COMPUTATIONAL

METAMATERIAL DESIGN

Víctor D. Fachinotti

Pontificia Universidad Católica de Chile, April-May 2018

OUTLINE
• Computational metamaterial design
• Microscale analysis
• Multiscale problem as a macroscopic one with inhomogeneous material
• Macroscopic thermo-mechanical response as a function of microstructure
• Material design as an optimization problem
• Applications:
• Optimization of the mechanical response under thermal loads
• Optimization of the thermal response using free material optimization (FMO)
• Heat flux manipulation
• Design of easy-to-make devices using discrete material optimization (DMO)
• Design of easiest-to-make devices using topology optimization
• Mechanical cloaking
• Advantages of computational metamaterial design
• Perspectives
2
METAMATERIAL DESIGN

• MATERIAL DESIGN: to modify the microstructure of the material in a

macroscopic piece in order to obtain an optimal response of the
piece

• METAMATERIAL: the so-designed material, usually having

extraordinary effective properties:
• optical or acoustical camouflage /invisibility
• negative Poisson ratio
• negative thermal conductivity, thermal camouflage, etc.

3
Metamaterial with negative
Poisson ratio made by
dip-in direct-laser-writing
optical lithography
(Bückmann et al., Metamaterial that twitsts under compression
Advanced Materials, 2012) (Frenzel et al., Science, 2017)

4
 Metamaterial made of
PVC and PDMS for
cloaking elastic waves
(Stenger et al., PRL 2012)

Metamaterial as a laminate of copper (A)

and polyurethane (B) for heat flux
inversion (Narayana & Sato, PRL 2013)

5
MACROSCOPIC BODY WITH VARIABLE
MICROSTRUCTURE
• Let the microstructure vary
throughout the
macroscopic domain, being
sampled at a series of
points 𝑿𝛼

• Each 𝑿𝛼 has its own

Representative Volume
Element (RVE)

6
COMPUTATIONAL METAMATERIAL DESIGN

• Computational Metamaterial Design involves the computational

solution of a series of multiscale problems for changing
microstructure

MACRO-
micro-scale
SCALE Macroscopic
analysis Effective properties
response
ANALYSIS
(at each RVE)
(AT THE BODY W)
until finding the optimal macroscopic response

7
QUANTITATIVELY CHARACTERIZED
MICROSTRUCTURE
• Let the RVE at any sampling point 𝑿𝛼 ∈ Ω be characterized by a finite
(𝛼) (𝛼)
number of scalar (micro)parameters 𝑝1 , 𝑝2 , …

Ex.: Narayana &

Sato’s heat flux
inverter (PRL
2012)

Ω
𝛼 𝛼
 Effective properties at 𝑿𝛼 ∈ Ω = 𝑓(𝑝1 , 𝑝2 , … )
8
MACROSCOPIC BODY WITH VARIABLE
QUANTITATIVELY CHARACTERIZED
MICROSTRUCTURE

RVE
caracterized
by 𝒑 1

RVE
caracterized
by 𝒑 2

9
MICROSCALE ANALYSIS

10
MICROSCALE ANALYSIS

• Goal: determination of the effective properties as analytical functions

of the microparameters

Microscale
analysis
Analytical

𝜶 𝜶 𝐞𝐟𝐟 𝐩𝐫𝐨𝐩 𝐚𝐭 𝐗 𝜶
𝒑𝟏 , 𝒑𝟐 ,… Experimental 𝜶 𝜶
= 𝒇(𝒑𝟏 , 𝒑𝟐 , …)
+RSM

Numerical
+RSM

11
ANALYTICAL MICROSCALE ANALYSIS:
LAMINATE
Effective anisotropic conductivity
𝑑𝐴 𝑘𝐴 + 𝑑𝐵 𝑘𝐵 + 𝑑𝐶 𝑘𝐶
𝑘𝜆𝜆 =
𝑑𝐴 + 𝑑𝐵 + 𝑑𝐶
𝑑𝐴 + 𝑑𝐵 + 𝑑𝐶
𝑘𝜏𝜏 =
𝑑𝐴 𝑑𝐵 𝑑𝐶
+ +
𝑘𝐴 𝑘𝐵 𝑘𝐶
𝑘𝑥𝑥 = 𝑘𝜆𝜆 cos 2 𝜃 + 𝑘𝜏𝜏 sin2 𝜃
𝑘𝑦𝑦 = 𝑘𝜆𝜆 sin2 𝜃 + 𝑘𝜏𝜏 cos 2 𝜃
𝑘𝑥𝑦 = 𝑘𝑦𝑥 = (𝑘𝜆𝜆 −𝑘𝜏𝜏 )cos 𝜃 sin 𝜃

12
EXPERIMENTAL+NUMERICAL MICROSCALE
ANALYSIS: PAPER
• Using upscaling techniques, discrete element
simulations and X-ray microtomography of
the geometry of wood fibers and their bonds
and the architecture of the fibrous network,
Marulier (PhD thesis, 2013) determined the
homogenized elastic moduli:

𝑪orth = 1.14 × 109 𝜙 − 0,02 2 𝑨 𝑎

⟹ 𝑪𝑥𝑦 = 𝚯 𝜃 𝑪orth 𝜙, 𝑎 𝚯 𝜃 𝑇
– 𝜙: fiber content
– 𝑨(𝑎): fiber orientation tensor (response surface from experiments), 𝑎: orientation
intensity
– 𝚯 𝜃 : serves to rotates from 𝜆𝜏 to 𝑥𝑦, 𝜃: angle between the 𝑥 and 𝜆
* Collaboration with S. Le Corre (LTN Nantes) and L. Orgéas (LCNRS Grenoble)
13
NUMERICAL MICROSCALE ANALYSIS:
CANCELLOUS BONE
• Using FEM for a geometrically
parameterized cell, Kowalczyk (2006)
determined the homogenized elastic
moduli:
′
𝐶𝑖𝑗𝑘𝑙 = 𝑓 𝑡𝑐 , 𝑡𝑣 , 𝑡ℎ
′
⟹ 𝐶𝑖𝑗𝑘𝑙 = 𝑅𝑚𝑖 𝑅𝑛𝑗 𝑅𝑝𝑘 𝑅𝑞𝑙 𝐶𝑚𝑛𝑝𝑞
– 𝑡𝑐 , 𝑡𝑣 , 𝑡ℎ : geometric parameters
– 𝑹(𝜓1 , 𝜓2 , 𝜓3 ): 3D rotation tensor,
𝝍: rotation vector

* Collaboration with A. Cisilino & L. Colabella (INTEMA, Argentina)

14
 NUMERICAL MICROSCALE ANALYSIS: SOLID
WITH INCLUSIONS
• Using FEM on RVEs with variable
𝑏 and ℎ, we determined the effective
thermomechanical properties
𝑘𝑖𝑗 = 𝑘𝑖𝑗 𝑏, ℎ GRIDS FROM FEM
PARAMETRIC
𝐶𝑖𝑗𝑘𝑙 = 𝐶𝑖𝑗𝑘𝑙 𝑏, ℎ ANALYSIS

𝜕𝜎𝑖𝑗
= 𝑑𝑖𝑗 (𝑏, ℎ) POLYNOMIAL
𝜕𝑇
RESPONSE
SURFACES

* Fachinotti, Toro, Sánchez & Huespe,

Int. J. Solids & Struct. 2015
15
REDUCTION OF THE MULTISCALE PROBLEM

• Once you know the effective material properties as functions of the

microparameters 𝒑 from the microscale analysis, the multiscale
problem becomes a classic macroscopic problem with
inhomogeneous material properties

16
MACROSCOPIC THERMO-MECHANICAL
RESPONSE AS A FUNCTION OF
MICROSTRUCTURE

17
THERMOMECHANICAL RESPONSE IN A BODY
WITH VARIABLE MICROSTRUCTURE

18
THERMOMECHANICAL RESPONSE AS A
FUNCTION OF MICROSTRUCTURE
• Given the microstructure 𝑷 = 𝒑 1 , 𝒑(2) , … througout Ω:
1) solve the steady state FEM heat equation:
𝑇
𝛻𝝓 𝒌 𝒑 𝛻𝝓𝑑𝑣 𝑻= 𝝓𝑇 𝑞 wall 𝑑𝑠
Ω 𝜕Ωq
𝑲ther 𝑷 𝑸
⟹ 𝑻 = 𝑲ther 𝑷 −𝟏 𝑸 = 𝑻 𝑷
2) solve the FEM equilibrium equation:
𝑩𝑇 𝑪 𝒑 𝑩 𝑑𝑣 𝑼 = 𝚽 𝑇 𝒕wall 𝑑𝑠 − 𝑩𝑇 𝝈,𝑇 𝒑 Δ𝑇 𝑷 𝑑𝑣 = 𝟎
Ω 𝜕Ω𝜎 Ω
𝑲mech 𝑷 𝑭mech 𝑭ther 𝑷
−𝟏 [𝑭
⟹ 𝑼 = 𝑲mech 𝑷 mech − 𝑭ther 𝑷 ] = 𝑼 𝑷
• The macroscopic thermo-mechanical response is the nonlinear function
ℛ = 𝑓 𝑼 𝑷 , 𝑻 𝑷 , 𝑷 = ℛ(𝑷) 19
MATERIAL DESIGN AS AN OPTIMIZATION
PROBLEM
• To design a material consists of finding the optimal set
(1) (1) (2) (2) opt
𝑷opt = 𝑝1 , 𝑝2 , … , 𝑝1 , 𝑝2 , …
such that
ℛ 𝑷opt = min ℛ 𝑷
𝑷
subject to
𝑎𝑖 ≤ 𝑃𝑖 ≤ 𝑏𝑖 Bound constraints
𝑐𝑖 𝑷 ≥ 0 Inequality constraints
𝑑𝑖 𝑷 = 0 Equality constraints
• This is a nonlinear constrained optimization problem with a usually
large number of design variables

20
 MATERIAL DESIGN FOR OPTIMAL
MACROSCOPIC MECHANICAL RESPONSE
UNDER THERMAL LOADS

with
S. Toro, P. Sánchez & A. Huespe (CIMEC)

21
THERMAL DEFLECTION OF A CANTILEVER
PLATE
T=50°C
STEEL
0.3 m E=2e11Pa
T=0°C n=0.3
a=1e-5/°C
3m k=36.5W/(m°C)

COPPER
E=1.2e11Pa
n=0.34
a=1.7e-5/°C
k=384W/(m°C)
22
MAXIMAL DEFLECTION
To maximize the deflection of the tip of the beam consists of finding
1 1 2 2 opt
𝑷opt = 𝑏 ,ℎ ,𝑏 ,ℎ ,… such that
−𝑢𝑦(𝑷opt) = min(−𝑢𝑦 (𝑷)) Periodic RVE
𝑷 𝑏 𝑛

1 𝑛
ℎ
𝑛
𝑿

uy

Copper
Steel
23
MAXIMAL DEFLECTION:
OPTIMIZATION PROBLEM
To maximize the tip deflection, we solve the nonlinear constrained
optimization problem
min
2𝑁
(−𝑢𝑦 (𝑷))
𝑷∈ℝ

subject to
𝑛
0≤ℎ ≤1
0≤𝑏 𝑛 ≤1 𝑛 = 1,2, … , 𝑁 = 287 = #𝑛𝑜𝑑𝑒𝑠
 EFFECTIVE PROPERTIES AS FUNCTIONS OF
MICROSTRUCTURE

Grids from FEM

microscale
homogenization
analysis

Polynomial
response
surfaces

25
MAXIMAL DEFLECTION:
OPTIMAL SOLUTION

Thickness of the vertical layers Thickness of the horizontal layers

(100% copper)

(100% steel)

26
MAXIMAL DEFLECTION:
VERTICAL DISPLACEMENTS

Copper beam Optimal beam

uy = 1.361 uy,copper

27
MINIMAL DEFLECTION:
OPTIMIZATION PROBLEM
To minimize the tip deflection, we solve the nonlinear constrained
optimization problem
min
2𝑁
(𝑢𝑦 (𝑷))
𝑷∈ℝ

subject to
𝑛
0≤ℎ ≤1
0≤𝑏 𝑛 ≤1 𝑛 = 1,2, … , 𝑁 = 287 = #𝑛𝑜𝑑𝑒𝑠
MINIMAL DEFLECTION:
OPTIMAL SOLUTION

Thickness of the vertical layers Thickness of the horizontal layers

(100% copper)

(100% steel)

29
MINIMAL DEFLECTION:
VERTICAL DISPLACEMENTS

Steel beam Optimal beam

uy = 0.527 uy,steel

30
 MATERIAL DESIGN FOR OPTIMAL
MACROSCOPIC THERMAL RESPONSE USING
FREE MATERIAL OPTIMIZATION (FMO)

with S. Giusti
(GIDMA, UTN Córdoba, Argentina)

31
 FREE MATERIAL OPTIMIZATION OF THE
THERMAL RESPONSE
• FREE MATERIAL OPTIMIZATION (FMO): the design variables are the
effective properties themselves
1 1 2 2 𝑛 𝑛 𝑛
• For 𝑷 = 𝑘𝑥𝑥 , 𝑘𝑦𝑦 , 𝑘𝑥𝑥 , 𝑘𝑦𝑦 , … (with 𝑘𝑥𝑥 , 𝑘𝑦𝑦 , and 𝑘𝑥𝑦 = 0 being
the effective conductivities at node 𝑛), let us find
𝑷opt = arg min 𝑖∈𝐴𝐵 (𝑇𝑖 (𝑷) − 200°C) 2
𝑷
subject to
𝑛 , 𝑛 ≤1
0.001 ≤ 𝑘𝑥𝑥 𝑘𝑦𝑦

32
INITIAL TEMPERATURE DISTRIBUTION
Initial guess: 𝑘𝑥𝑥 = 𝑘𝑦𝑦 = 0.5

33
OPTIMAL DISTRIBUTIONS OF CONDUCTIVITIES

kxx

kyy
34
TEMPERATURE FOR THE OPTIMAL SOLUTION

35
DETERMINATION OF THE MICROSTRUCTURE
• Knowing the optimal macroscopic 𝑘𝑥𝑥 and 𝑘𝑦𝑦 at a point of
the mesh, a topology optimization problem is solved to
determine the microstructure we need to achieve such 𝑘𝑥𝑥
and 𝑘𝑦𝑦

• Topology optimization using the topology derivative

approach

36
TOPOLOGY OPTIMIZATION AT THE MICROSCALE
3 6 9 3 6 9

2 5 8

2 5 8
1 4 7

1 4 7

37
COMPUTATIONAL METAMATERIAL DESIGN
FOR HEAT FLUX MANIPULATION

with
I. Peralta, A. Ciarbonetti (CIMEC)

38
 MANIPULATING THE HEAT FLUX
Prescribed
2 1 1 boundary flux
𝑿 3 𝒒 𝒒 𝑿
𝒒 3 2
𝑿
Prescribed
boundary 𝛀
temperature

• Given 𝒒 𝑞 as the desired heat flux at 𝑿 𝑞

, 𝑞 = 1,2, … , 𝑁𝑞 ,
you have to find 𝑷 such that
𝑞
−𝒌 𝒑 grad 𝑇 𝑷 𝑿𝑞 =𝒒 for 𝑖 = 1,2, … , 𝑁𝑞
39
HEAT FLUX MANIPULATION AS AN
OPTIMIZATION PROBLEM
• In order to perform the given task as well as possible, let us solve
the nonlinear optimization problem
1 𝑁𝑞 𝑞 2
min 𝑞=1 −𝒌 𝒑 grad 𝑇 𝑷 𝑿𝑞 −𝒒
feasible 𝑷 𝑁𝑞
MSE 𝑷
subject to constraints accounting for, at least, the feasibility of the
microstructure.

 Maybe, MSE 𝑷 ≠ 0 for feasible 𝑷

 We’ll find the “optimal” feasible 𝑷

40
 DESIGN OF A HEAT FLUX CONCENTRATION
AND CLOAKING DEVICE
• To find 𝑷opt = [𝑑 1 , 𝜃 1 ,…, 𝑑 𝑁 ,𝜃 𝑁 ]opt (𝑁 = 1896 is the # elems
in Ωdevice ) such that
1 2
𝑷opt = arg min 𝑞 −𝒌 𝒑 grad 𝑇 𝑷 𝑿𝑞 −𝒒 𝑞
𝑷 𝑁𝑞
with
𝒒 𝑞 = 5𝒒0 in Ωconc
1 2
𝒒 𝑞 = 𝒒0 in Ωcloak and Ωcloak
subject to the box constraints
0 ≤ 𝑃2𝑒−1 ≡ 𝑑 𝑒 ≤1
0 ≤ 𝑃2𝑒 ≡ 𝜃 𝑒 ≤𝜋

41
HEAT FLUX CONCENTRATION AND CLOAKING:
OPTIMAL METAMATERIAL DISTRIBUTION

Fraction of copper Fraction of PDMS Orientation

42
HEAT FLUX CONCENTRATION AND CLOAKING:
OPTIMAL CONDUCTIVITY DISTRIBUTION

43
HEAT FLUX CONCENTRATION AND CLOAKING:
OPTIMAL TEMPERATURE DISTRIBUTION

* Peralta, Fachinotti & Ciarbonetti, Scientific Reports, Jan. 2017

(http://www.nature.com/articles/srep40591)
44
EASY-TO-MAKE HEAT FLUX MANIPULATING
DEVICES USING DISCRETE MATERIAL
OPTIMIZATION (DMO)

with
I. Peralta (CIMEC)

45
MULTIPHASE TOPOLOGY OPTIMIZATION

• The material at the element Ω 𝑒 is either one of 𝑀 predefined,

candidate materials with conductivities 𝒌1 , 𝒌2 , … , 𝒌𝑀
• Each material maybe a metamaterial itself
𝑒
• The design variables for Ω 𝑒 are the fractions 𝑓𝑚 of each material
𝑚 = 1,2, … , 𝑀
• The conductivity at Ω 𝑒 is defined by the mixture law
𝑒 𝑒 𝑒
𝒌 𝑒 = 𝑓1 𝒌1 + 𝑓2 𝒌2 + ⋯ + 𝑓𝑀 𝒌𝑀
• We must use an optimization algorithm driving to optimal solutions
𝑒 𝑒
with 𝑓𝑚 = 1 or 𝑓𝑚 = 0
• An integer optimization algorithm (e.g. GA) is unaffordable given the
large number of design variables

46
DISCRETE MATERIAL OPTIMIZATION

• Using the Discrete Material Optimization (DMO) approach

proposed by Stegmann & Lund (IJNME, 2005), we define:
𝑓𝑚∗ (𝒑 𝑒 )
𝑒 𝑒
𝑓𝑚 = ∗ 𝑒 = 𝑓𝑚 (𝒑 )
𝑖 𝑓𝑖 𝒑
𝑝 𝑝
𝑒 𝑀 𝑒
with 𝑓𝑚∗ 𝒑 𝑒
= 𝜌𝑚 𝑗=1,𝑗≠𝑚 1− 𝜌𝑚
𝑒 𝑒 𝑒 𝑒 𝑒
• The design variables for the finite element Ω are 𝑝 = [𝜌1 , 𝜌2 , … , 𝜌𝑀 ]
𝑒 𝑒
𝜌𝑖 : artificial density of material 𝑚 at Ω , like in Topology Optimization
• 𝑝 ≥ 3, like in SIMP for Topology Optimization
𝑒 𝑒
• This definition forces 𝜌𝑗≠𝑖 → 0 when 𝜌𝑖 →1
𝑒
• It doesn’t need a constraint (one per finite element!) to make 𝑖 𝑓𝑖 =1
47
 DESIGN OF A HEAT FLUX CONCENTRATION
AND CLOAKING DEVICE USING DMO
1 1 1 1896 1896 1896 opt
• To find 𝑷opt = [𝜌1 , 𝜌2 , 𝜌3 ,…, 𝜌1 , 𝜌2 , 𝜌3 ] such that
1 2
𝑷opt = arg min 𝑞 −𝒌 𝒑 grad 𝑇 𝑷 𝑿𝑞 −𝒒 𝑞
𝑷 𝑁𝑞
subject to the box constraints
0 ≤ 𝑃𝑖 ≤ 1
with
𝒒 𝑞 = 5𝒒0 in Ωconc
𝑞 1 2
𝒒 = 𝒒0 in Ωcloak and Ωcloak

48
HEAT FLUX CONCENTRATION AND CLOAKING
USING DMO:
OPTIMAL METAMATERIAL DISTRIBUTION

49
HEAT FLUX CONCENTRATION AND CLOAKING
USING DMO:
FULLY DISCRETE METAMATERIAL DISTRIBUTION

50
HEAT FLUX CONCENTRATION AND CLOAKING
USING DMO:
OPTIMAL TEMPERATURE DISTRIBUTION

* Peralta & Fachinotti, Scientific Reports, July 2017

(http://www.nature.com/articles/s41598-017-06565-6)
51
 EVEN EASIER-TO-MAKE HEAT FLUX
MANIPULATING DEVICES USING TOPOLOGY
OPTIMIZATION

with
A. Ciarbonetti, I. Peralta (CIMEC),
I. Rintoul (INTEC)

52
TOPOLOGY OPTIMIZATION

• The material at the element Ω 𝑒 is either one of two predefined

candidate materials with conductivities 𝒌1 , 𝒌𝟐
• Each material is isotropic
𝑒 𝑒
• There is only one design variable for Ω : the artificial density 𝜌
of material 1
• The conductivity at Ω 𝑒 is defined using SIMP (Solid Isotropic
Material with Penalization)
𝑒 𝑒 𝑝 𝑒 𝑝
𝒌 = 𝜌 𝒌1 + 1 − 𝜌 𝒌2
𝑒 𝑒
• A priori, using 𝑝 ≥ 3, 𝜌 → 0 or 𝜌 →1 for the optimal solution

53
DESIGN OF HEAT FLUX INVERTER USING
TOPOLOGICAL OPTIMIZATION
• To find 𝑷opt = [𝜌 1
,…, 𝜌 4000 opt
] such that
1 2
𝑷𝑜𝑝𝑡 = arg min 𝑞 −𝒌 𝒑 grad 𝑇 𝑷 𝑿𝑞 −𝒒 𝑞
𝑷 𝑁𝑞
subject to the box constraints
0 ≤ 𝜌(𝑒) ≤ 1
with
𝒒 𝑞 = −𝒒0 in Ωinvert

54
HEAT FLUX INVERTER:
TOPOLOGY OPTIMIZATION SOLUTION

55
HEAT FLUX INVERTER:
BLACK AND WHITE FILTERING

• For manufacturability, regions with intermediate material fractions

(“grey zones”) must be avoided
• Black and white filters (Sigmund 2007) serve to this end
• Here, a simple a posteriori b&w
filter is preferred: material
fraction greater than 𝑤 ∗ is taken
to 1; otherwise, it is taken to 0

56
HEAT FLUX INVERTER:
TOPOLOGY OPTIMIZATION SOLUTION
+ BLACK AND WHITE FILTERING

57
HEAT FLUX INVERTER:
TOPOLOGY OPTIMIZATION WITH AND
WITHOUT BLACK AND WHITE FILTERING

WITHOUT B&W FILTER WITH B&W FILTER

58
HEAT FLUX INVERTER:
TOPOLOGY OPTIMIZATION WITH AND
WITHOUT BLACK AND WHITE FILTERING

WITHOUT B&W FILTER WITH B&W FILTER

59
HEAT FLUX INVERTER:
EXPERIMENTAL VALIDATION
Computationally
designed device

60
HEAT FLUX INVERTER:
EXPERIMENTAL VALIDATION

61
HEAT FLUX INVERTER:
EXPERIMENTAL VALIDATION

V. Fachinotti et al., “Optimization-based design of easy-to-make

devices for heat flux manipulation”, Int. J. Ther. Sci., 2018 62
HEAT FLUX INVERTER:
COMPARISON WITH NARAYANA AND SATO´S
INVERTER
• Accomplishment of the inversion task
𝑇𝐷 −𝑇𝐶
𝑞invert = −𝑘agar = −𝛼𝑞0
𝐶𝐷

𝐶 𝐷

𝛼 = 0.997 𝛼 = 0.778 𝛼 = 0.774 𝛼 = 0.395

S. Narayana & Y. Sato, “Heat Flux Manipulation with Engineered
Thermal Materials”, Physical Review Letters 2012 63
MECHANICAL CLOAKING

with
I. Peralta (CIMEC)

64
MECHANICAL CLOAKING

Displacement 𝒖

Displacement 𝒖0 Ωdev
Ωincl Ωcloak

• To cloak the inclusion Ωincl , let us design the material in Ωdev such
that the displacement in Ωcloak ressembles 𝒖0
• The cloaking task consists of finding the material distribution 𝑷 in
Ωdev such that
𝒖 𝒙𝑖 , 𝑷 = 𝒖0 𝒙𝑖 ∀𝒙𝑖 ∈ Ωcloak
65
MECHANICAL CLOAKING AS AN
OPTIMIZATION PROBLEM
Let us accomplish the cloaking task as well as possible by solving the
nonlinear constrained optimization problem
1
min 𝒖 𝒙𝑖 , 𝑷 − 𝒖0 𝒙𝑖 2
𝑷 𝑁cloak
𝒙𝑖 ∈Ωcloak
subject to bound, equality and/or inequality constraints

66
CLOAKING OF A HOLE IN A PLATE
• The displacement in a nylon plate is
originally 𝒖0
• Once the plate is holed, let us design a
device Ωdev to cloak the hole Ωincl
• Using DMO,
1 1 15084 15084
𝑷 = 𝜌Al 𝜌PTFE … 𝜌Al 𝜌PTFE
• The optimal 𝑷 is the solution of
1 2
min 𝒖 𝒙𝑖 , 𝑷 − 𝒖0 𝒙𝑖
𝑷 19972
𝒙𝑖 ∈Ωcloak PTFE or aluminum
Nylon
subject to
0 ≤ 𝑃𝑖 ≤ 1 𝑖 = 1,2, … , 30168

67
CLOAKING OF A HOLE IN A PLATE

68
PERSPECTIVES
• Robustness
• instabilities (checkerboard)
• grey zones
• convergence
• 3D
• Multiobjective optimization
• Applications
• Isolation: to deviate the heat flux from the zones where it is undesired, to
drive it to somewhere where it maybe useful
• Optimization of Austempered Ductile Iron (with B. Tourn)
• Mechanical properties depend on the thermal history
• Topology and heat treatment optimization to make a macroscopic
piece have a given mechanical response
• Metamaterials for wind turbine blades (with A. Albanesi)
• Fabrication, patents 69
ACKNOWLEDGEMENTS

• European Research Council, Grant Agreement n. 320815 “COMP-

DES-MAT: Advanced Tools for Computational Design of Engineering
Materials”, leaded by X. Oliver (CIMNE-UPC), 2013-2017.
• National Littoral University (UNL), Santa Fe, Argentina, research
project CAI+D 2016 50420150100087LI, “Metamaterials:
Computational Design, Thermal, Mechanical and Acoustic
Applications, and Prototyping”, 2017-2019.
• National Agency for the Promotion of Science and Technology of
Argentina (ANPCyT), research project PICT 2016-2673
“Computational Design of Metamaterials”, 2017-2020.

70