergence of microscopic boundary layer vortices along the upper and lower walls, continues with their gradual aggregation and growth, and culminates with the bending and rotation of the two shear layers, once they depart into the wake. The recirculation zone is associated with a pressure drop in the immediate wake, generating a drag force on the bluff body. The Reynolds number Re 5 LUr LU 5 m n

is a scaling coefficient that reflects the ambient velocity U, the characteristic length scale L of structures of interest, the characteristic density r

of the fluid, and the reciprocal of the fluids dynamic and kinematic viscosity coefficients m and n, respectively. Higher Reynolds numbers are associated with thinner and increasingly unstable shear layers. As an illustration, Re 5 O 1102 2 for very small insect flight, Re 5 O 1105 2 for trucks in highway cruise, and Re 5 O 1 108 2 2 O 1 109 2 for commercial airliners at cruise flight. In comparison, the critical Reynolds number for the emergence of instabilities in wake and separated flows, as in Figure 2, are Re 5 O 1 10 2 , while the transition to full three dimensionality occurs at Re 5 O 1 100 2 . Considering the bluff body wake, the symmetry shown in Figure 2(b) becomes unstable at a critical Reynolds number. Beyond that point, vortices grow, alternately, at the lower and upper lips of the bluff body, until they detach and drift into the wake, illustrated by Figure 2(c). This instability is associated with much larger wake vortices and drag forces. Additionally, periodic vortex shedding induces an oscillatory lift force that can cause the bluff body to vibrate. This dynamic pattern affects vehicles on a highway, and it governs the impact of wind on tall buildings and bridges, as well as the impact of ocean streams on oil drills that reach from the ocean surface to the ocean floor.

2 0.8 0.6 0.4 0.5 0 (a) 20 10 0 10 20 0.5 0.5 0 0.5 0.5 1 1.5 (b) 2 0 2 2 0 2 (c) 4 6 1 0 1 2

2 (d)

50

100 (e)

150

200

250

FIGURE 2 Illustrations of shear layers and related instabilities. (a) An example of a laminar shear layer. Here the color code represents the gradient of the horizontal velocity component. The color map in (b)(d) represents the flow by levels of the vertical velocity. (b) A symmetric vortex pair forms in the wake of a bluff body. (c) Periodic vortex shedding in an unstable wake. (d) The evolution and shedding of leading- and trailing-edge vortices in an unstable flow over an airfoil at a high angle of attack. (e) Vortex growth and pairing in a two-dimensional mixing layer. The color code represents levels of vorticity, which is a quantity reflecting the orientation and intensity of rotational motion in the flow.
20 IEEE CONTROL SYSTEMS MAGAZINE

APRIL 2011

Figure 2(d) illustrates a similar phenomenon in the flow over an airfoil at a high angle of attack. The strong shear near the leading and trailing edges of the airfoil give rise to the alternating evolution and shedding of a large leading-edge vortex and a somewhat smaller trailing-edge vortex. The effect of this phenomenon is mixed, namely, a large and energetic leading-edge vortex enhances lift forces. This added lift is a critical benefit of flapping flight. Yet, at sufficiently high angles of attack, the airfoil stalls, that is, the vortical flow becomes stagnant and the produced lift falls sharply.

TICKLING THE FLOW: BOUNDARY LAYER ACTUATION IN OPEN AND CLOSED LOOP
Flow-control actuators typically act by injecting small, pulsating disturbances into shear layers. Examples of hardware used to create these disturbances include piezoelectric membranes in small resonance chambers that generate zero-net-flux micro-jets, surface plasma actuators, and even small-scale combustion-powered fluidic jets [8][11]. Even very low amplitude actuation, sometimes referred to as tickling the flow, can have a significant impact when its interactions with the early stages of a shear layer instability are used to or-

chestrate the process by which small vortices pair and aggregate, as the instability evolves. This orchestration may require a feedback mechanism, following the standard control-theoretic paradigm. Surprisingly, this approach may be fairly successful even in steady, open-loop implementations, with the possibility of slow adaptation of actuation amplitude and frequency to changing inflow conditions. To illustrate these three types of strategies, as well as the underlying mechanisms that make them effective, we use results from an experimental study of vortex-shedding attenuation behind a three-dimensional bluff body [12]. Both vortex shedding and strong wake asymmetry are apparent in the experimental snapshot of the natural flow shown in Figure 3(a). A cross section of the velocity field is visualized by the black arrows, and a color code represents the respective vorticity. These data were obtained by an optical technique known as particle image velocimetry (PIV).

Open-Loop Control
Periodic, open-loop actuation is commonly used as a means to attenuate flow instabilities. In the bluff-body example of Figure 3(b), periodic pulsatile actuation, represented by the red arrows, is synchronously applied at the

upper and lower bluff-body lips. The effect on the flow field, shown in Figure 3(b), is dramatic, namely, a much elongated and nearly symmetric recirculated zone is created, wake vortices are less energetic, and the pressure drop behind the bluff-body wake is reduced by 40%. To recap, periodic excitation of flow oscillations at a selected frequency can be used to attenuate far larger oscillations at another frequency. The obvious advantages of this choice include its simplicity and resistance to sensor noise and poor observability properties, which make feedback difficult. The key to understanding the mechanism that makes this control possible is the nonlinearity of fluid dynamics. Obviously, this approach would not work in a linear system, where the response to periodic actuation is merely superimposed on the natural instability. Quadratic terms in a dynamical system, such as the Navier-Stokes equation, enable the exchange of energy between oscillations at different frequencies. Cross-frequency energyflow rates are proportional to the amplitudes of the respective quadratic terms. Exciting flow structures at one frequency can therefore have the secondary effect of enhancing the rate of energy flow into those structures. That energy comes from oscillations at other

0.5 y / y / 0 0.5

0.5 0 0.5 y /

0.5 0 0.5

0.5

1 x / (a)

1.5

0.5

1 x / (b)

1.5

0.5

1 x / (c)

1.5

FIGURE 3 Experimental results of the natural and actuated flows behind a three-dimensional bluff body. The black arrows represent the optically measured velocity field. The color code represents vorticity. (a) The natural flow is characterized by large vortices, alternately shed from the upper and lower lips. (b) Synchronized, open-loop pulsatile actuation symmetrizes the near wake, reducing the pressure drop along the rear face by 40% compared with the natural flow. The actuators are represented by the red arrows located at the lower and upper lips. (c) The same improvement is obtained by a single actuator, along the top lip, when actuation is synchronized with pressure measurements at the lower lip, where the sensor is represented by the green tab.
APRIL 2011

IEEE CONTROL SYSTEMS MAGAZINE 21

A basic system-theoretic premise is that a dynamic model is the starting point in a systematic control design.
frequencies, including the undesirable, naturally unstable oscillations, which are thus attenuated. This mechanism is often mediated by the slowly varying, nonoscillatory, mean flow [13], [14]. The idea is therefore to select an actuation frequency that strikes a balance between two competing objectives. On the one hand, in order to not substitute one evil for another, the actuation frequency needs to be associated with flow structures that are naturally stable and decay without excitation. On the other hand, the response to the selected excitation frequency needs to be sufficiently energetic to enable the flow to lock in, that is, to sustain oscillations at that frequency, under acceptable, that is, low, actuation amplitude. This way, open-loop control substitutes high-amplitude natural oscillations by low-amplitude excited oscillations. The cost is the energy invested in the excitation of an inefficient frequency. Optimal actuation frequency and amplitude are commonly found by extensive experimental parametric studies. body height H, and the inflow velocity U. A change in the inflow velocity U therefore requires a change in the actuation frequency f . That is, an effective implementation does require feedback, but that feedback is relatively rudimentary; it operates at long time constants and requires information that is easy to measure and filter. To put this statement in perspective, we do not require any information about the phase of the oscillatory instability, which a linear vibration control method would require.

removed, feedback is far more agile and can respond to changing conditions within a few actuation periods. Beyond these arguments and this example, there are instabilities where the open-loop option is not available, but where simple mechanisms, such as the one described here, are feasible. Connecting with a familiar nonlinear control-theoretic concept, the problem can often be framed as that of designing a dissipative control.

MODEL-BASED DESIGN AND REDUCED-ORDER FLOW MODELS


A basic system-theoretic premise is that a dynamic model is the starting point in a systematic control design. Illustrated by the bluff-body example, where highly simplified versions of reality suffice, the question of what constitutes an ample, and a feasible model, deserves at least a few sentences. Let us step back and begin with the challenges. Kolmogorovs estimate of the number of dynamic variables per unit volume that are necessary in a numerical implementation of the Navier-Stokes equations is Re9/4. Considering rank-and-file engineering applications, where the Reynolds number commonly ranges between Re 5 O 1 104 2 and Re 5 O 1 109 2 and beyond, this estimate is truly staggering. In fact, even offline numerical simulations resort to significant structural simplifications of the NavierStokes equation to remain practically useful. Familiar model-reduction methods, such as balanced truncation, become untenable, and only approximations based on empirical or simulation data may be feasible [15]. This difficulty is compounded by the fact that linear models are inherently insufficient. Once again, the success of open-loop excitation in attenuating instabilities is a vivid demonstration of this fact. Galerkin models are obtained as projections of the NavierStokes equation on low-order modal expansions, that is, u 1 x, t 2 < u0 1 x 2 1 a ai 1 t 2 ui 1 x 2 ,
i51 N

Closed-Loop Control
Given the success of quasi-open-loop actuation, the benefits for closed-loop control need to be well established and proven to be worth the trouble. To that end, let us revisit our bluff-body study, this time with a closed-loop method. This method requires only a single actuator, say, at the top lip. A pressure gauge replaces the actuator at the lower lip. This sensor signal is processed by a robust, narrowband filter to derive reliable information on the phase of the oscillatory growth and release of vortices into the lower shear layer. To synchronize the two shear layers, pulsed actuation is applied at the upper lip with an optimized phase difference, with respect to the oscillations measured by the sensor, at the lower lip. The result, visualized in Figure 3(c), is similar to what was achieved by open-loop actuation. In particular, the pressure gains achieved by the two methods are essentially identical. The advantage of using feedback is threefold. First, using a single actuator reduces hardware cost, weight, and bulk. The second advantage is the reduction by about 40% of the required actuation energy. Finally, with the slow adaptation postulate

Slow Adaptation of Open-Loop Actuation


The optimal actuation parameters in an open-loop strategy depend on the ambient flow, represented by the Reynolds number, as well as on properties of the actuator and other elements of the flow configuration. For example, the optimal open-loop actuation frequency in the bluff-body example is characterized by the nondimensional Strouhal frequency value St 5 0.15. The Strouhal frequency St 5 fH U

depends on the dimensional actuation frequency f (in hertz), the bluff


22 IEEE CONTROL SYSTEMS MAGAZINE

APRIL 2011

and thus
N # ai 5 ci 1 a ,i, jaj 1 a qi, j, kajak. j51 j, k

Efficient expansion sets can be computed by an empirical balanced truncation, as in [15], by proper orthogonal decomposition (POD), a method that requires only a singular value decomposition of a data matrix, or by temporal filtering methods [6], [14]. The nonlinearity of the flow is reflected by this option, but new caveats emerge. Not only is the representation of actuation intrinsically difficult in this framework, but the effect of successful actuation is often reflected by making the original expansion set obsolete. Finally, we need to keep in mind that no low-order model can successfully represent the complete repertoire of feasible dynamics of the actuated flow. Any model is therefore useful only inasmuch as control policies based on it maintain the flow state within its narrow validity envelope. While these challenges and restrictions define the confines of modelbased design, there are redeeming factors that make low-order models feasible and useful. The bluff-body example illustrates some of the most common ones. When the objective of control is to regulate slowly varying, phase-independent properties, such as the mean aerodynamic forces, and when open-loop actuation is sufficient, at least over a short time horizon, design models need to resolve only the input-output relations associated with these slow dependencies. This approach is taken in diverse applications that fit this description, with black-box linear time-invariant models and adaptive variants [2], [5]. A slowly varying periodic dominance is yet another common feature. When present, once again, a slow adaptation of a linear time-invariant model may well suffice [16]. With closer connection to the underlying physics, parameterized mode-sets and mean-field Galerkin models [6] may be advanta-

geous. In fact, the bluff-body control study, which we used as an illustration, grew out of the insights gained through the latter option. Lastly, a useful di stinction is between the role of a model as a crisp encapsulation of an understanding of the dynamics of a given system and its role as a means for control implementation. In the reviewed example, the former may be a Galerkin system. The latter is a traveling wave representation, capturing the desired phas e difference between a pulsed actuation and an oscillatory sensor signal.

Award (1995), the Richard von Mises Award (GAMM, 2005), two Gallery of Fluid Motion Awards (APS-DFD), and the ANR Chaire dExcellence.

REFERENCES
[1] R. D. Joslin and D. Miller, Eds., Fundamentals and Applications of Modern Flow Control (Progress in Aeronautics and Astronautics Series). Washington, DC: AIAA, 2009. [2] R. King, Ed. Active Flow Control (Lecture Notes on Numerical Fluid Mechanics and Interdisciplinary Design). Berlin: Springer-Verlag, 2007. [3] F. Browand, R. McCallen, and J. Ross, Eds. The Aerodynamics of Heavy Vehicles. II: Trucks, Buses, and Trains (Lecture Notes in Applied and Computational Fluid Mechanics). New York: Springer-Verlag, 2009. [4] L. Cortelezzi and I. Mezic , Eds. Analysis and Control of Mixing with an Application to Micro and Macro Flow Processes. Berlin: Springer-Verlag, 2009. [5] R. King, Ed. Active Flow Control II (Lecture Notes on Numerical Fluid Mechanics and Interdisciplinary Design). Berlin: Springer-Verlag, 2010. [6] B. R. Noack, M. Morzyn ski, and G. Tadmor, Eds. Reduced-Order Modelling for Flow Control. Berlin: Springer-Verlag, 2010. [7] F. E. Fish and G. V. Lauder, Passive and active flow control by swimming fishes and mammals, Annu. Rev. Fluid Mech., vol. 38, pp. 193224, 2006. [8] A. Glezer and M. Amitay, Synthetic jets, Annu. Rev. Fluid Mech., vol. 34, pp. 503529, 2002. [9] T. C. Corke, C. L. Enloe, and S. P. Wilkinson, Dielectric barrier discharge plasma actuators for flow control, Annu. Rev. Fluid Mech., vol. 42, pp. 505529, 2010. [10] L. N. Cattafesta and M. Sheplak, Actuators for active flow control, Annu. Rev. Fluid Mech., vol. 43, 2011. [11] A. Rajendar, T. Crittenden, and A. Glezer, Characterization of the internal flow dynamics of combustion powered actuators, in Proc. 4th Flow Control Conf., Seattle, WA, 2008, AIAA Paper 2008-3760. [12] M. Pastoor, L. Henning, B. R. Noack, R. King, and G. Tadmor, Feedback shear layer control for bluff body drag reduction, J. Fluid Mech., vol. 608, pp. 161196, 2008. [13] M. Luchtenburg, B. R. Noack, B. Gnther, R. King, and G. Tadmor, A generalised mean-field model of the natural and high-frequency actuated flow around a high-lift configuration, J. Fluid Mech., vol. 623, pp. 283316, 2009. [14] G. Tadmor, O. Lehmann, B. R. Noack, and M. Morzyn ski, Mean field representation of the natural and actuated cylinder wake, Phys. Fluids, vol. 22, p. 034102, 2010. [15] C. W. Rowley, Model reduction for fluids, using balanced proper orthogonal decomposition, Int. J. Bifurcation Chaos, vol. 15, pp. 997 1013, 2005. [16] L. N. Cattafesta, Q. Song, D. R. Williams, C. W. Rowley, and F. S. Alvi, Active control of flowinduced cavity oscillations, Prog. Aerospace Sci., vol. 44, pp. 479502, 2008.

AUTHOR INFORMATION
Gilead Tadmor (tadmor@coe.neu. edu) received the B.Sc. in 1977 from Tel Aviv University and the M.Sc. in 1979 and the Ph.D. in 1984 from the Weizmann Institute of Science, all in mathematics. In 1989 he joined Northeastern University in Boston, where he is a professor of electrical and computer engineering and of mathematics. His research background is in robust and optimal control and in distributed parameter dynamical systems. His current interests include applications of reduced-order system identification and control with applications to fluid dynamics, aeroelasticity and flow control, crowd motion modeling and anomaly detection, and biomedical imaging. He can be contacted at 409 DA, Northeastern University, Boston, MA 02115 USA. Bernd R. Noack received the diplom in 1989 and doctorate in 1992 in physics from the Georg-August-University, Gttingen, Germany. He held positions at the Max-Planck-Institut, the German Aerospace Center, Gttingen University, United Technologies Research Center (UTRC), and TU Berlin (20002009). Since 2010 he has been a directeur de recherche at CNRS, located at Institute P. His current research concerns fundamental and applied issues in reduced-order modeling of fluid dynamics and turbulence and closed-loop flow control. He is the recipient of the Hugo-Denkmeier

APRIL 2011

IEEE CONTROL SYSTEMS MAGAZINE 23

Potrebbero piacerti anche