A compressible boundary layer optimal control approach

using nonlinear boundary region equations

∗1
Omar Es Sahli , Adrian Sescu†1 , Mohammed Afsar‡2 , Yuji Hattori§3 , and Makoto
Hirota¶3
1
Department of Aerospace Engineering, Mississippi State University, MS 39762
2
Department of Mechanical & Aerospace Engineering, Strathclyde University, 75 Montrose
St. Glasgow, G1 1XJ, UK
3
Institute of Fluid Science, Tohoku University, 2 Chome-1-1 Katahira, Aoba Ward, Sendai,
980-8577, Japan

It is generally agreed that high-amplitude free-stream disturbances, as well as large

surface roughness elements, excite a laminar boundary layer and engender stream-wise
oriented vortices accompanied by streaks of varying amplitudes, which lead to secondary
instabilities and ultimately to transition to turbulence. In this paper, we aim at deriving
and testing an optimal control algorithm in the context of compressible boundary layers,
in an attempt to suppress or at least limit the growth of stream-wise vortices or reduce
the wall shear stress and heat transfer. The goal is to achieve optimum control of stream-
wise vortices, which is expected to result in a reduction of vortex energy and ultimately
delay the transition from laminar to turbulent flow. Flow instabilities are introduced either
through roughness elements equally separated in the span-wise direction or via free-stream
disturbances. The compressible Navier-Stokes equations are reduced to the boundary
region equations (BRE) in a high Reynolds number asymptotic framework, wherein the
stream-wise wavelengths of the disturbances are assumed to be much larger than the span-
wise and wall-normal counterparts. Lagrange multipliers are utilized to derive the adjoint
compressible boundary region equations, by transforming the constrained optimization
problem into its unconstrained form. In the implemented optimal control strategy, the
wall transpiration velocity represents the control variable whereas the wall shear stress or
the vortex energy designates the cost functional.

I. Introduction
Flow control techniques have been the focus of many studies due to their wide range of applications as
well as the impact that they could have on the various parameters of the flow, and thus on the efficiency,
performance and noise levels. For instance, wings, engine inlets and nozzles, combustors, automobiles,
aircraft, and marine vehicles all offer suitable configurations for the implementation of such control methods.
Depending on the application and the desired result, one might want to reduce drag, increase lift, delay or
accelerate the transition, enhance mixing, postpone flow separation or even manipulate a turbulence field.
Controlling transitional or fully developed turbulent boundary layers mainly intend to achieve a reduction
in the energy carried by structures in the stream-wise direction, which take the form of high- and low-velocity
streaks originating in the near-wall region (e.g., Kendall,22 Matsubara & Alfredsson,36 or Landahl26 ). For
boundary layers over flat plates or wings, these streaks appear when the upstream roughness’ height surpasses
a specific critical value (e.g., Choudhari & Fischer,6 White,53 White et al.,54 Goldstein et al.,10–12 or Wu
& Choudhari56 ), or when the amplitude of the free-stream disturbances exceeds a certain threshold (e.g.,
∗ Research Assistant, Department of Aerospace Engineering, Mississippi State University, MS 39762; member AIAA.
† Associate Professor, Department of Aerospace Engineering, Mississippi State University, MS 39762; Associate Fellow AIAA.
‡ Chancellor Fellow, Department of Mechanical & Aerospace Engineering; AIAA member.
§ Professor, Institute of Fluid Science, Tohoku University.
¶ Assistant Professor, Institute of Fluid Science, Tohoku University.

1 of 15

American Institute of Aeronautics and Astronautics

Kendall,22 Westin et al.,52 Matsubara & Alfredsson,36 Leib et al.,28 Zaki & Durbin,59 Goldstein & Sescu,13
or Ricco et al.41 ). As for the boundary layer of flow along concave surfaces, the elongated streaks forming
as stream-wise (Görtler) vortices develop as a result of the imbalance between the centrifugal forces and the
wall-imposed pressure gradient (e.g., Gortler,14 Hall,16–18 Swearingen & Blackwelder,51 Malik & Hussaini,34
Saric,42 Li & Malik,29 Boiko et al.,2 Wu et al.,55 or Sescu et al.,44, 45 Ren & Fu,40 Dempsey et al.,8 Xu
et al.58 ). In the same context of curved walls, the higher the curvature the more rapid vortices form,
thus pronouncedly affecting the mean flow and resulting in the breakdown of laminar flow into turbulence.
Similarly, small to medium curvature also induce vortex formation which alters the mean flow and causes
the laminar flow to breakdown.
Like many other types of instabilities, stream-wise oriented vortices and their accompanying streaks are
present and very important in many engineering applications. Examples of such applications include the
flow around wings or turbo-machinery blades, and the flow in the proximity to the walls of wind tunnels or
turbofan engine intakes. Not only that these instabilities result in the transition from laminar to turbulent
flow, but this transition is also a significant source of noise in supersonic and hypersonic wind tunnels,
subsequently causing high levels of interference with the measurements of the test section (Schneider43 ),
which in turn makes it a challenging task to compare between wind tunnel measurements and real flight
conditions. With this in mind, one would want to lessen the energy associated with such stream-wise vortices,
which is expected to delay early nonlinear breakdown and the transition to turbulence. Moreover, since the
instability’s transient part is the one influencing the growth of three-dimensional instabilities, responsible for
the breakdown, the most, the control strategy must primarily focus on the restriction of the development of
such transient modes.
Although significant achievements have been made in the last decades toward understanding the phe-
nomenology behind the formation and the influence of the stream-wise vortices and streaks in incompressible
boundary layers, the progress in the compressible regime has been slow and the bypass transition remains
largely unexplored. This work is directed towards scratching the surface of the subject, even more, and to
further investigate the effects of wall transpiration in a nonlinear compressible boundary layer framework.
Wall transpiration control techniques of boundary layers can be applied via localized suction and blowing
regions, respectively, underneath low- and high-velocity streaks. The net result is a decrease in the span-wise
variation of the stream-wise velocity and, therefore, a corresponding reduction in the number and strength
of the bursting events.
Alternatively, this can be achieved through active wall control, which has been utilized in the context
of turbulent channel flow (see Choi, Moin & Kim5 ) as a technique to reduce skin friction drag. Choi et
al.5 conducted direct numerical simulations using active wall control based on wall transpiration, by placing
sensors in a sectional plane that is parallel to the wall; the technique achieved approximately 25% reduction
in frictional drag. Practically, it is very challenging and difficult, if not impossible, to install sensors in the
flow as they may interact with it and either induce new or amplify existing disturbances. Thus, to avoid
interference issues, Choi et al.5 investigated the same control algorithm but with sensors placed at the wall
with information based on the leading term in the Taylor series expansion of the vertical component of velocity
near the wall; however, this approach resulted only in a 6% reduction. Koumoutsakos24, 25 implemented a
similar feedback control algorithm informed by flow quantities at the wall. Taking vorticity flux components
as inputs to the control algorithm resulted in a more significant skin friction reduction (approximately 40%).
Lee et al.27 derived new suboptimal feedback control laws based on blowing and suction to manipulate the
flow structures in the proximity to the wall, using surface pressure or shear stress distribution (the reduction
in the frictional drag was in the range of 16-20%). Observing that the opposition control technique is more
effective in low Reynolds number turbulent wall flows, Pamies et al.38 proposed the utilization of the blowing
only at high Reynolds numbers, and by doing so they obtained significant reduction in the skin-friction drag
for these flows. Recently, Stroch et al.50 conducted a comparison between the opposition control applied in
the framework of turbulent channel flow and a spatially developing turbulent boundary layer. They found
that the rates of frictional drag reduction are approximately similar in both cases. An overview of the issues
and limitations associated with the opposition control type is given in the review article of Kim.23
Hogberg et al.19 reported the first successful relaminarization of a Reτ = 100 turbulent channel flow by
applying zero mass flux blowing and suction at the wall in the framework of linear full-state optimal control
theory. They showed that the information available in the linearized equations may be sufficient to construct
linear controllers able to relaminarize a wall turbulent flow, but this may be limited to low Reynolds number
flows.

2 of 15

American Institute of Aeronautics and Astronautics

 A number of experiments aiming to control disturbances in laminar or turbulent boundary layers by
blowing and suction have been conducted over the years. Several of them are briefly mentioned here. Gad-el-
Hak & Blackwelder9 used continuous or intermittent suction to eliminate artificially generated disturbances in
a flat-plate boundary layer. The same idea was used in the experiments conducted by Myose & Blackwelder37
to delay the breakdown of Görtler vortices. Jacobson & Reynolds20 developed a new type of actuation based
on a vortex generator to control disturbances generated by a cylinder with the axis normal to the wall, and
unsteady boundary layer streaks generated by pulsed suction. Regarding the latter, the actuation was able to
significantly reduce the span-wise gradients of the stream-wise velocity, which are known to be an important
driving force of secondary instabilities (see Swearingen & Blackwelder51 ). In the experiments of Lundell &
Alfredsson,33 stream-wise velocity streaks in a channel flow were controlled by localized regions of suctions
in the downstream, which are found to be effective in delaying secondary instabilities and consequently the
transition onset.
Optimal control in the framework of laminar or turbulent boundary layers has been utilized in a number
of studies. There are numerous studies pertaining the application of optimal control of shear flows (see the
review of Gunzburger15 or a more recent review of Luchini & Bottaro,32 although the latter is in a slightly
different context). The following studies have targeted the control of disturbances evolving in laminar or
turbulent boundary layers (e.g.1, 3, 4, 7, 19, 21, 31, 48, 49, 60 )
A relevant work in the present context is that of Joslin et al.,21 where a mathematical framework for
optimal control of disturbances in three-dimensional boundary layers based on Lagrange multipliers was
introduced; the analysis included in this work is largely based on their derivation. Optimal control of
turbulent channel flows by blowing and suction was employed previously by Bewley & Moin,1 who claimed
a 17% frictional drag reduction as a result of this scheme. Blowing and suction based optimal control was
also applied by Cathalifaud & Luchini3 to reduce the energy of disturbances in a flat-plate and a concave
boundary layer. In the study of Zuccher et al.,60 an optimal and robust control strategy was discussed and
tested in the framework of steady three-dimensional disturbances (in the form of streaks) that form in a
flat-plate boundary layer. It was based on an adjoint-based optimization technique to first find the optimal
state for given initial conditions, and then to determine what the worst initial conditions for the optimal
control are. Lu et al.31 derived an optimal control scheme within the linearized unsteady boundary region
equations which are the asymptotic reduction of the Navier-Stokes equations under the assumption of low
frequency and low stream-wise wavelength. Their study aimed at controlling both streaks developing in
flat-plate boundary layers and Görtler vortices evolving along concave surfaces. Cherubini et al.4 applied a
nonlinear optimal control strategy with blowing and suction, starting with the full Navier-Stokes equations,
and using the method of Lagrange multipliers to determine the largest decrease of the disturbance energy.
A closed-loop optimal control technique based on wall transpiration was derived and tested by Papadakis,
Lu & Ricco,39 in the framework of a flat-plate laminar boundary layer excited by freestream disturbances.
The optimal control was split into two sequences that can be obtained by marching the corresponding equa-
tions in forward and backward directions, and it was found that the feedback sequence is more important than
the feed-forward sequence. The study of Xiao & Papadakis57 employs an optimal control algorithm based
on Lagrange multipliers, aimed at delaying transition in a flat-plate boundary layer excited by freestream
vortical disturbances, is based on blowing and suction, and is derived in the framework of full Navier-Stokes
equations.

II. Basic scalings and governing equations

A. Scalings
All dimensional spatial coordinates (x∗ , y ∗ , z ∗ ) are normalized by the spanwise separation λ∗ , while the
dependent variables by their respective freestream values, except the pressure, which is normalized by the
dynamic pressure:

t∗ x∗ y∗ z∗
t̄ = ; x̄ = ; ȳ = ; z̄ = (1)
λ∗ /V∞
∗ λ∗ λ∗ λ∗

u∗ v∗ w∗ ρ∗
ū = ∗
; v̄ = ∗
; w̄ = ∗
; ρ̄ = (2)
V∞ V∞ V∞ ρ∗∞

3 of 15

American Institute of Aeronautics and Astronautics

 p∗ − p∗∞ T∗ µ∗ k∗
p̄ = ; T̄ = ; µ̄ = ; k̄ = (3)
ρ∗∞ V∞
∗2 ∗
T∞ µ∗∞ ∗
k∞
where λ∗ is the spanwise wavelength of the disturbances, (u∗ , v ∗ , w∗ ) are the velocity components, ρ∗ the
density, p∗ is pressure, T ∗ temperature, µ∗ dynamic viscosity, k ∗ thermal conductivity, and all quantities
with ∞ at the subscript represent conditions at infinity.
Reynolds number based on the spanwise separation, Mach number and Prandtl number are defined as

ρ∗∞ V∞
∗ ∗
λ V∞∗
µ∗∞ Cp
Rλ = , Ma = , Pr = (4)
µ∗∞ a∗∞ ∗
k∞
where µ∗∞ , a∗∞ and k∞∗
are freestream dynamic viscosity, speed of sound and thermal conductivity, respec-
tively, and Cp is the specific heat at constant pressure. For boundary layer flows over curved surfaces, we
define the global Görtler number as

Rλ2 λ∗
Gλ = (5)
r∗
where r∗ is the radius of the curvature.

B. Compressible Navier-Stokes equations (N-S)

For a full compressible flow, the primitive form of the Navier-Stokes equations in non-dimensional variables
are written as
Continuity equation:

 
Dρ̄ ∂ ū ∂v̄ ∂ w̄
+ρ + + =0 (6)
Dt ∂ x̄ ∂ ȳ ∂ z̄
x-momentum equation:

        
Dū ∂ p̄ 1 ∂ 2 ∂ ū ∂v̄ ∂ w̄ ∂ ∂ ū ∂v̄ ∂ ∂ w̄ ∂ ū
ρ̄ =− + µ 2 − − + µ + + µ + (7)
Dt̄ ∂ x̄ Reλ ∂ x̄ 3 ∂ x̄ ∂ ȳ ∂ z̄ ∂ ȳ ∂ ȳ ∂ x̄ ∂ z̄ ∂ x̄ ∂ z̄
y-momentum equation:

        
Dv̄ ∂ p̄ 1 ∂ 2 ∂v̄ ∂ ū ∂ w̄ ∂ ∂v̄ ∂ ū ∂ ∂v̄ ∂ w̄
ρ̄ =− + µ 2 − − + µ + + µ + (8)
Dt̄ ∂ ȳ Reλ ∂ ȳ 3 ∂ ȳ ∂ x̄ ∂ z̄ ∂ x̄ ∂ x̄ ∂ ȳ ∂ z̄ ∂ z̄ ∂ ȳ
z-momentum equation:

        
Dw̄ ∂ p̄ 1 ∂ 2 ∂ w̄ ∂ ū ∂v̄ ∂ ∂ w̄ ∂ ū ∂ ∂v̄ ∂ w̄
ρ̄ =− + µ 2 − − + µ + + µ + (9)
Dt̄ ∂ z̄ Reλ ∂ z̄ 3 ∂ z̄ ∂ x̄ ∂ ȳ ∂ x̄ ∂ x̄ ∂ z̄ ∂ ȳ ∂ z̄ ∂ ȳ
Energy equation:

      
DT̄ 1 ∂ ∂ T̄ ∂ ∂ T̄ ∂ ∂ T̄
ρ̄ = k + k + k
Dt̄ P rReλ ∂ x̄ ∂ x̄ ∂ ȳ ∂ ȳ ∂ z̄ ∂ z̄
"    2 #
∂ ū ∂v̄ ∂ w̄ 2 ∂ ū ∂v̄ ∂ w̄
2
− (γ − 1)M∞ p + + − µ + + (10)
∂ x̄ ∂ ȳ ∂ z̄ 3 ∂ x̄ ∂ ȳ ∂ z̄
"   2 #
2  2  2  2  2 
2 µ ∂ ū ∂v̄ ∂ w̄ ∂ ū ∂v̄ ∂ w̄ ∂ ū ∂v̄ ∂ w̄
+ (γ − 1)M∞ 2 +2 +2 + + + + + +
Reλ ∂ x̄ ∂ ȳ ∂ z̄ ∂ ȳ ∂ x̄ ∂ x̄ ∂ z̄ ∂ z̄ ∂ ȳ

4 of 15

American Institute of Aeronautics and Astronautics

where
D ∂ ∂ ∂ ∂
= + ū + v̄ + w̄ (11)
Dt̄ ∂ t̄ ∂ x̄ ∂ ȳ ∂ z̄
is the substantial derivative (for the what follows, we consider the steady-state case for the N-S equations,
i.e. ∂/∂ t̄ = 0). The pressure p, the temperature T and the density of the fluid are combined in the
equation of state in non-dimensional form, p̄ = ρ̄T̄ /γM∞ 2
, assuming that non-chemically-reacting flows are
considered. Other notations include the dynamic viscosity µ, Reynolds number Re = ρ∞ V∞ λ /µ based on a
∗ ∗

characteristic velocity V∞ , and a characteristic length L,the free-stream Mach number M∞ = V∞

∗ ∗
/a∗∞ . The
dynamic viscosity and thermal conductivity k is linked to the temperature using the Sutherland’s equations
in dimensionless form,

T 3/2
 
C2 Cp µ
µ= 1+ ; or µ = T b; k= (12)
T∞ T + TC2 Pr
∞

where b = 0.76, C1 = 1.458 × 10−6 , C2 = 110.4, Cp = γR/(γ − 1), γ = 1.4, and P r = 0.72 for air.

C. Compressible Nonlinear Boundary-Region Equations (BRE)

We consider a compressible flow of uniform velocity V∞ ∗
and temperature T∞ ∗
past a flat
p or curved surface.
The air is treated as a perfect gas so that the sound speed in the free-stream c∗∞ = γRT∞ ∗ , where γ =

1.4 is the ratio of the specific heats, and R = 287.05N m/(kgK) is the universal gas constant; Mach number
is assumed to be of order one. The flow is divided into four regions as in Leib et al.,28 Ricco & Wu41 or
Marensi et al.35 (see figure 1):
• Region I in proximity to the the leading edge, where the flow is assumed inviscid and the disturbances
are treated as small perturbations of the base flow.
• Region II is the boundary layer in the vicinity of the leading edge, with the thickness much smaller
than the spanwise separation associated with the freestream disturbances
• Region III is the viscous region that follows in the downstream of region II; here, the boundary layer
thickness is in the same order of magnitude as the spanwise separation
• Region IV is above region III, and the flow is assumed again inviscid since the viscous effects are
negligible

Figure 1: Fhe flow domains illustrating the asymptotic structure (Leib et al.28 ).

The focus in this paper is the region III, where the streamwise wavenumber of disturbances are expected
to be small, and the flow is governed by the boundary region equations. In this region, the compressible

5 of 15

American Institute of Aeronautics and Astronautics

boundary region equations can be derived from the full steady-state compressible N-S equations. Based in
the above assumption, the streamwise distance can be scaled as x = x̄/Rλ while the other two are the same
y = ȳ, z = z̄, and the time as t = t̄/Rλ . Also, the crossflow components of velocity are expected to be
small compared to the streamwise component, and variations of pressure are expected to be very small. This
suggest the introduction of the scaling of dependent variables as:

u = ū; v = v̄/Rλ ; w = w̄/Rλ ; ρ = ρ̄; p = p̄/Rλ2 ; T = T̄ ; µ = µ̄; k = k̄; (13)

Plugging these into the full steady-state N-S equations, and performing on order of magnitude analysis, the
boundary region equations are obtained in the form

Continuity equation:

V · ∇ρ + ρ∇ · V = 0 (14)
x-momentum equation:

ρV · ∇u = ∇c · (µ∇c u) (15)
y-momentum equation:
       
∂p ∂ 2 ∂v ∂ ∂u ∂ ∂v ∂w
2
ρV · ∇v + Gλ u = − + µ 3 −∇·V + µ + µ + (16)
∂y ∂y 3 ∂y ∂x ∂y ∂z ∂z ∂y
z-momentum equation:
       
∂p ∂ 2 ∂w ∂ ∂u ∂ ∂v ∂w
ρV · ∇w = − + µ 3 −∇·V + µ + µ + (17)
∂z ∂z 3 ∂z ∂x ∂z ∂y ∂z ∂y
Energy equation:
" 2  2 #
1 ∂u ∂u
ρV · ∇T = 2
∇c · (k∇c T ) + (γ − 1)M∞ µ + (18)
Pr ∂y ∂z
where V is the velocity vector and ∇c is the crossflow nabla operator:
∂~ ∂~
V = u~i + v~j + w~k; ∇c = j+ k (19)
∂y ∂z
The effect of the wall curvature is contained in the term involving the global Görtler number Gλ .

Equations of state in nondimensional form:

ρT
p= 2
Re2λ ; µ = T b; k=µ (20)
γM∞
where b = 0.76.
This set of equations is parabolic in the streamwise direction and elliptic in the spanwise direction.
Appropriate initial/upstream and boundary conditions are necessary to close the problem; these conditions
will be the same as those used by Ricco & Wu.41
The nonlinear compressible boundary equations will be numerically solved using an algorithm similar to
the one used in Sescu & Thompson,45 adapted to the compressible regime.

III. Optimal control problem in the nonlinear regime

While it is common for an optimal flow control problem to be formulated in the framework of a dynamical
system (usually, described by a set of equations that are parabolic in time), here we replace the time
direction by the x-direction owing to the parabolic character of the BRE in the stream-wise direction, which
necessitates stream-wise marching to obtain a solution.

6 of 15

American Institute of Aeronautics and Astronautics

A. Generic optimal control formalism
To fix ideas, we first write equations (14)-(18) in the generic and more compact form

G(q) = 0, (21)
for brevity, with initial and boundary conditions

q(0, y, z) = q0 (y, z) (22)

q(x, 0, z) = φ, lim q(x, y, z) = qB (x, z), (23)
y→∞

along the wall-normal direction, and periodic or symmetry boundary conditions in the span-wise direction,
z. In equation (21), G() is the non-linear BRE differential operator in abstract notation, q = (ρ, u, v, w, T )
is the vector of state variables, φ is the control variable associated with the boundary conditions (e.g., the
transpiration velocity at the wall, vw ), q0 (y, z) represents the upstream or initial condition in x = 0, and
qB is a given function that specifies the boundary condition at infinity. We define an objective (or cost)
functional as

J (q, φ) = E(q) + σ kφx kβ2 + kφkβ2 , (24)

where E(q) is a specified target function to be minimized (e.g., the energy of the disturbance, or the wall
shear stress; the latter is considered in this study), the second term on the right hand side of (24) is a
penalization term depending on the norm of the control variable (usually, this quantity place a constraint
on the magnitude of the admissible control variable, since it cannot increase or decrease indefinitely), σ and
β are given constants, and subscript x denotes derivative with respect to x. The norm kk in equation (24)
is associated with an appropriate inner product of two complex functions, f and g, defined as
ˆ Xt
hf, gi = f ∗ gdX (25)
0
in the space [0, Xt ], with Xt being the terminal stream-wise location (the star in (25) denotes complex
conjugate).
A common approach to transform a (nonlinear) constrained optimization problem into an unconstrained
problem is by using the method of Lagrange multipliers (see, for example,,21 ,1560 ). To this end, we consider
the Lagrangian

L(q, φ, qa ) = J (q, φ) − hG(q), qa i, (26)

where qa is the vector of Lagrange multipliers (ρa , ua , v a , wa , T a ), also known as the adjoint vector. In other
words, the Lagrange multipliers are introduced in order to transform the minimization of J (q, φ) under the
constraint G(q) = 0 into the unconstrained minimization of L(q, φ, qa ). The unconstrained optimization
problem can be formulated as:
Find the control variable φ, the state variables q, and the adjoint variables qa such that the Lagrangian
L(q, φ, qa ) is a stationary function, that is
∂L ∂L ∂L a
δL = δq + δφ + δq = 0 (27)
∂q ∂φ ∂qa
where

∂L L(a + δa) − L(a)

δa = (28)
∂a 
represents directional differentiation in the generic direction δa. All directional derivatives in (27) must
vanish, providing different sets of equations:

• adjoint BRE equations are obtained by taking the derivative with respect to q,

∂L
=0 ⇒ G a (qa ) = 0 (29)
∂q

7 of 15

American Institute of Aeronautics and Astronautics

 • optimality conditions are obtained by taking the derivatives with respect to φ,

∂L
=0 ⇒ O(qa , q, φ) = 0 (30)
∂φ

• the original BRE equations are obtained by taking the derivative with respect to qa ,

∂L
=0 ⇒ G(q, ψ) = 0. (31)
∂qa

Equations (29)-(31) form the optimal control system that can be utilized to determine the optimal states
and the control variables. One can note that the stationarity of the Lagrangian with respect to the adjoint
variables qa = (ρa , ua , v a , wa , T a ) essentially yields the original state equations, while the stationarity with
respect to the state variables q = (ρ, u, v, w, T ) yields the adjoint equations that depend on the state
variables. The relationship between the state variables and the adjoint variables can be expressed by the
adjoint identity,

hG(q), qa i = hq, G a (qa )i + B(φ) (32)

where the last term, B, represents a residual from the boundary conditions.

B. Adjoint compressible BRE

In our particular case of a boundary layer flow over a flat or concave surface with wall transpiration, the
adjoint compressible BRE are derived starting with the integral

L(ρ, u, v, w, T, p, µ, k, vw , ρa , ua , v a , wa , T a , pa , µa , k a , s)
ˆ ˆ 2 ˆ X0 ˆ ˆ X1 ˆ ˆ Xt ˆ
σ X1 ∂vw 2
= + kvw k dΓdX − svdΓdX − s (v − vw ) dΓdX − svdΓdX
2 X0 Γ ∂X 0 Γ X0 Γ X1 Γ
ˆ ˆ
α Xs1
+ kτw − τ0 k2 dΓdX
2 Xs0 Γ
ˆ Xt ˆ
− ρa (V · ∇ρ + ρ∇V) dΩdX
0 Ω
ˆ Xt ˆ
− ua [ρV.∇u − ∇c · (µ∇c u)] dΩdX
0 Ω
ˆ Xt ˆ         
∂p ∂ 2 ∂v ∂ ∂u ∂ ∂v ∂w
− v a ρV · ∇v + GΛ u2 + − µ 3 −∇·V − µ − µ + dΩdX
0 Ω ∂y ∂y 3 ∂y ∂x ∂y ∂z ∂z ∂y
ˆ Xt ˆ         
a ∂p ∂ 2 ∂w ∂ ∂u ∂ ∂v ∂w
− w ρV · ∇w + − µ 3 −∇·V − µ − µ + dΩdX
0 Ω ∂z ∂z 3 ∂z ∂x ∂z ∂y ∂z ∂y
ˆ Xt ˆ ( "  2  2 #)
1 ∂u ∂u
− T a ρV · ∇T − ∇c · (k∇c T ) − (γ − 1)M∞2
µ + dΩdX
0 Ω P r ∂y ∂z
ˆ Xt ˆ  
ρT
− pa p − 2
R 2
eλ dΩdX
0 Ω γM∞
ˆ Xt ˆ
− µa (µ − T b )dΩdX
0 Ω
ˆ Xt ˆ
− k a (k − µ)dΩdX (33)
0 Ω

8 of 15

American Institute of Aeronautics and Astronautics

 In the above equations, the control is applied in specified intervals [X0 , X1 ] only, Ω is the cross-section
domain [0, ∞] × [z1 , z2 ] ranging from the wall (y = 0) to infinity and from z1 to z2 in the span-wise direction,
Γ is the wall boundary line for a given X, τw is the wall shear stress, τ0 is a target shear stress (equal to
the value corresponding to the Blasius solution), and [Xs1 , Xs2 ] is the interval where the cost function is
defined. The last three integrals in (33) are used to enforce the boundary condition on v, which includes the
wall transpiration. If we take the directional derivative of the Lagrangian with respect to ρ, the result is
ˆ Xt ˆ
δρT 2
ρa (V · ∇δρ + δρ∇ · V) + δρV · (ua ∇u + v a ∇v + wa ∇w + T a ∇T ) − pa 2
Reλ dΩdX = 0 (34)
0 Ω γM ∞

ˆ Xt ˆ  2

a a a a a T Reλ a
δρ V · (u ∇u + v ∇v + w ∇w + T ∇T ) − p 2
T − V · ∇ρ dΩdX
γM∞
0
ˆ Ω

+ ρa uδρ|X
0 dΩ
t

Ω
ˆXt ˆ
+ ρa (v + w)δρ|Γ dΓdX = 0 (35)
0 Γ

The second and third integrals are obtained from integration by parts, respectively, in [0, Xt ] and in Ω;
assuming arbitrary variations of δρ, the first adjoint equation is obtained in the form

2
T Reλ
V · (ua ∇u + v a ∇v + wa ∇w + T a ∇T ) − pa 2
− V · ∇ρa = 0 on [0, Xt ] × Ω (36)
γM∞
a~
where δρ|Γ = 0, δρ|X0 = 0, and V = u i + v j + w k.
t a a~ a~

Similarly, we take the directional derivative of the Lagrangian with respect to rest of the state variables
u, v, w, T and obtain

∂ρa
 
∂V ∂T 2 ∂µ
ρ Va · + Ta + 2GΛ uv a − ρ − ρV · ∇ua − (µ∇c · (∇c ua ) + ∇c µ · ∇c ua ) + (∇c · Va )
∂x ∂x ∂x 3 ∂x
∂Va ∂Va
 
1
− µ∇c · + ∇c µ · = 0 on [0, Xt ] × Ω (37)
3 ∂x ∂x

" 
∂ρa 4 ∂µ ∂v a ∂ 2 va ∂µ ∂v a ∂ 2 va 1 ∂ 2 wa
  
∂V a a ∂T a
ρ V · +T −ρ − ρV · ∇v − +µ 2 + +µ 2 + µ
∂y ∂y ∂y 3 ∂y ∂y ∂y ∂z ∂z ∂z 3 ∂y∂z
#
∂µ ∂wa 2 ∂µ ∂wa
+ + = 0 on [0, Xt ] × Ω (38)
∂z ∂y 3 ∂y ∂z

" 
∂ρa 4 ∂µ ∂wa ∂ 2 wa ∂µ ∂wa ∂ 2 wa 1 ∂ 2 va
  
∂V
a a ∂T a
ρ V · +T −ρ − ρV · ∇w − +µ 2
+ +µ 2
+ µ
∂z ∂z ∂z 3 ∂z ∂z ∂z ∂y ∂y ∂y 3 ∂y∂z
#
∂µ ∂v a 2 ∂µ ∂v a
+ + = 0 on [0, Xt ] × Ω (39)
∂y ∂z 3 ∂z ∂y

2
1 ρReλ
ρV · ∇T a + ∇c k · ∇c T a + k∇2c T a − pa + µa bT b−1 = 0 on [0, Xt ] × Ω (40)


Pr γM∞2

The initial and boundary conditions associated with the adjoint equations are

(ρa , ua , v a , wa , T a )|X=Xt = (0, 0, 0, 0, 0) in Ω, (41)

9 of 15

American Institute of Aeronautics and Astronautics

 
(α(τ − τ ), 0, 0) f or X ∈ [X , X ]
w 0 s0 s1
(ua , v a , wa )|Γ = (42)
(0, 0, 0) otherwise

and

(ρa , ua , v a , wa , T a )|Y →∞ = (0, 0, 0, 0) (43)

where α is a constant pre-factor that controls the penalization of the wall shear stress. With the state
variables (ρ, u, v, w, T ) determined from equations (14)-(17), the adjoint equations (36)-(46) are linear and
parabolic, and can be solved via a marching procedure in the backward direction, starting from the terminal
stream-wise location, Xt , towards the initial stream-wise location X0 . In the present study, the values of the
constant factors α and σ are set to 1 and 0.1, respectively.
The non time-dependent adjoint variables pa , µa , k a are determined in a similar fashion

∂v a ∂wa
pa = + (44)
∂y ∂z

∂T a ∂T ∂T a ∂T
 
1
ka = − + (45)
Pr ∂y ∂y ∂z ∂z

"
2 ∂v a 2 ∂wa
   
a a a ∂v ∂u ∂w ∂w ∂u ∂v
µ = k − ∇c u · ∇ c u + 2 + + + 2 + +
3 ∂y ∂y ∂x ∂z 3 ∂z ∂z ∂x ∂y
#
∂V a
 a
∂wa

∂v ∂v ∂w
∇c u · + + + (46)
∂x ∂z ∂y ∂z ∂y

The control variables can be updated using optimality conditions or by a steepest descent method ac-
cording to

dJ (n)
(n+1)
vw (n)
= vw −α (n)
(47)
dvw
where n represent the iteration index, and α in this case is the descent parameter.
The control algorithm starts with the solution to the compressible BRE for the uncontrolled boundary
layer, followed by the solution to the adjoint compressible BRE (note that the adjoint BRE depend on the
solution to the BRE). The difference between the wall shear stress and the original laminar wall shear stress
is then compared to a desired value; if the difference is larger than a given threshold then the steepest descent
method is utilized to determine the new wall transpiration velocity vw . Once these are determined, the loop
is repeated, and the calculation finishes when the difference between the wall shear stress and the desired
value is less than the threshold.
Adjoint equations (36)-(46) and the associated initial and boundary conditions (41)-(43) are solved nu-
merically on the same grid as the original BRE state equations (14)-(17), and utilizing the same numerical
algorithm (as for the BRE equations), except that the marching is preformed backwards, starting from the
terminal stream-wise location.

IV. Preliminary results

Numerical simulations of the nonlinear compressible boundary region equations (14)-(18), for a range of
Mach numbers covering subsonic (M = 0.8), supersonic (M = 2, 4), and hypersonic (M = 6) regimes were
carried out. Four spanwise separations have been considered: 0.3, 0.5, 0.7 and 0.9 cm. Since there are no
obstructions in the flow no shock waves are generated (this has been previously considered in several studies;
see, for example, Li et al.30 or Ren & Fu40 ), while the mean inflow condition is generated from the similarity
solution, which is obtained by means of the Dorodnitsyn-Howarth coordinate transformation
ˆ y
Ȳ (x, y) = ρ(x, ỹ)dỹ (48)
0

10 of 15

American Institute of Aeronautics and Astronautics

 A similarity variable is defined as
 1/2
Rλ
η = Ȳ , (49)
2x
and the base velocity and temperature can be expressed as

U = F 0 (η), V = (2xRλ )−1/2 (ηc T F 0 − T F ),

T = T (η) (50)
´η
where prime represents differentiation with respect to η, and ηc = 1/T 0 T (η̃)dη̃. F and T satisfy the
following coupled equations

µ 0
F 00 + F F 00 = 0,
T
1  µ 0 0 µ
T + F T 0 + (γ − 1)M 2 F 002 = 0, (51)
Pr T T
satisfying the boundary conditions

F (0) = F 0 (0) = 0, T 0 (0) = 0, F 0 → 1, T → 1 → as → η → ∞ (52)

The dependence of the viscosity on the temperature is assumed to be described by the power law,

µ = T b, b = 0.76 (53)

Equations (51) were solved by a shooting method for both adiabatic and isothermal conditions. Figure 3
shows this solution for Mach number equal to 4, and for both isothermal and adiabatic wall conditions.
20 20
isothermal
adiabatic

15 15

10 10

5 5

0 0
0 0.2 0.4 0.6 0.8 1 1 1.5 2 2.5
F' T a)
b)

Figure 2: Base flow solution.

Gortler vortices are excited by a small disturbance imposed on the base flow at the inflow boundary. The
global Gortler number is 3.34 × 106 corresponding to a radius of curvature of 2.4 m, the Reynolds number
based on the spanwise separation and the freestream velocity is 16, 755. In figure 3, we plot contours of
velocity magnitude in several cross-stream planes, for both Mach numbers, and with wall temperature set
at 300 K, where the same disturbances amplitude (relative to the free stream velocity) has been imposed at
the inflow boundary (there are multiple rows of mushroom shapes in figure 3 for presentation purposes, but
we only computed one row, and multiply it taking advantage of the periodicity in the spanwise direction).

11 of 15

American Institute of Aeronautics and Astronautics

 We quantify the vortex energy as

ˆz2 ˆ∞ h i
2 2 2
E(x) = |u(x, y, z) − um (x, y)| + |v(x, y, z) − vm (x, y)| + |w(x, y, z) − wm (x, y)| dzdy, (54)
z1 0

where um (x, y), vm (x, y), and wm (x, y) are the spanwise mean components of velocity, and z1 and z2 are the
coordinates of the boundaries in the spanwise direction.

Figure 3: Contours of velocity magnitude in several cross-stream planes for the disturbance amplitude of
0.005 and different spanwise separations.

Figure 4a shows the kinetic energy distribution along the streamwise direction for the two wall temper-
atures. A reduction of the energy is noticed as the wall is heated at Tw = 450 K. In figure 4b, the spanwise
averaged wall shear stress is plotted as a function of the streamwise coordinate; here, an increase in the wall
shear stress is observed as the wall is heated.
Streak amplitude is defined here as the difference between the maximum and minimum streamwise
velocity disturbance over 0 < y < ∞ and 0 < z < πl, where λ/2 is half from the spanwise separation,

S(x) = max [u(x, y, z) − um (x, y)] − min [u(x, y, z) − um (x, y)]

0<y<∞ 0<y<∞
0<z<λ/2 0<z<λ/2

In figure 5, the streak amplitude is plotted as a function of the streamwise coordinate for different Mach
numbers and spanwise separations. As the spanwise separation increases, the maximum of streak amplitude
moves in the downstream.

12 of 15

American Institute of Aeronautics and Astronautics

 Figure 4: Flow vortex energy for consecutive iterations

Figure 5: Flow shear stress for consecutive iterations

The final paper will include results from the adjoint equations with wall transpiration as a control
function. A parametric study based on the variation of the Mach number, temperature of the wall, and
spanwise separation will be conducted.

References
1 Bewley, T., and Moin, P. (1994) Optimal control of turbulent channel flows, ASME DE, Vol. 75, pp. 221-227.
2 Boiko, A.V., Ivanov, A.V., Kachanov, Y.S. and Mischenko, D.A. (2010) Steady and unsteady Görtler boundary-layer
instability on concave wall, Eur. J. Mech. B - Fluids, Vol. 29, pp. 61-83.
3 Cathalifaud, P., and Luchini, P. (2000) Algebraic growth in boundary layers: optimal control by blowing and suction at

the wall, Eur. J. Mech. B - Fluids. Vol. 19, pp. 469-490.

4 Cherubini, S., Robinet, J.-C., and De Palma, P. (2013) Nonlinear control of unsteady finite-amplitude perturbations in

the Blasius boundary-layer flow, J. Fluid Mech.. Vol. 737, pp. 440-465.
5 Choi, H., Moin, P., and Kim, J. (1994) Active turbulence control for reduction in wall-bounded flows, J. Fluid Mech.. Vol.

262, pp. 75-110.

6 Choudhari, M. and Fischer, P. (2005) Roughness induced Transient Growth, AIAA Paper 2005-4765.

13 of 15

American Institute of Aeronautics and Astronautics

 7 Corbett, P. and Bottaro, A. (2001) Optimal Control of Nonmodal Disturbances in Boundary Layers, Theoret. Comput.

Fluid Dynamics. Vol. 15, pp. 65-81.

8 Depmsey, L.D., Hall, P. and Deguchiu, K. (2017) The excitation of Gortler vortices by free stream coherent structures, J.

Fluid Mech., Vol. 826, pp. 60-96.

9 Gad-el-Hak, M. (1989) Flow Control, Applied Mechanics Review. Vol. 42, pp. 261-293.
10 Goldstein, M., Sescu, A., Duck, P. and Choudhari, M. (2010) The Long Range Persistence of Wakes behind a Row of

Roughness Elements, J. Fluid Mech., Vol. 644, pp. 123-163.

11 Goldstein, M., Sescu, A., Duck, P. and Choudhari, M. (2011) Algebraic/transcendental Disturbance Growth behind a

Row of Roughness Elements, J. Fluid Mech., Vol. 668, pp. 236-266.

12 Goldstein, M., Sescu, A., Duck, P. and Choudhari, M. (2016) Nonlinear wakes behind a row of elongated roughness

elements, J. Fluid Mech., Vol. 796, pp. 516-557.

13 Goldstein, M., and Sescu, A. (2008) Boundary-layer transition at high free-stream disturbance levels - beyond Klebanoff

modes. J. Fluid Mech., Vol. 613, pp. 95-124.

14 Görtler, H . (1941) Instabilita-umt laminarer Grenzchichten an Konkaven Wanden gegenber gewissen dreidimensionalen

Storungen, ZAMM, Vol. 21, pp. 250–52; english version: NACA Report 1375 (1954)
15 Gunzburger, M. (2000) Adjoint Equation-Based Methods for Control Problems in Incompressible, Viscous Flows, Flow,

Turbulence and Combustion. Vol. 65, pp. 249.

16 Hall, P. (1982) Taylor-Görtler vortices in fully developed or boundary-layer flows: linear theory, J. Fluid Mech., Vol. 124,

pp. 475-494.
17 Hall, P. (1983) The linear development of Görtler vortices in growing boundary layers, J. Fluid Mech., Vol. 130, pp.

41-58.
18 Hall, P. and Horseman. N. (1991) The linear inviscid secondary instability of longitudinal vortex structures in boundary

layers. J. Fluid Mech., Vol. 232, pp. 357-375.

19 Högberg, M., Bewley, T.R., Henningson, D.S. (2003) Relaminarization of Re = 100 turbulence using gain scheduling
τ
and linear state feedback control, Phys. Fluids, Vol. 15, pp. 3572.
20 Jacobson, S.A., and Reynolds, W.C. (1998) Active control of streamwise vortices and streaks in boundary layers J. Fluid

Mech., Vol. 360, pp. 179-211.

21 Joslin, R., Gunzburger, M., Nicolaides, R., Erlebacher, G., and Hussaini, M. (1197) Self-Contained Automated Method-

ology for Optimal Flow Control, AIAA J.. Vol. 35, pp. 816-824.
22 Kendall, J.M. (1998) Experiments on boundary-layer receptivity to freestream turbulence, AIAA Paper 2004-2335.
23 Kim, J. (2003) Control of turbulent boundary layers, Phys. Fluids. Vol. 15, pp. 1093.
24 Koumoutsakos, P. (1997) Active control of vortex-wall interactions, Phys. Fluids. Vol. 9, pp. 3808.
25 Koumoutsakos, P. (1999) Vorticity flux control for a turbulent channel flow, Phys. Fluids. Vol. 11, pp. 248.
26 Landahl, M.T. (1980) A note on an algebraic instability of inviscid parallel shear flows. J. Fluid Mech., Vol. 98, pp.

243-251.
27 Lee, C., Kim, J., and Choi, H. (1998) Suboptimal control of turbulent channel flow for drag reduction, J. Fluid Mech..

Vol. 358, pp. 245-258.

28 Leib, S.J., Wundrow, W., and Goldstein, M. (1999) Effect of free-stream turbulence and other vortical disturbances on a

laminar boundary layer. J. Fluid Mech., Vol. 380, pp. 169-203.

29 Li, F., and Malik, M. (1995) Fundamental and subharmonic secondary instabilities of Görtler vortices, J. Fluid Mech.,

Vol. 297, pp. 77-100.

30 Li, F., Choudhari, M., Chang, C.-L., Greene, P., and Wu, M. (2010) Development and Breakdown of Gortler Vortices

in High Speed Boundary Layers", 48th AIAA Aerospace Sciences Meeting Including the New Horizons Forum and Aerospace
Exposition, Aerospace Sciences Meetings.
31 Lu, L., Agostini, L., Ricco, P., & Papadakis, P., (2014) Optimal state feedback control of streaks and Gortler vortices

induced by free-stream vortical disturbances, UKACC International Conference on Control. Loughborough, U.K.
32 Luchini, P., and Bottaro, A. (2014) Adjoint Equations in Stability Analysis, Annu. Rev. Fluid Mech.. Vol. 46, pp. 493-517.
33 Lundell, F., and Alfredsson, P.H. (2003) Experiments on control of streamwise streaks, Eur. J. Mech. (B/Fluids). Vol.

22, pp. 1279-1290.

34 Malik, M.R., and Hussaini, M.Y. (1990) Numerical simulation of interactions between Gortler vortices and Tollmien-

Schlichting waves. J. Fluid Mech., Vol. 210, pp. 183-199.

35 Marensi, E., Ricco, P., and Wu, X. (2017) Nonlinear unsteady streaks engendered by the interaction of free-stream

vorticity with a compressible boundary layer. J. Fluid Mech.. Vol. 817, pp. 80-121.
36 Matsubara, M., and Alfredsson, P.H. (2001) Disturbance growth in boundary layers subjected to free stream turbulence,

J. Fluid Mech., Vol. 430, pp. 149.

37 Myose, R. Y., and Blackwelder, R. F. (1995) Control of streamwise vortices using selective suction, AIAA Journal, Vol.

33, pp 1076-1080.
38 Pamies, M., Garnier, E., Merlen, A., and Sagaut, P. (2007) Response of a spatially developing turbulent boundary layer

to active control strategies in the framework of opposition control, Phys. Fluids. Vol. 19, pp. 108102.
39 Papadakis, G., Lu,L., and Ricco, P. (2016) Closed-loop control of boundary layer streaks induced by free-stream turbu-

lence, Phys. Rev. Fluids. Vol. 1, pp. 043501.

40 Ren, J., and Fu, S. (2017) Secondary instabilities of Gortler vortices in high-speed boundary layer flows, J. Fluid Mech.,

Vol. 781, pp. 388-421.

41 Ricco, P. (2011) Laminar streaks with spanwise wall forcing, Phys. Fluids, Vol. 23, pp. 064103.
42 Saric, W.S. (1994) Görtler Vortices, Annu. Rev. Fluid Mech., Vol. 26, pp. 379-409.

14 of 15

American Institute of Aeronautics and Astronautics

 43 Schneider, P. S. (2001) Effects of high-speed tunnel noise on laminar-turbulent transition, J. Spacecr. Rockets, Vol 38, pp

323-333.
44 Sescu, A., Pendyala, R., and Thompson, D. (2014) On the Growth of Görtler Vortices Excited by Distributed Roughness

Elements, AIAA Paper 2014-2885.

45 Sescu, A., and Thompson, D. (2015) On the Excitation of Görtler Vortices by Distributed Roughness Elements, Theoret.

Comput. Fluid Dynamics, Vol. 29, pp. 67-92.

46 Sescu, A., Taoudi, L., Afsar, M., and Thompson D. (2016) Control of Gortler Vortices by Means of Surface Streaks, AIAA

Paper 2016-3950.
47 Sescu, A., Taoudi, L. & Afsar, M. (2018) Iterative control of Gortler vortices via local wall deformations, Theoret. Comput.

Fluid Dynamics, Vol. 32, pp. 63-72.

48 Sescu, A., and Afsar, M. (2018) Hampering Gortler vortices via optimal control in the framework of nonlinear boundary

region equations, J. Fluid Mech., Vol. 848, p. 5-41.

49 Sescu, A., Alaziz, R., and Afsar, M.Z. (2019) Effect of Wall Transpiration and Heat Transfer on Nonlinear Gortler Vortices

in High-speed Boundary Layers, AIAA Journal, doi.org/10.2514/1.J057330

50 Stroh, A., Frohnapfel, B., Schlatter, P., and Hasegawa, Y. (2015) A comparison of opposition control in turbulent

boundary layer and turbulent channel flow, Phys. Fluids. Vol. 27, pp. 075101.
51 Swearingen, J.D. and Blackwelder, R.F. (1987) The growth and breakdown of streamwise vortices in the presence of a

wall. J. Fluid Mech., Vol. 182, pp. 255-290.

52 Westin, K.J.A., Boiko, A.V., Klingmann, B.J.B., Kozlov, V.V., and Alfredsson, P.H. (1994) Experiments in a boundary

layer subjected to free stream turbulence. part 1. Boundary layer structure and receptivity. J. Fluid Mech., Vol. 281, pp. 193-218.
53 White, E.B. (2002) Transient growth of stationary disturbances in a flat plate boundary layer, Phys. Fluids, Vol. 14, pp.

4429-4439.
54 White, E.B., Rice, J.M. and Ergin, F.G. (2005) Receptivity of stationary transient disturbances to surface roughness,

Phys. Fluids, Vol. 17, pp. 064109.

55 Wu, X, Zhao, D. and Luo, J (2011) Excitation of steady and unsteady Görtler vortices by free-stream vortical disturbances,

J. Fluid Mech., Vol. 682, pp. 66-100.

56 Wu, X. and Choudhari, M. (2011) Linear and nonlinear instabilities of a blasius boundary layer perturbed by streamwise

vortices. Part 2. Intermittent instability induced by long wavelength Klebanoff modes. J. Fluid Mech.. Vol. 483, pp. 249-286.
57 Xiao, D., and Papadakis, G. (2017) Nonlinear optimal control of bypass transition in a boundary layer flow, Phys. Fluids.

Vol. 29, pp. 054103.

58 Xu, D., Zhang, Y., and Wu, X. (2017) Nonlinear evolution and secondary instability of steady and unsteady Gortler

vortices induced by free-stream vortical disturbances, J. Fluid Mech., Vol. 829, pp. 681-730.
59 Zaki, T.A., and Durbin, P. (2005) Mode interaction and the bypass route to transition, J. Fluid Mech., Vol. 531, pp.

85-111.
60 Zuccher, S., Luchini, P., and Bottaro, A. (2004) Algebraic growth in a blasius boundary layer: optimal and robust control

by mean suction in the nonlinear regime, Eur. J. Mech. B - Fluids. Vol. 513, pp. 135-160.

15 of 15

American Institute of Aeronautics and Astronautics