QMC: Operator Splitting Workshop, Optimization and Control in Free ans Moving Boundary Fluid-Structure Interactions - Lorena Bociu, Mar 22, 2018

Optimization and Control in Free and
Moving Boundary Fluid-Structure
Interactions
Lorena Bociu 1
SAMSI - Operator Splitting Methods in Data Analysis
March 21, 2018
1
The research was supported by NSF-DMS 1312801 and NSF CAREER 1555062

Fluid-Structure Interactions (FSI)
Interaction of some movable and/or deformable structure with
an internal or surrounding ﬂuid ﬂow
industrial processes, aero-elasticity, and biomechanics
The boundary of the domain is not known in advance,
but has to be determined as part of the solution.
Free boundary: steady-state problem.
Moving boundary: time dependent problems and the position of the
boundary is a function of time and space.

Lagrangian vs Eulerian
Coupling of incompressible Navier-Stokes equations with
an elastic solid.
Solid: displacements are often relatively small
computational domain: Ω
Lagrangian formulation: focus on the material particle x and its
evolution
Fluid: displacements are large and usually irrelevant
we are mostly interested in the velocity ﬁeld
Eulerian framework: observe what happens at a given point x in the
physical space.

an elastic solid.
evolution
physical space.
FSI: Fluid + elasticity + interface conditions between solid and ﬂuid

an elastic solid.
evolution
physical space.
FSI: Fluid + elasticity + interface conditions between solid and ﬂuid
match these two diﬀerent frameworks

Fluid - Elasticity Interaction: PDE model
Configuration: the elastic body moves and deforms inside the fluid.
Elastic body located at time t ≥ 0 in a domain Ω(t) ⊂ R3
with
boundary Γ(t).
The fluid occupies domain Ωf
(t) = D ¯Ω(t), with smooth boundary
Γ(t) ∪ Γf .
Let D ⊂ R3
be the control volume. D contains the solid and the
fluid at each time t ≥ 0, i.e. D = Ω(t) ∪ Ωf
(t), with smooth
boundary ∂D = Γf .

Navier-Stokes - Eulerian Framework
Fluid: Newtonian viscous, homogeneous, and incompressible.
Its behavior is described by its velocity w and pressure p.
The viscosity of the fluid is ν > 0, and the fluid strain and
stress tensors are given by
ε(w) =
1
2
[Dw + (Dw)∗
], σ(p, w) = −pI + 2νε(w),
where Dw is the gradient matrix of w, and (Dw)∗ represents
the transpose of Dw.
The fluid state satisfies the following Navier-Stokes equations:



wt − ν ∆w + Dw · w + p = v1 on Ωf (t)
div w = 0 on Ωf (t)
w = 0 on Γf

Structural Deformation: Lagrangian formulation
The evolution of the fluid domain Ωf
(t) is induced by the
structural deformation through the common interface Γ(t).
O ⊂ D: reference configuration for the solid; ∂O = S
Of
= D ¯O: reference fluid configuration. T
D is described by a smooth, injective map:
ϕ : ¯D × R+
−→ ¯D, (x, t) → ϕ = ϕ(x, t).
For x ∈ O, ϕ(x, t): the position at time t of the material point x.
On Of
, ϕ(x, t) is defined as an arbitrary extension of the restriction
of ϕ to S, which preserves the boundary Γf , i.e. ϕ = IΓf
on Γf .
J(ϕ) > 0: Jacobian of the deformation ϕ(t)

Nonlinear elasticity
St. Venant - Kirchhoﬀ equations: large displacement, small
deformation elasticity. Green-St. Venant nonlinear strain tensor:
σ(ϕ) =
1
2
[(Dϕ)∗
Dϕ − I].
Piola transform of the Cauchy stress tensor:
P(x) = Dϕ(x)[λTr[σ(ϕ)]I + 2µσ(ϕ)])
Equilibrium equations for elasticity :
Jρ∂ttϕ − DivP = Jρv2 on O

FSI - PDE model



wt − ν ∆w + Dw · w + p = v1 on Ωf
(t)
div w = 0 on Ωf
(t)
w = 0 on Γf
Jρ∂ttϕ − DivP = Jρv2 on O
w ◦ ϕ = ϕt on S
Pn = J(ϕ)(σ(p, w) ◦ ϕ)(Dϕ)−∗
n on S
ϕ = IΓf
on Γf ,
IC:ϕ(·, 0) = ϕ0
, ϕt(·, 0) = ϕ1
, w(·, 0) = w0
, p(·, 0) = p0
on (O)2
× (Oc
)2
.

PDE-constrained Optimization Problems governed by FSI
In most of the applications, the goal is the
optimization or optimal control of the considered process,
related sensitivity analysis (with respect to relevant physical parameters).
minimize turbulence in the ﬂuid
optimize ﬂuid velocity or pressure
optimize the deformation of the structure
minimize wall shear stresses

Control problems in FSI: most of the literature is focused on the
assumption of small but rapid oscillations of the solid body, so that
the common interface may be assumed ﬁxed:
Lasiecka and Bucci ’05, ’10, Lasiecka, Triggiani, and Zhang ’11,
Lasiecka and Tuﬀaha, ’08-’09, Avalos-Triggiani ’08-’12.

Recently, PDE constrained optimization problems governed by free
boundary interactions have been considered, with most research
studies mainly addressed in the context of the numerical analysis
of the ﬁnite element methods [Antil-Nochetto-Sodre ’14,
Richter-Wick ’13, Van Der Zee et al ’10]

Steady State Navier-Stokes and Elasticity



−ν ∆w + Dw · w + p = v|Ωf
on Ωf
divw = 0 on Ωf
w = 0 on Γ := ϕ(S)
−DivP = v|Ωe
on O
Pn = J(ϕ)(σ(p, w) ◦ ϕ)(Dϕ)−∗
n on S
w = 0, ϕ = IΓf
on Γf
2
P.G. Ciarlet, Mathematical Elasticity Vol. I: Three-dimensional Elasticity, North-Holland Publishing Co.,

Steady State Navier-Stokes and Elasticity



−ν ∆w + Dw · w + p = v|Ωf
on Ωf
divw = 0 on Ωf
w = 0 on Γ := ϕ(S)
−DivP = v|Ωe
on O
Pn = J(ϕ)(σ(p, w) ◦ ϕ)(Dϕ)−∗
n on S
w = 0, ϕ = IΓf
on Γf
Cauchy Stress Tensor T : Ωe → S3
, T = [J−1
P · (Dϕ)∗
] ◦ ϕ−1
[2
]



−ν ∆w + Dw · w + p = v|Ωf
on Ωf
divw = 0 on Ωf
w = 0 on Γ := ϕ(S)
−DivT = v|Ωe
on Ωe = ϕ(O)
T n = σ(p, w)n on Γ
w = 0, ϕ = IΓf
on Γf .
2
P.G. Ciarlet, Mathematical Elasticity Vol. I: Three-dimensional Elasticity, North-Holland Publishing Co.,

OCP
We consider the optimal control problem:
min J(w, v) = 1/2 w − wd
2
L2(Ωf ) + 1/2 v 2
H3(D) (1)
subject to



−ν ∆w + Dw · w + p = v|Ωf
on Ωf
divw = 0 on Ωf
w = 0 on Γ := ϕ(S)
−DivT = v|Ωe
on Ωe = ϕ(O)
w = 0, ϕ = IΓf
on Γf .
distributed control v ∈ H3
(D)
wd ∈ L2
(Ωf ) is a desired ﬂuid velocity.

OCP
min J(w, v) = 1/2 w − wd
2
L2(Ωf ) + 1/2 v 2
H3(D) (2)
subject to
(E)



−ν ∆w + Dw · w + p = v|Ωf
on Ωf
divw = 0 on Ωf
w = 0 on Γ := ϕ(S)
−DivT = v|Ωe on Ωe = ϕ(O)
w = 0, ϕ = IΓf
on Γf .
Goals:
1. Existence of an optimal control
2. First-order necessary conditions of optimality (NOC)

Well-posedness Analysis
FSI: parabolic-hyperbolic coupled system
regularity gap of the fluid and structure velocities on the common
interface: the traces of the elastic component at the energy level are
not defined via the standard trace theory, and this induces a loss of
regularity at the boundary of the coupled system.
Coutand-Shkoller ’05-’06: Existence of strong solutions for the case of a linear and then quasi-linear elastic
body flowing within a viscous, incompressible fluid, under the assumptions of smooth initial data (i.e., the
initial fluid velocity w0
belongs to H5
, and the initial data for elasticity (ϕ0
, ϕ1
) belong to H3
× H2
).
Due to the incompressibility condition of the fluid, uniqueness of solution for the model required higher
regularity for the initial data (i.e., (w0
, ϕ0
, ϕ1
) ∈ H7
× H5
× H4
).
Kukavica-Tuffaha-Ziane ’09-’11, Ignatova-Kukavica-Lasiecka-Tuffaha ’12-’14, Raymond-Vanninathan ’15
for N-S coupled with linear elasticity/wave equation.
The authors of [Ignatova-Kukavica-Lasiecka-Tuffaha] also prove global in time well-posedness for small
initial data of the Navier–Stokes-elasticity model involving a wave equation with frictional damping, and
they show that the energy associated with smooth and sufficiently small solutions of the damped model
decay exponentially to zero.
Canic-Muha ’13-’14: dynamical coupling (which is of great interest in the modeling and analysis of the
cardiovascular system) - 2D
Grandmont’02, Wick-Wollner’14: steady state NS-St. Venant elasticity equations.

OCP
min J(w, v) = 1/2 w − wd
2
L2(Ωf ) + 1/2 v 2
H3(D) (3)
subject to
(E)



−ν ∆w + Dw · w + p = v|Ωf
on Ωf
divw = 0 on Ωf
w = 0 on Γ := ϕ(S)
−DivT = v|Ωe on Ωe = ϕ(O)
w = 0, ϕ = IΓf
on Γf .
Goals:
1. Existence of an optimal control (EOC)
2. First-order necessary conditions of optimality (NOC)
Compute the gradient of the functional J.
Characterization of the optimal control will pave the way for a
numerical study of the problem.

NOC: Main Challenge
Lagrangian: L = J − (weak form of the system)
Not convex-concave, due to the nonlinearity of the
control-to-state map.
Min-Max theory does not apply, i.e., one can not reduce the
cost function gradient to the derivative of the Lagrangian with
respect to the control, at its saddle point [ Delfour-Zolesio ’86]
Optimality conditions must be derived from diﬀerentiability
arguments on the cost functional J with respect to the control v.
Main challenge: dependence of the cost integrals in J on the
unknown domain Ωf , which also depends on the control v.
min J(w, v) = 1/2 w − wd
2
L2(Ωf ) + 1/2 v 2
H3(D)
Directional derivative of J with respect to v in the direction of v :
for small parameter s ≥ 0, consider the perturbed functional
J(v + sv ) and then calculate the derivative at s = 0 of the function
s → J(v + sv ).

With the following notation for the s-derivatives at s = 0,
ϕ =
∂
∂s
ϕs
s=0
, U = ϕ ◦ϕ−1
, w =
∂
∂s
ws
s=0
, and p =
∂
∂s
ps
s=0
,
we can compute the directional derivative of J as
∂J(v; v ) = lim
s→0
J(v + sv ) − J(v)
s
=
∂
∂s
J(v + sv )
s=0
=
∂
∂s
1
2 (Ωf )s
|ws − wd |2
+
1
2
v + sv 2
H3(D)
s=0
=
Ωf
(w − wd ) · w +
1
2 Γ
|w − wd |2
U · nf + (v, v )H3(D)

With the following notation for the s-derivatives at s = 0,
ϕ =
∂
∂s
ϕs
s=0
, U = ϕ ◦ϕ−1
, w =
∂
∂s
ws
s=0
, and p =
∂
∂s
ps
s=0
,
we can compute the directional derivative of J as
∂J(v; v ) = lim
s→0
J(v + sv ) − J(v)
s
=
∂
∂s
J(v + sv )
s=0
=
∂
∂s
1
2 (Ωf )s
|ws − wd |2
+
1
2
v + sv 2
H3(D)
s=0
=
Ωf
(w − wd ) · w +
1
2 Γ
|w − wd |2
U · nf + (v, v )H3(D)
The challenge of applying optimization tools to free boundary FSI is
the proper derivation of the sensitivity and adjoint sensitivity
information with correct balancing conditions on the common
interface.
As the interaction is a coupling of Eulerian and Lagrangian
quantities, sensitivity analysis on the system falls into the framework
of shape analysis.

Sensitivity System [LB - J.-P. Zolesio]



−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
−Div T(U’) = v Ωe
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf




−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Θ = Dϕ ◦ ϕ−1
DU := Θ∗
(DU )Θ,
T(U ) := (DU )T +
1
det Θ
Θ · {λTr(DU )I + µ[DU + (DU )∗
]}Θ∗
,




−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Θ = Dϕ ◦ ϕ−1
DU := Θ∗
(DU )Θ,
T(U ) := (DU )T +
1
det Θ
]} Θ∗
,
Σ(U )




−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Θ = Dϕ ◦ ϕ−1
DU := Θ∗
(DU )Θ,
T(U ) := (DU )T +
1
det Θ
]} Θ∗
,
Σ(U )
˜Σ(U )




−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Θ = Dϕ ◦ ϕ−1
DU := Θ∗
(DU )Θ,
T(U ) := (DU’)T +
1
det Θ
]} Θ∗
,
Σ(U )
˜Σ(U )

Sensitivity System [LB - J.-P. Zolesio ]



−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
−DivT(U ) = v Ωe
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf

Sensitivity System [LB - J.-P. Zolesio ]



−ν∆w + (Dw’) w + (Dw) w’ + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U’ = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Γ U , n
B(U ) =(T + pI − 2νε(w)) · [(DΓU )∗
n + (D2
bΩe )UΓ] +(DT )U · n
+ div(U )T · n − T · (DU )∗
· n−
− U , n (−DivΓ(T ) + [∂νpI − 2ν∂νε(w)] · n).

Notation
(Df)ij = ∂j fi ∈ M3
is the gradient matrix at a ∈ X of any vector
field f = (fi ) : X ⊂ R3
→ R3
.
div f = ∂i fi ∈ R is the divergence of f : X ⊂ R3
→ R3
.
Div T = ∂j Tij ∈ R3
is the divergence of any second-order tensor
field T = (Tij ) : X ⊂ R3
→ M3
.
A∗
= transpose of A, for any A ∈ M3
.
dΩ(x) =
infy∈Ω |y − x| Ω = ∅
∞ Ω = ∅
is the distance function
bΩ(x) = dΩ(x) − dΩc (x) , ∀x ∈ Rn
is the oriented distance fn.
from x to Ω, for any Ω ⊂ Rn
.
H = ∆bΩ = Tr(D2
bΩ) is the additive curvature of Γ = ∂Ω. [3
]
3
M.C. Delfour and J.P. Zolesio, Shapes and Geometries: Analysis, Differential Calculus and Optimization,
SIAM 2001.




−ν∆w + (Dw )w + (Dw)w + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Γ U , n
B(U ) =(T + pI − 2νε(w)) · [(DΓU )∗
n + (D2
bΩe
)UΓ] +(DT )U · n
+ div(U )T · n − T · (DU )∗
· n−




−ν∆w + (Dw )w + (Dw)w + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U’) on Γ
w = 0, U = 0 on Γf
Γ U , n
B(U ) =(T + pI − 2νε(w)) · [(DΓU )∗
n + (D2
bΩe
)UΓ] +(DT )U · n
+ div(U )T · n − T · (DU )∗
· n−
= B1 · Γ U , n − U , n B2 + (DT )U · n + div(U )T · n − T · (DU )∗
· n

Connection to Shape Analysis
As vs = v + sv , the geometry of the problem moves with the
flow of a vector field that depends on the deformation ϕs.
The perturbation Γs of the boundary is built by the flow of the
vector field V (s, x) = ∂
∂s ϕs ◦ ϕ−1
s , i.e.,
Γs = Ts(V )(S), where Ts(V ) : Ωe → (Ωe)s, Ts(V ) = ϕs◦ϕ−1
.

Connection to Shape Analysis
As vs = v + sv , the geometry of the problem moves with the
flow of a vector field that depends on the deformation ϕs.
The perturbation Γs of the boundary is built by the flow of the
vector field V (s, x) = ∂
∂s ϕs ◦ ϕ−1
s , i.e.,
Γs = Ts(V )(S), where Ts(V ) : Ωe → (Ωe)s, Ts(V ) = ϕs◦ϕ−1
.
(ϕ , w , p ): ‘shape’ derivatives with respect to the speed V , which
is a vector field that depends on ϕs and is not given a priori.
Standard theory on shape derivatives: the domain is perturbed
by an a priori given vector field and then the speed method is
applied.
s-derivatives: ‘pseudo-shape derivatives’, in the sense that
much of the theory of shape calculus remains applicable.

Goal: ﬁnd the gradient of J at v: J (v; v )
∂J(v; v ) =
Ωf
(w − wd ) · w +
1
2 Γ
|w − wd |2
U · nf + (v, v )H3(D)
Sensitivity system provides the characterization for (U , w , p ):



−ν∆w + (Dw )w + (Dw)w + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U ) on Γ
w = 0, U = 0 on Γf
v does not appear in the chain rule computation, since it is hidden
in the sensitivity equations for w , p , and U .

Goal: ﬁnd the gradient of J at v: J (v; v )
∂J(v; v ) =
Ωf
(w − wd ) · w +
1
2 Γ
|w − wd |2
U · nf + (v, v )H3(D)
Sensitivity system provides the characterization for (U , w , p ):



−ν∆w + (Dw )w + (Dw)w + p = v Ωf
in Ωf
divw = 0 in Ωf
w + (Dw)U = 0 on Γ
in Ωe
T(U ) · n = (−p I + 2νε(w )) · n + B(U ) on Γ
w = 0, U = 0 on Γf
v does not appear in the chain rule computation, since it is hidden
in the sensitivity equations for w , p , and U .
Idea: Introduce a suitable adjoint problem that eliminates the
s-derivatives and provides an explicit representation for J (v; v ).

Theorem (LB - K. Martin)
For the optimal control problem:
min J(w, v) = 1/2 w − wd
2
L2(Ωf ) + 1/2 v 2
H3(D),
subject to FSI, the gradient of the cost functional is given by
J (v; v ) = (v , v)D + (v |Ωf
, Q) + (v |Ωe
, R),
where Q, P, and R solve the following adjoint sensitivity problem:



−ν∆Q + (Dw)∗
Q − (DQ)w + P = w − wd Ωf
div(Q) = 0 Ωf
−Div ¯T (R) = 0 Ωe
Q = R Γ
¯T (R)n + (Dw)∗
σ(P, Q)n + divΓ[B1R]n − (DT ∆
· n)∗
R
−H T n, R n + Γ T n, R
−DivΓ(n ⊗ T R) + B2, R n = 1
2 |w − wd |2
nf Γ
Q = 0 Γf
(4)

Matching of Normal Stress Tensors
¯T (R)n+(Dw)∗
σ(P, Q)n+divΓ[B1R]n−(DT ∆
·n)∗
R−H T n, R n+ Γ T n, R
−DivΓ(n ⊗ T R) + B2, R n =
1
2
|w − wd |2
nf
B1 = T + pI − 2νε(w) and B2 = −DivΓ(T ) + [∂νpI − 2ν∂νε(w)] · n
(DT ∆
· f )ik := ∂k Tij fj

¯T (R)n+(Dw)∗
·n)∗
−DivΓ(n ⊗ T R) + B2, R n =
1
2
|w − wd |2
nf
(DT ∆
DT is deﬁned as (DT .e)ij = (∂k Tij )ek . With the above notation,
we can IBP
˜Γc
{(DT )γ} · ne, R =
˜Γc
(∂k Tij γk )(ne)j Ri
=
˜Γc
γk (∂k Tij (ne)j Ri ) =
˜Γc
γ, (DT ∆
· ne)∗
R .

¯T (R)n+(Dw)∗
·n)∗
−DivΓ(n ⊗ T R) + B2, R n =
1
2
|w − wd |2
nf
(DT ∆
DT is deﬁned as (DT .e)ij = (∂k Tij )ek . With the above notation,
we can IBP
˜Γc
{(DT )γ} · ne, R =
˜Γc
(∂k Tij γk )(ne)j Ri
=
˜Γc
γk (∂k Tij (ne)j Ri ) =
˜Γc
γ, (DT ∆
· ne)∗
R .
˜B(R) = divΓ[B1R]n − (DT ∆
· n)∗
R − H T n, R n + Γ T n, R
−DivΓ(n ⊗ T R) + B2, R n

Theorem (LB - K. Martin)
For the optimal control problem:
min J(w, v) = 1/2 w − wd
2
L2(Ωf ) + 1/2 v 2
H3(D),
subject to FSI, the gradient of the cost functional is given by
J (v; v ) = (v , v)D + (v |Ωf
, Q) + (v |Ωe
, R),
where Q, P, and R solve the following adjoint sensitivity problem:



−ν∆Q + (Dw)∗
Q − (DQ)w + P = w − wd Ωf
div(Q) = 0 Ωf
−Div ¯T (R) = 0 Ωe
Q = R Γ
¯T (R)n + (Dw)∗
σ(P, Q)n + ˜B(R) = 1
2 |w − wd |2
nf Γ
Q = 0 Γf
(5)

References
1. F. Abergel and R. Temam, On Some Control Problems in Fluid Mechanics, Theoret. Comput. Fluid
Dynamics 1, (1990), 303-325.
2. H. Antil, R. H. Nochetto, and P. Sodré, Optimal Control of a Free Boundary Problem: Analysis with
Second-Order Sufficient Conditions, SIAM J. Control Optim. 52, 5, (2014), 2771-2799.
3. L. Bociu, L. Castle, K. Martin, and D. Toundykov, Optimal Control in a Free Boundary Fluid-Elasticity
Interaction, AIMS Proceedings, (2015), 122-131.
4. L. Bociu, D. Toundykov, and J.-P. Zolésio, Well-Posedness Analysis for a Linearization of a Fluid-Elasticity
Interaction, SIAM J. Math. Anal., 47, 3, (2015), 1958-2000.
5. L. Bociu, J.-P. Zolésio, Linearization of a coupled system of nonlinear elasticity and viscous fluid, “Modern
Aspects of the Theory of Partial Differential Equations”, in the series “Operator Theory: Advances and
Applications”, 216, (Springer, Basel, 2011), 93-120.
6. L. Bociu, J.-P. Zolésio, Existence for the linearization of a steady state fluid - nonlinear elasticity
interaction, DCDA-S, (2011), 184-197.
7. L. Bociu, J.-P. Zolésio, Sensitivity analysis for a free boundary fluid-elasticity interaction, EECT 2, (2012),
55-79.
8. P.G. Ciarlet, Mathematical Elasticity Volume I: Three-dimensional Elasticity, North-Holland Publishing Co.,
Amsterdam, 1988.
9. M.C. Delfour and J.P. Zolesio, Shapes and Geometries: Analysis, Differential Calculus and Optimization,
SIAM 2001.
10. L. Formaggia, A. Quarteroni, A. Veneziani Eds., Cardiovascular Mathematics. Modeling and simulation of
the circulatory system. Vol I., MS & A, (Springer-Verlag Italia, Milano, 2009).
11. T. Richter and T. Wick, Optimal Control and Parameter Estimation for Stationary Fluid-Structure
Interaction Problems, SIAM J. Sci. Comput., 35, 5, (2013), 1085-1104.
12. T. Wick and W. Wollner, On the differentiability of fluid-structure interaction problems RICAM-Report,
16, (2014).

QMC: Operator Splitting Workshop, Optimization and Control in Free ans Moving Boundary Fluid-Structure Interactions - Lorena Bociu, Mar 22, 2018

More Related Content

What's hot (17)

Similar to QMC: Operator Splitting Workshop, Optimization and Control in Free ans Moving Boundary Fluid-Structure Interactions - Lorena Bociu, Mar 22, 2018 (20)

More from The Statistical and Applied Mathematical Sciences Institute (20)

Recently uploaded (20)

QMC: Operator Splitting Workshop, Optimization and Control in Free ans Moving Boundary Fluid-Structure Interactions - Lorena Bociu, Mar 22, 2018