Optimization of Functionally Graded

Materials Considering Dynamical Analysis

F.J. Ramírez-Gil, J.E. Murillo-Cardoso, E.C.N. Silva

and W. Montealegre-Rubio

Abstract Functionally graded materials (FGMs) are a new class of bio-inspired

composite materials made from different material phases, in which their volume
fraction changes gradually towards a particular direction. Accordingly, continuous
changes in the composition, microstructure and porosity of the graded materials
results in material properties gradients; for this reason, the material properties move
smoothly and continuously from one surface to another, eliminating the interface
problem. Hence, with appropriate design, FGMs can develop better properties than
their homogeneous counterpart due to their better designability. One potential
employment of FGMs is as damper or energy absorber in dynamic applications, in
which optimization techniques such as the topology optimization method
(TOM) can contribute to a better performance in relation to a non-optimized design.
In this chapter, functionally graded structures are designed with and without the
TOM in order to explore the advantages of the FGM concept in low-velocity impact
condition, which is a special case in the world of dynamic analysis, and has interest
for designing machinery parts and components.

⋅
Keywords Structural optimization Topology optimization ⋅ FGM ⋅ Composite
⋅ ⋅
materials Low velocity impact Finite element modeling

F.J. Ramírez-Gil ⋅ J.E. Murillo-Cardoso ⋅ W. Montealegre-Rubio (&)

Faculty of Mines, Department of Mechanical Engineering,
Universidad Nacional de Colombia, 050034 Medellín, Colombia
e-mail: [email protected]
E.C.N. Silva
Department of Mechatronics and Mechanical Systems Engineering,
Escola Politécnica of the University of São Paulo,
Butantã, São Paulo 05508-900, Brazil

© Springer International Publishing Switzerland 2016 205

P.A. Muñoz-Rojas (ed.), Computational Modeling, Optimization
and Manufacturing Simulation of Advanced Engineering Materials,
Advanced Structured Materials 49, DOI 10.1007/978-3-319-04265-7_8
206 F.J. Ramírez-Gil et al.

1 Introduction

An impact is defined as the collision of two or more solid bodies within a short time
(typically, in the order of millisecond and microseconds depending on the velocity
of the impacting objects), in which the load generated has high intensity. Situations
involving impact have historically received great attention. Traditionally, the prime
interest in this area has been for military applications. However, advances in
technology have placed such demands on materials and structures on other problem
applications such as [1]:
• safety transportation of hazardous materials;
• vehicle crashworthiness;
• safety of nuclear reactor structures;
• protection of military vehicles, structures, and aircraft;
• design of lightweight armor systems;
• erosion and fracture of solids due to liquid and solid particle;
• impact protection of spacecraft from meteoroid impact;
• explosive forming and welding of metals;
• and the design of machinery tools for manufacturing processes.
Impact on structures involves dynamic effects of particular interest, because their
effect is potentially catastrophic although their probability of occurrence is gener-
ally lower than for other types of loads. Velocity is perhaps the simplest parameter
to classify the different types of impacts, even though other variables of geometric
type and/or related to material properties are critical. Several classifications have
been suggested, but the most common is the following, which summarizes the
effects on the material [2]:
• Low-velocity impacts (v < 50 m/s). Elastic effects or localized plastic
deformation.
• Medium-velocity impacts (50 m/s < v < 500 m/s). Generalized plastic
deformation.
• High-velocity impacts (500 m/s < v < 2000 m/s). The viscous resistance of the
material still matters.
• Hypervelocity impacts (2000 m/s < v). The material can be considered a
hydrodynamic fluid where the density is the main parameter.
Additionally to the above classification, there are some phenomena that can occur
in the structure depending on the impact velocity, some of them are:
• Dynamic and structural vibrations: relevant in low-velocity impacts in which the
structure geometry is predominant, and it can be studied by implicit or explicit
integration methods.
• Stress and shock waves propagation: the effect of stress waves are important in
low- and medium-velocity impacts, which become shock waves at hyperve-
locity impact.
Optimization of Functionally Graded Materials … 207

• Nonlinear material behavior: some phenomena such as plasticity, fracture, and

dependence on strain rate and temperature can occur. The nonlinearities increase
with the impact velocity; even for very high velocities, the material has a fluid
behavior where its resistance can be neglected.
• Large displacements and deformations: changes in geometry and finite rotations
that influence the loads and their effect.
• Contacts and interface phenomena: contact is the key to any impact model, since
loads are transmitted by it.
• Penetration and perforation: when the impacting object has enough velocity, it
can penetrate (does not cross the material thickness) or perforate (crossing the
material thickness) the impacted object.
• Local phenomena of rupture, such as spalling, scabbing, petalling and plugging.
As shown above, depending on the impact type or impact velocity, the analysis
developed is different. For example, at low-velocity impact, the material properties
have influence since the behavior is predominantly elastic and eventually occurs
local plasticity. In the elastic region, the Young´s modulus, Poisson’s ratio, and
material’s density and damping have the primary influence. Herein, impact forces
classified as low-velocity are considered, in which only dynamic phenomena,
structural vibrations and stress wave propagation are presented. Therefore, material
nonlinearities are avoided, which means that plasticity, penetration, perforation and
local rupture are not considered because these phenomena mainly appear above
medium-velocity impacts and require more elaborated analysis and pose greater
difficulties to simulation. In that sense, the material is modelled as a linear elastic
isotropic material; however, geometric nonlinearities are taken into account to
include large-deflection effects in the transient analysis. Finally, contact is not
considered since an impulsive pressure load is used to represent the impact. More
details about simulation considerations are described in subsequent sections.

2 Functionally Graded Materials

The material used for impact depends on the specific application. For example,
ceramics are hard, which is an important characteristic in high-velocity impacts to
decelerate and to erode the impacting object. However, the lack of ductility in
ceramics is a problem because a ductile element is required to absorb, through
deformation, whole or part of the kinetic energy of the impacting object. On other
hand, metals are good energy absorbers due to their enhanced ductile behavior;
however, their high density represents the major problem because the structures
built with these materials are heavy. For that reason, polymers appear as an alter-
native due to their low density, ranging from 1.0 to 1.4 g/cm3 [3]. The most
common polymers for impact applications are polymeric fibers. These fibers have
been found to be very efficient due to their improved specific strengths (the
strength/density ratios). More efficient materials are based on aramid fibers
(e.g. Kevlar6®), ultra-high molecular weight polyethylene fibers (e.g. Spectra® and
208 F.J. Ramírez-Gil et al.

Dyneema®) and the PBO fiber (e.g. Zylon®). Despite the advantages of polymers,
these materials are not commonly used alone for impact applications due to their
low resistance. Thus, materials with different properties need to be assembled in the
most advantageous way for impact applications [4].
The discussion above shows that to dissipate the energy of an impact, the
material should contain several properties. Composite materials are an adequate
option for this kind of application since they can combine different properties. This
idea is supported by the pioneering research on ceramic armor, which concluded
that the material performance could be significantly enhanced with materials
macrostructurally designed for grading from a pure and stiff ceramic to a ductile
material on the back surface [5]. In recent years, these types of materials have been
referred to as functionally graded materials (FGMs) [6].
FGMs are also frequently found in nature. Bones, teeth and tree stalks are FGMs
designed by nature to meet their expected service requirements. For this reason, the
human designs based on the FGM concept are considered bioinspired
materials/structures. Some examples of biological FGM are shown in Fig. 1; these
nature designs combine high strength with lightweight materials, suitable properties
for many applications. Using FGMs means to emulate nature to solve structural
engineering problems, in the same way as an artificial neural network is used to
emulate the human brain.
An example of a human-made FGM (artificial) is shown in Fig. 2 in which a
micrograph reveal the spatially varying composition between two materials, where
the matrix and inclusions change through thickness following a graded function

Pore gradation

Fiber gradation

Cross section of a palm stem and its Cross section of dinosaur

graphical representation of the fiber bone
distribution

Graphical representation
of the pore distribution

Pore gradation

Cross section of an elk horn

Fig. 1 Natural lightweight FGM examples for impact applications

Optimization of Functionally Graded Materials … 209

Material A on the top surface (Ni)

From properties
Rich material A (Ni) region
of material A (Ni)

Material A (Ni) with

inclusions of material B (Cu)

Transition region

Material B (Cu) with

To properties of
inclusions of material A (Ni)
material B (Cu)
Rich material B (Cu) region

Material B on the bottom surface (Cu)

Fig. 2 Microstructure of a FGM graded from material A (Ni) to material B (Cu)

intentionally selected. This gradation provides FGMs with the ability to combine
the desired properties of different materials; for example, a metal and a ceramic are
used as base materials for taking advantage of the high temperature and corrosion
resistance of ceramics, as well as the ductility and toughness of metals.
Functionally graded or gradient materials were first proposed in 1984 aiming to
prepare thermal barrier materials exclusively for space application [6]. Functionally
graded means that continuous and smooth changes can be obtained in the com-
positions, microstructure, porosity, etc. in some direction in the material [7]. Hence,
the properties of a typical FGM plate change continuously from one surface to
another through its thickness direction. This FGM design is intended for taking
advantage of certain desirable features of each of the constituent phases [8]. Thus,
FGMs have a number of advantages that make them attractive in potential appli-
cations, including a possible reduction of in-plane and transverse
through-the-thickness stresses, improved residual stress distribution, enhanced
thermal properties, higher fracture toughness, and reduced stress intensity factors
[9]. Due to its advantages as compared with homogeneous materials, the FGM have
found many applications in engineering, not only in the aerospace industry, but also
in the nuclear and the automobile industries [10].
Despite the number of FGM advantages, this class of materials has two major
drawbacks in their design [7]. The first problem is presented in the materials
selection to form the FGM, and the second problem consists in determining the
optimal spatial distribution of the material phases (composition profile).
In the context of low-velocity impacts, the first problem lies in the selection of a
set of suitable materials to support impact loads and simultaneously get a light-
weight structure. As stated before, metals are useful as energy absorbers; however,
they are high-density materials. In consequence, herein an artificially graded
porosity is introduced in the material, specifically, in a circular steel plate. The steel
is selected as a base material due to its popularity, manufacturability and avail-
ability, which makes steel less expensive and more usable than other metals.
Moreover, the induced porosity is meant to reduce the total weight of the structure
210 F.J. Ramírez-Gil et al.

designed with this material, similar to bamboo, bones and horns (see Fig. 1).
However, the material is modeled as a linear elastic isotropic material and not as a
porous one.
The second problem is the lack of a systematic design approach that allows the
selection of the composition profile which best accomplishes the intended purpose
of the material application, while maintaining other properties within limits that
ensure acceptable performance. Therefore, in this work the topology optimization
method (TOM) is proposed for finding the composition profile (gradation function)
along the thickness direction to provide an optimal pore distribution in a circular
plate under low-velocity impact loads. In addition, since for low-velocity impacts
where dynamics and structural vibrations effects mainly occur, the geometry is
predominantly as stated before. Thus, the design of mechanical parts subjected to
this kind of loads are suitable for TOM since this optimization technique provides
novel geometries, apart from other benefits, such as weight reduction.

3 Topology Optimization Method for FGM Design

Currently, there is a need of efficient engineering methodologies to design parts and

structures in order to save material, time and costs [11]. Thus, in recent years, there
has been increasing interest in methods that automatically obtain optimized struc-
tures from performance conditions specified by the designer [12]. Optimization is a
field of engineering that can be used to facilitate and to automate the design process.
A relatively new method for obtaining optimized designs automatically is the TOM.
The next subsections discuss the basis of TOM and its application to FGM design.

3.1 Basics of the Topology Optimization Method

The TOM is a powerful method in which the basic idea is to distribute one or more
materials within a predefined design domain containing the geometry and boundary
conditions of the problem. In TOM, an objective function is extremized seeking a
desired behavior in the structure to meet some designer requests [12]. An additional
typical goal in structural optimization is to obtain optimal structures with reduced
mass. The TOM commonly produces designs that require some post-processing;
TOM is thus considered to produce conceptual designs. These designs are mostly
non-intuitive and are usually novel designs [13].
The first work that served as the basis for the TOM was by Michell in 1904 [14],
who established the optimality conditions on loaded structures. From this work,
many others have contributed to the area. Specifically, in 1988 Bendsøe and
Kikuchi developed a computational method for topology optimization [15]. This
method consists mainly of two modules, the analysis module and the optimization
module [16]. The first is used to calculate the response of the structure against the
Optimization of Functionally Graded Materials … 211

physical phenomena considered, for which the finite element method (FEM) is the
tool most commonly used. The optimization module is used to update the design
variables that improve the designed structure.
The TOM can be applied to design discrete or continuum structures [13]. For
inherently discrete structures such as those based on beams and/or trusses, the TOM
must determine the optimal number, position and connectivity of structural mem-
bers. This area of research has been active for several decades and was mainly
developed by Prager and Rozvany [13]. Moreover, the TOM in continuum struc-
tures must determine the number of holes and the external and internal contour
shape. There are several investigations concerning the topology optimization of
continuous structures; it can roughly distinguish between two types of approaches:
the material or microstructural approach and the geometric or macrostructural
approach [13].
The geometric approach consists of an iterative process of positioning new holes
or “bubbles” at specific points in the design domain. At each iteration, the holes can
appear and disappear, and their contours are subjected to a shape optimization
process. This procedure is complicated since the mesh must change every iteration
to accurately approximate the new generated contours. Additionally, the “birth” and
“death” of the holes require another optimization technique [13].
In the material or microstructural approach, frequently a constant finite element
(FE) mesh is used to describe the geometry. Typically, the mesh is uniform, and the
design variables are assumed constant for each FE. The method consists in deter-
mining if an element in a continuous medium should contain material or not. Then,
the design variable is defined for each element with values of 1 to indicate solid
elements and 0 for void elements. The result is an approximate description of the
outer and inner contours of the continuous structure that represents the optimized
topology design.
The microstructural approach is preferred for simplicity. However, the topology
optimization problem formulated as 0–1 is a discrete optimization problem, which
is an ill-conditioned problem. Then, in order to obtain a well-conditioned problem,
an adjustment is required in the problem formulation, which can be achieved by
including additional restrictions on the problem formulation and/or by introducing a
continuous design variable [13]. Consequently, the microstructural approach uses a
basic concept known as material model.
The material model addresses the topology optimization problem by relaxing it
by varying design variables continuously between 0 and 1. There are several
material models used in the literature to solve topology optimization problems [17].
However, the models most commonly used in topology optimization apply the
homogenization method, which was first introduced in 1988 by Kikuchi and
Bendsøe [15], and the “power-law” or SIMP (solid isotropic material with penal-
ization) model, proposed a year later by Bendsøe [18].
The material models based on the homogenization method provide a regular-
ization of the optimization problem via relaxation (extension) of the design space,
and its periodicity implies that the effective mechanical properties of the
microstructures can be determined through homogenization [13]. There are some
212 F.J. Ramírez-Gil et al.

variants of this model, such as the hole-in-cell microstructures and layered 2D/3D
microstructures. Even though these methods are useful, the number of design
variables used in the optimization problem are higher than those used in the SIMP
model. Additionally, homogenization models require a dependence model for the
material properties with respect to the geometric parameters of the unit cell, which
complicates their implementation [11].
The SIMP model is a simple approximation used to relax the space of possible
solutions of the optimization problem without increasing the number of design
variables [19]. This material model does not regularize the topology optimization
problem; however, it can be well-conditioned if an additional constraint on the
formulation is used; for example, using a constraint on the perimeter for 2D
structures or surfaces in 3D structures or by using a filtered technique [13]. In the
SIMP model, the design variables are assigned to each FE and the material prop-
erties are assumed constant on it. The general form of SIMP is:

A = ρðxÞp A0 , 0 ≤ ρðxÞ ≤ 1, p > 1, x∈Ω ð1Þ

where A represents the material interpolated property, ρ is the design variable also
called pseudo-density, p is the penalization factor and A0 is the base or reference
material property. Vector x represents the position in the design domain Ω.
Depending on the dimensionality used in the TOM problem, vector x can be x = x
for one-dimension, x = f x y gT for two-dimension or x = f x y z gT for
three-dimension problems. With SIMP can appear some gray areas with interme-
diate pseudo-densities (composite material) that avoid the convergence to a
black-and-white design (solid and void—porous material), then, the penalization
factor is set to a value higher than 1 [19].
The SIMP model provides good results in topology optimization and is relatively
easy to implement in commercial finite element codes contrasted with the
homogenization methods [20, 21]. However, this model has disadvantages, such as
the topology dependence on input design parameters and mesh discretization;
nevertheless, these problems can be solved by using several strategies, such as those
presented in [22].
For solving the optimization problems there are several approaches which can be
grouped in three main categories: mathematical programming (MP), optimality
criteria (OC) and evolutionary programming (EP) methods. The MP techniques are
math-based methods for optimization [12]. The OC methods are rules intuitively or
rigorously derived if a closed-form formulation is expressed [23]. The EP methods
are heuristic or intuition-based approaches that use mechanisms inspired by bio-
logical evolution, such as reproduction, mutation and survival for finding an opti-
mal solution to the problem [24]. The MP and OC methods use continuous design
variables whereas EP methods typically use discrete representations as design
variables. However, the MP methods are preferred over the others due to their
generality, allowing the solution of complex and non-linear optimization problems.
Additionally, MP methods are more suitable to deal with optimization problems
with more than one constraint.
Optimization of Functionally Graded Materials … 213

The most common searching technique used in MP is the sequential linear

programming (SLP) due to its straightforward implementation. The SLP algorithm
searches for the optimum in a nonlinear design space using a sequence of linear
approximations for the objective function and constraints. It uses linear program-
ming to solve each linear sub-problem. The linear approximation is calculated with
gradient information by using the first-order Taylor expansion series, process
commonly known as sensitivity analysis.
Figure 3 shows the flowchart of the SLP-based topology optimization algorithm.
The first step of this process defines the design domain with loads and boundary
conditions. Next, the optimization iterative process starts with the finite element
analysis (FEA) of the structure under some considered phenomena. After this, the

Fig. 3 SLP-based topology optimization flowchart

214 F.J. Ramírez-Gil et al.

sensitivity analysis, which determines the change in the structural response of small
changes in the design variables, is carried out. The move limits are calculated in the
next step. The move limits are additional side constraints on the optimization
problems that define a region of the design space where the solution of the lin-
earized sub-problem will lie [25]. Next, in order to avoid some numerical insta-
bilities, such as mesh dependence and grey areas, a filter on the design variables,
sensitivities or move limits is computed [22]. After the above steps are finalized, a
linear optimization solver optimizes the linearized objective function. Some typical
linear optimization solvers are based on the simplex or Karmarkar algorithms.
Finally, a convergence criterion is evaluated to verify if the optimization process
can finish. When the iterative process ends, the optimal topology of the structure is
obtained.

3.2 Topology Optimization of FGMs

From the designer viewpoint, the introduction of gradient distribution in a material

provides a larger design domain, therefore providing a better designability. Con-
sequently, the graded material will outperform (or perform just as) its uniform
counterpart [26]. Optimization techniques are hence required in order to find an
optimal gradation for this type of materials.
The optimum topology optimization result is known to consist of a structure with
gray-scale areas, which means intermediate or composite material. In a typical
topology optimized structure, the gray areas are undesirable since they avoid a
precise definition of the boundaries. However, these gray-scale results given by the
TOM are strongly related to the concept of FGM materials, which essentially
considers a continuous transition of material properties [27].
The design of FGM materials using topology optimization has two approaches.
The first consists in finding the optimal property variation of the FGM in the
domain for achieving the design requirements. The second approach consists of a
predefined functionally graded design domain, where the properties change in a
particular direction according to a specified model. These FGM approaches differ in
that in the first case, all the elements are in the final design, i.e., they fill the whole
design domain; however, their material properties are not known a priori, and the
TOM finds them. In the second case, each FE has a specified property; however,
this varies from one element to the other continuously following a predefined
graded function. Hence, the elements that occupy the final design are not known
and the TOM has to find them.
Depending on the FGM optimization approach, a material model is used. For the
first approach, a layered FGM material model is used. This is because current FGM
manufacturing techniques emphasize layered systems, and a layered material
constraint is adopted. In this case, two-phase material systems are considered. Thus,
Optimization of Functionally Graded Materials … 215

the FGMs represent the transition between two base materials, and the objective is
to find the optimal FGM property variation such that the material model allows
local distribution of the two materials in the domain. The material model used is
based on a density method approach, defined for the Lame constants [27].
The material model for the second FGM optimization approach is the
FGM-SIMP material model. This new material model is similar to the traditional
SIMP; the difference is that in the FGM-SIMP, a design domain known as FGM
domain is used. The FGM domain presents a continuous variation in the material
properties, defined by the designer. The TOM must find the elements that satisfy the
objective function and constraints, as in a traditional TOM problem. Thus, in the
FGM-SIMP material model, property A0 considered in Eq. (1) for the traditional
SIMP is not constant along the design domain, but it depends on position x as:

A = ρðxÞp A0 ðxÞ, x∈Ω ð2Þ

In traditional topology optimization formulations, the design variable is defined in a

piecewise fashion in the discretized domain, which means that the continuity of the
material distribution does not occur between FEs (see Fig. 4a). However, consid-
ering the topology optimization results as an FGM-type, a more natural way of
representing the material distribution emerges by using the concept of the graded
finite elements (GFE) [27]. The GFE leads to a continuous representation of
material properties that are interpolated within the finite element using the FE shape
functions (see Fig. 4b) [28]. Some works have suggested the continuum distribution
of the design variable within the FE in the topology optimization formulation [29].

Property defined at
Property defined integration points of FE
at center of FE

*
* * *
*
x, y or z direction

x, y or z direction

Property Property

Fig. 4 Finite element modeling of FGMs: a Homogeneous finite element and, b graded finite
element
216 F.J. Ramírez-Gil et al.

This formulation is known as CAMD (continuous approximation of material dis-

tribution) which is a variation of the SIMP model. By means of the continuum
model approach, the design of FGM structures can be fully achieved by applying
topology optimization because a continuous change in material properties is con-
sidered within the design domain.

4 Topology Optimization of Dynamically Loaded

Structures

In this section, the TOM for solving dynamic optimization problems is described.
Since a TOM problem requires two modules (the analysis and optimization mod-
ules) to be solved, some methods for solving the dynamic analysis problem are first
introduced; after this, the addressed methodology to solve the optimization problem
is exposed.

4.1 Dynamic Finite Element Analysis

The structures under low-velocity impacts can be analyzed as an elastic dynamic

phenomenon as previously stated. However, there are three main dynamics analysis
types: modal, harmonic and transient analysis [31]:
• The modal analysis is used to calculate the natural frequencies and mode shapes
of a structure or a machine component. It can also serve as a starting point for
another, more detailed, dynamic analysis, such as a transient dynamic analysis, a
harmonic analysis, or a spectrum analysis.
• The harmonic analysis is used to determine the response of a structure to har-
monically time-varying loads (cyclic loads). This type of analysis is useful to
analyze structures subjected to resonance, fatigue, and other harmful effects of
forced vibrations.
• The transient analysis is used to evaluate structures under loads that are arbi-
trarily time-varying. In an interval of time, the loads can change their magnitude,
direction and/or position. Within this analysis, the inertial and damping effects
are considered important.
Consequently, since for low-velocity impacts the structure is under the action of a
general time-dependent load, thus the transient analysis is preferred over the others.
However, there are two solution methods for the transit analysis, which are
mode-superposition and full transient. The mode-superposition method is com-
monly used in analysis where nonlinearities are not included for being less com-
putationally expensive that the full method. Nevertheless, the full method is a more
general method allowing any kind of nonlinearity and any type of load. Therefore,
Optimization of Functionally Graded Materials … 217

since geometric nonlinearities are considered herein in order to allow

large-deflection effects in the transient analysis, the full method is selected.
The most generic approach for solving the dynamic response in structures is the
direct numerical integration of the dynamic equilibrium equation [31]:
..
M U + CU̇ + Rint = Rext ð3Þ

where M is the mass matrix, C is the damping matrix, and Rint and Rext are the
internal and external loads, respectively. Eq. (3) is valid for both linear and non-
linear material properties. If the material is linearly elastic, then internal loads are
Rint = KU, where K is the stiffness matrix, and Eq. (3) becomes:
..
M U + CU̇ + KU = Rext ð4Þ
..
being U, U̇ , U the displacements, velocities, and accelerations of the structure,
respectively. Matrices M, C and K are obtained as a sum (assembly) of local
matrices:

nel nel nel

M = ∑ m, C = ∑ c, K= ∑ k ð5Þ
e=1 e=1 e=1

where the symbol ∑ represent the assembly operation, nel is the number of FEs in
the mesh (discretized design domain) and the local matrices can be obtained as:
Z Z Z
m= ρN NdV,
T
c= T
cd N NdV, k= BT DBdV ð6Þ

with shape functions matrix N, gradient matrix B and material properties matrix D,
ρ represents mass density and cd is a damping parameter analogous to viscosity. In
Eq. (6), m and c are identified as “consistent” element mass and damping matrices,
which emphasizes that these forms follow directly from the FE discretization, and
use the same shape functions as the element stiffness matrix [31]. However,
non-consistent forms of these matrices are also used.
The consistent mass matrix m is symmetric, full, and positive definite. However,
there is a simpler formulation for the mass matrix known as “lumped”. This
approach consists in placing concentrated masses at nodes, which produce a
diagonal mass matrix. Although the lumped formulation has computational
advantages, such as less storage space and processing time, this formulation can
require especial solution algorithms once the lumped matrices are positive
semidefinite or, indeed, indefinite. Additionally, for implicit integration methods, a
diagonal mass matrix provides little computational economy and a consistent for-
mulation provides more accuracy [31]. Thus, the consistent formulation is chosen
for this work.
218 F.J. Ramírez-Gil et al.

The damping matrix can be generated from Eq. (6) or by using other formula-
tions. Damping dissipates energy by limiting the amplitude of vibration produced
by the loads and produces a decay of these vibrations over time. Several types of
damping can be included in a dynamic analysis; some of them are the viscous or
Rayleigh damping, hysteresis or solid damping, Coulomb or friction damping, and
numerical damping. Here, the numerical damping is considered, which is further
discussed in section “implicit integration”.
Finally, Eq. (4) is discretized in time and solved by some direct integration
method, which makes Eq. (4) be held for all t and then it is also valid at time t + Δt:
..
MUt + Δt + CU̇t + Δt + KUt + Δt = Ft + Δt ð7Þ

Various numerical techniques have been developed in direct numerical inte-

gration for solving the system of Eq. (7). However, the next subsections describe
the most common methods: explicit and implicit direct time integration methods.
Explicit Integration
Explicit methods only require information on time step t to predict the structural
response at time t + Δt , e.g.:
 .. 
Ut + Δt = f Ut , U̇ t , Ut , Ut − Δt , . . . ð8Þ

These methods are conditionally stable, i.e., the time step size (Δt) has to be
smaller than a critical value Δtcr; otherwise, the solution is not stable, that is, the
solution diverges [31]. With time step Δt, the velocity and acceleration at time step
t are approximated through conventional central differential equation (in the nota-
tion, time t is equivalent to n, e.g. Ut ≡ Un ):

1
U̇ n = ðUn + 1 − Un − 1 Þ → Un + 1 = Un − 1 + 2ΔtU̇ n
2Δt ð9Þ
.. 1
Un = 2 ðUn + 1 − 2Un + Un − 1 Þ
Δt

Substituting Eq. (9) into Eq. (3), and reorganizing it:

   
1 1 2 1 1
M + C U n+1 = R ext
− R int
+ MU n − M − C Un − 1 ð10Þ
Δt 2 2Δt n n
Δt 2 Δt 2 2Δt

Equation (10) is conditionally stable with the condition [31]:

qffiffiffiffiffiffiffiffiffiffiffiffi 
2
Δt ≤ 1 − ξ2 − ξ ð11Þ
ωmax
Optimization of Functionally Graded Materials … 219

with ξ as the rate of damping and ωmax as largest contributing frequency to the
dynamic response. Thus, the critical time step (time discretization) can be expressed
by the so-called Courant–Friedrichs–Lewy (CFL) condition [31]:

Lmesh
Δt cr ≤ ð12Þ
c

where c is the velocity of sound in the material. In practice, the spatial discretization
is obtained as:

λ c pffiffiffiffiffiffiffi
Δx = Δy = with λ = ; c = E ̸ρ ð13Þ
20 f0

Term λ is the wavelength, f0 is the fundamental frequency, E is the Young modulus

and ρ is the density.
Implicit Integration
The most common implicit integration method is the Newmark method, which
computes the structural response at time t + Δt based on the solution at current and
previous times as follows:
 .. .. 
Ut + Δt = f Ut + Δt , U̇ t + Δt , Ut + Δt , Ut , U̇ t , Ut , . . . ð14Þ

Specifically, Newmark relations are (as above, the time t + Δt is equivalent to

n + 1, e.g. Ut + Δt ≡ Un + 1 ):
 .. .. 
U̇ n + 1 = U̇ n + Δt γUn + 1 + ð1 − γ ÞUn
1  .. ..  ð15Þ
Un + 1 = Un + ΔtU̇ n + Δt 2 2βUn + 1 + ð1 − 2βÞUn
2

Manipulating Eqs. (4) and (11), and reorganizing:

 
1 1 ̇ 1 ..
K eff
Un + 1 = Rext +M Un + Un + − 1 Un
n+1
βΔt 2 βΔt 2β
    ð16Þ
γ γ ̇ γ ..
+C Un + − 1 Un + Δt − 1 Un
βΔt β 2β

with:

1 γ
Keff = M+ C+K ð17Þ
βΔt 2 βΔt

where γ and β are constant terms that determine the algorithm characteristics,
such as accuracy, numerical stability and the amount of algorithmic damping.
220 F.J. Ramírez-Gil et al.

When γ > 1 ̸2, Newmark methods display algorithmic damping and, in order to
retain unconditional stability, the following choice of β is appropriate:
 
1 1 1 2
γ≥ , β= γ+ ð18Þ
2 4 2

For γ = 0.5 and β = 0.25, Eq. (15) produces the average acceleration method,
while for γ = 0.5 and β = 1 ̸6, Eq. (15) is the linear acceleration method. Moreover,
an amplitude decay factor δ can be introduced and Eq. (18) is rewritten as:

1 1
γ= + δ, β= ð1 + δÞ2 , δ≥0 ð19Þ
2 4

Although Eqs. (18) and (19) impose a condition in order to get an uncondi-
tionally stable solution, i.e., the convergence is independent of the time integration
step Δt, this value must be appropriately selected for accuracy purposes.
At this point, a choice must be made between these two direct integration
methods: explicit or implicit. The explicit integration methods require many, but
low computational cost, integration sub-steps and are useful for short time impact
problems. Conversely, the implicit methods require less, but computational-intense,
integration sub-steps and are useful for structural dynamic problems. Hence, since a
low-velocity impact with material linear effects is considered herein, the Newmark
implicit integration method is suitable. Finally, since only numerical damping is
considered, the Newmark method presented in Eqs. (16) and (17) becomes:
 
1 1 1 ̇ 1 ..
ext
M + K Un + 1 = Rn + 1 + M Un + Un + − 1 Un ð20Þ
βΔt 2 βΔt 2 βΔt 2β

which will be used for the simulation later in this chapter with K = KðUÞ since
geometric nonlinearities are considered. Consequently, the Newmark integration
scheme may be used in association with the Newton-Raphson iterative algorithm.

4.2 Topology Optimized Structures Under Impact Loads

The TOM has been applied successfully to many fields [12]. However, most
attention has been paid to structural optimization, especially under static loading
condition. As optimization techniques and computers have increased their power,
new major complex problems can be solved by TOM. One example is the opti-
mization of structures under dynamic loading conditions. These kinds of problems
are complex since they require dynamic finite element and sensitivity analysis.
Topology optimized structures under transient loads are of primary interest to
automotive safety, specifically regarding crashworthiness [32].
Optimization of Functionally Graded Materials … 221

In a dynamic problem, such as impacts involving two or more bodies, the loads
are applied over a short period and transient effects must be considered.
Dynamic FEA is required to obtain the numerical dynamic response of the system.
Dynamic FEA commonly requires geometric and/or materials and/or contact non-
linearities. Even though the dynamic FEA does not consider any nonlinearities and
is evaluated as an elastic analysis, large simulation times are needed.
The complexity of dynamics TOM problems is increased due to the dynamics
sensitivity analysis, which is computational expensive and due to nonlinearities, the
gradient calculation is usually a difficult task or maybe impossible for the most
complex problems. Consequently, these difficulties have hindered research in the
structure optimization for impact loading conditions. To overcome these problems,
surrogate models or non-gradient methodologies are often employed [32].
Surrogate methodologies use design variables in the range of 10–50, while the
topology optimization problems typically need from thousands to millions of design
variables, which increase the problem complexity and the computational time
required for finding the solution. Then, fewer methods using topology optimization
for dynamical analysis exist due to the large number of design variables involved.
Surrogate models reduce the cost of expensive analysis methods as well as alleviate
issues with sensitivities. Additionally, they are good alternatives when dealing with
the highly nonlinear and noisy design spaces. Examples of surrogates methods are
RSM (response surface method), Kriging, artificial neural networks, and RBF
(radial basis functions) [32].
TOM based on non-gradients have populated the dynamic problems avoiding the
complexity imposed by the dynamics sensitivity analysis. An often used, but inef-
ficient approach, is to utilize genetic algorithms (GAs) or semi-stochastic techniques.
These methods may be more likely to find global solutions, but they require thou-
sands of function calls. Another non-gradient based methodology developed by Xie
and Stevens is called evolutionary structural optimization (ESO) [33]. It is based on
the concept of progressively removing inefficient material from a structure so that it
evolves into an optimal design. Another approach that requires non-gradient infor-
mation is the hybrid cellular automaton (HCA) method [34].
Although the design of energy absorption structures has been studied for many
years, the implementation of TOM in this kind of problems is still relatively new
[32]. Since most applications under impact events aim at maximizing the energy
absorption by the structure during the collision, other structures that have shown a
good performance as energy absorbers under impact loads are cellular materials.
Cellular materials, including honeycombs and foams, have been widely applied
to various engineering fields and, among these applications, their employment as
structural materials of dampers or energy dissipation devices has been the most
focused application [35]. These materials with periodic cells are typically config-
ured as cores of panels, tubes and shells, which produce a sandwich-type structure.
The properties in these materials that appear most attractive for impact applications
are energy absorption, vibration control and reduced weight. Additionally, cellular
222 F.J. Ramírez-Gil et al.

materials are topology-sensitive; that is, relevant properties are sensitive to the
micro-architecture of the cells [35]. Two kinds of predominant topologies are
distinguished in cellular materials: stochastic and periodic microstructures. The
second configuration offers structures with superior performance than those con-
structed with the analogous stochastic configuration. Cellular materials, particularly
porous materials are closely related to natural structures and the FGM concept (see
Fig. 1).

4.3 Equivalent Static Loads

Dynamic loads are forces that change in the time domain, while static loads are
forces that are constant, regardless of the time, and the inertial effects are not
considered. Accordingly, structures under dynamic loads cannot be represented by
static loads, although at any given time t, Eqs. (3) and (7) can be thought of as a set
..
of “static” equilibrium equations that also take into account inertia forces (M U) and
damping forces (CU̇). Thus, there are several methods to transform dynamic loads
into static loads. One transformation method is the equivalent static load
(ESL) method [36]. An ESL is defined as a static load that produces the same
displacement field as a dynamic load at an arbitrary time t. A static analysis is
expressed as [37]:

KX = S ð21Þ

where X is the static displacement vector and S is the vector of external static loads.
The dynamic displacement vector U(td) at an arbitrary time td can be obtained from
Eq. (3), and substituting X for U(td) in Eq. (21), the ESLs are obtained as follows:

S = K Uðtd Þ ð22Þ

Then, a set of ESLs vectors are obtained, and its number depends on the time
discretization used. This set of ESLs vectors are used as multiple load conditions in
the optimization process [37].

4.4 The Optimization Process with ESLs

For a structural optimization problem subjected to an impact load, the objective

function is to minimize the compliance (or maximize the stiffness) of the structure.
This is a classic objective function in structural optimization, the convexity of
which has been mathematically proved. Additionally, this objective function is
Optimization of Functionally Graded Materials … 223

usually accompanied by a volume constraint. Thus, the formulation for a linear

static optimization problem is:

Min Fo ¼ U T F ¼ UT KU
ρ
V  Vmax ≤ 0 ð23Þ
Such that : ρ  ρmin ≥ 0
With: KU ¼ F

where the structural stiffness is represented as UTF, V is the maximum allowed

volume in the final optimal design, which is commonly a fraction of the design
domain volume Vmax, ρ is the design variables vector and ρmin is a minimum value
used to avoid numerical singularities. Furthermore, the equilibrium equation must
be satisfied. The sensitivities of the objective function Fo with respect to the design
variable ρe are given by:

∂Fo
¼ pρep1 E uTe k0 ue ð24Þ
∂ρe

where ue and k0 are the element displacement vector and local stiffness matrix,
respectively. The k 0 matrix is computed with a Young’s module E = 1 and based on
isotropic material and axisymmetric modeling. Once the optimization problem has
been formulated, their solution is the next phase. The optimization process with
ESL method consists of two parts [37]: the analysis and the design domains. Based
on the results of the analysis domain, ESLs are calculated for the design domain. In
the design domain, static response optimization is conducted with the ESL. The
modified design domain is incorporated to the analysis domain. The entire opti-
mization process iterates between the two domains until the convergence criteria are
satisfied. Figure 5 shows the optimization process using the ESL method where the
algorithm steps are [37]:
• Step 1: initial setting of parameters and design variables (number of iterations
k = 0, design variables ρk = ρ0, convergence parameters: ε a small value)
• Step 2: performs dynamic analysis with ρk (domain analysis)
• Step 3: calculates the ESLs at all-time intervals
• Step 4: solves the optimization problem with linear static response set of the
equivalent load (in the design domain), as shown in Fig. 3.
• Step 5: evaluates the convergence criteria
• Step 6: updates the design variables if the problem does not reach the conver-
gence and repeats the process.
224 F.J. Ramírez-Gil et al.

Fig. 5 Optimization process using equivalent static loads [37]

5 TOM-Based Design of FGMs Under Impact Loads

In this work, the design of FGMs under low-velocity impacts by the TOM is
conducted by using the ESL method [37]. In this section, cylindrical plates with
holes heuristically located are presented first, and next, the behavior of the same
cylindrical plate after an optimization procedure that introduces pores (or holes) in
the material by using the TOM is shown.

5.1 Heuristic Approach

In this section, functionally graded porous structures subjected to low-velocity

impact loads are analyzed. The graded function controls the holes diameter, which
is selected heuristically in this first part. Cylindrical steel plates with 10 cm in
diameter and 3 cm in thickness are used, as shown in Fig. 6. The plate is supported
at the ends; therefore, movements in X and Y are restricted. At the center, an
impulsive pressure load is applied, which simulates the impact of an element of
Optimization of Functionally Graded Materials … 225

Fig. 6 Steel plate geometry used in the simulation. a 3D model and. b Axisymmetric model

0.6 cm of diameter. Taking advantage of the geometry, a planar axisymmetric FEM

model is used, which reduces the solution time as compared to an equivalent model
made in a three-dimensional (3D) domain and produces approximated results.
The gradation in the plate is performed by circular holes,1 where their diameters
and distribution change in the same direction as the stress wave goes through the
plate (thickness direction). Figure 7 shows the functionally graded plates with their
respective graded function used in the simulations. Holes distribution 1 has rela-
tively small holes through the thickness of the plate with a constant function of
gradation (there is no variation in the distribution or diameter size of the holes).
Holes distribution 2 has similar characteristics as those of distribution 1, yet with
larger hole diameters. Holes distribution 3 begins with a larger diameter on the
surface near the impact area, reducing it as the holes approach the opposite surface.
Holes distribution 4 is opposed to distribution 3. Finally, holes distribution 5 begins
with relatively small holes on the surface near the impact area, then increases in size
in the middle and is reduced again on the opposite surface. These distributions are
arbitrarily selected for verifying the role of the holes, their distribution and their
diameters in the structural behavior subjected to impact loads, and their influence in
the stress wave propagation and weight reduction.
The simulations are performed by using the finite element software ANSYS with
the Newmark implicit integration method using interpolation parameters δ = 0.005,
γ = 0.5050 and β = 0.2525, which is an unconditionally stable scheme. Four types of
steel are used: a low carbon steel (AISI 1020), an alloy steel (AISI 4140), a tool steel
(AISI S2), and a ballistic steel (Mars 300). The material properties needed in the
transient simulation are presented in Table 1 for the different steels used. The applied
pressure is 350 MPa in 6.12 × 10−6 s (see Fig. 8), which represents the low-velocity
impact (an impulsive load applied over one integration time step). The mesh size and
the time step for the simulation is calculated by using the Courant condition, obtaining
Δt = 6.12 × 10−6 s and Δx = Δy = 25 × 10−5 m. Additionally, a convergence
test is performed to adjust the size of the mesh. The axisymmetric FE type used by

1
Since we use 2D domains, the holes seem circular; however, the model is asymmetric and thus the
holes are toroidal. We hereafter refer to them simply as holes.
226 F.J. Ramírez-Gil et al.

Plate thickness (mm)

Plate thickness (mm)

Hole diameter (mm) Hole diameter (mm)

Holes distribution 1 (HD1) Holes distribution 2 (HD2)

Plate thickness (mm)

Plate thickness (mm)

Hole diameter (mm) Hole diameter (mm)

Holes distribution 3 (HD) Holes distribution 4 (HD4)

Plate thickness (mm)

Hole diameter (mm)

Holes distribution 5 (HD5)

Fig. 7 Arbitrary hole distributions used in the plate

Table 1 Material properties used in the simulations

Material Density (kg/m3) Young’s modulus (GPa) Poisson’s ratio
AISI 1020 7500 200 0.29
AISI 4140 7850 205 0.29
AISI S2 7750 207 0.29
Mars 300 8100 210 0.32

ANSYS is the so-called Plane182, which is defined by four nodes having two degrees
of freedom at each node: translation in the X and Y directions.
For determining the structural behavior of the arbitrary cylindrical graded plates,
they are compared with a solid plate. By simulation, the stresses are obtained in five
Optimization of Functionally Graded Materials … 227

L.S #: load step number (1,2,3)

P0 = 350 MPa
dt = 6.12 x 10-6 s (time step)
tf = 6 x 10 -3 s (final simulation time)

Fig. 8 Load steps used in the simulations (unscaled)

points on the opposite surface where the impact load is applied, as shown in Fig. 9,
for the plate with holes distribution 1. Graphs from 1 to 5 show the value of the
equivalent von Mises stress versus time. As expected, the impact energy is atten-
uated by the structure over time. The stresses at point 1 are higher than those at
point 5.
Figure 10 shows the stress wave propagation at different times for a graded
plate. It can be seen that the stress wave propagates concentrically from the load
application point to the ends, across the entire plate. So as to know the stress wave
energy that crosses the plate, the von Mises stresses at each discretized point along
the entire opposite side where the load is applied are measured (line A–A in
Fig. 11). Figure 11 shows, for the graded plate with holes distribution 5, the von
Mises stresses at the load application time (6.12 × 10−6 s) and at the end of the
simulation time (6 × 10−3 s), when the load has been attenuated. However, the
total stress energy2 per unit volume is obtained as the area under the curve of the
stresses versus time. Thus, the von Mises equivalent stress for all points along line
A–A are summed and plotted against time as shown in Fig. 12 for the four materials
and for the five gradation configurations considered.
The gradation effect in the energy dissipation through the plate is compared with
the energy going through the solid plate. For this purpose, the area under the curves
in Fig. 12 is computed and a ratio between solid and graded plates are obtained.
Table 2 shows the comparison of the energy dissipated by the solid plate against
each graded plate and the percentage of weight reduction (relative to the solid
plate). It shows that the graded plates present a weight reduction of at least 14 % as
a consequence of the porosity introduced. Additionally, the energy going through
the graded plates increases in all cases as compared with the solid plate. It is
observed that as the material is more resistant to impact, the amount of energy
going through the plate is lower, as expected. The best result is for the Mars 300
with graded distribution 1. The worst result is for graded distribution 2, which has
the largest weight reduction (18.9 %); however, the energy going through the plate
presents the largest values. Accordingly, the results show that the gradation in the

2
The area under the curve indicates strictly power per unit volume, which is directly related to the
elastic deformation energy per unit volume.
228 F.J. Ramírez-Gil et al.

Fig. 9 von Mises stress (left) at points 1–5 of the plate (right) with holes distribution 1 (AISI
1020)

plates does not reduce the amount of energy going through them; depending on the
application, this increase may be relatively less critical as compared with the
advantages provided by the weight reduction. Additionally, the graded functions are
arbitrarily obtained, making it is important to seek for an optimum graded function
obtained by using optimization techniques, which is the subject of the subsequent
section.
Optimization of Functionally Graded Materials … 229

Fig. 10 Propagation of the von Mises stress waves through time (AISI 1020, Holes distribution 5)

Fig. 11 von Mises stresses along the opposite side where the load is applied. a At the time the
impact load is applied and, b at the end of the simulation (AISI 1020, holes distribution 5)

5.2 Optimized Approach

The optimized approach is carried out in ANSYS APDL by adapting its opti-
mization module with the ESL method, which requires four principal steps:
• Problem definition. In this step, the analysis type and the material properties
required by the analysis are defined.
• Active and passive regions definition. The active region corresponds to the
domain to be optimized with the objective function, while the passive region is
that area in which no optimization occurs. This design domain is defined sim-
ilarly to a sandwich-like structure: two solid panels enclosing a porous structure
(see Fig. 13).
• Loads and boundary conditions definition. The optimization problem requires a
FEA analysis at each iteration, which requires the use of loads and boundary
conditions. However, in this study, we use the ESL method to convert the
dynamic loads into equivalent static loads.
• Optimization process definition. Here, the objective function and constraints are
defined.
230 F.J. Ramírez-Gil et al.

Fig. 12 Summation of von Mises stress for all points along the line A–A against time simulation
for different materials. a AISI 1020, b AISI 4140, c AISI S2 and, d Mars 300

Table 2 Solid and graded plates comparison for the four materials
Ratio Weight Energy increment
reduction (%) AISI 1020 (%) AISI 4140 (%) AISI S2 (%) Mars 300 (%)
Solid/gradation 1 14.1 38.8 31.3 30.9 29.2
Solid/gradation 2 18.9 52.8 45.2 53.6 44.9
Solid/gradation 3 17.1 40.2 37.0 43.3 31.5
Solid/gradation 4 17.1 44.2 36.1 43.5 34.7
Solid/gradation 5 14.1 33.2 34.8 33.6 34.0

The material properties used for the optimization process are shown in Table 1.
The design domain is shown in Fig. 13a, which consists of active (the TOM can
modify it) and passive regions (the TOM cannot modify it). The FE type used in the
study is known as Plane82 in ANSYS. The objective function is the maximization
of the stiffness (minimum compliance problem) under low-velocity impact loads.
Since the set of ESLs are obtained from dynamic displacements at a particular time,
there is an ESL per node; however, for simplifying the study, the areas where the
Optimization of Functionally Graded Materials … 231

Fig. 13 a Two-dimensional design domain with active and passive regions, and b loads and
boundary conditions for two cases (OTi & OTii, respectively)

largest displacements are presented are used, as shown in Fig. 13b. Case I
(abbreviated as OTi – optimized topology for case I) obtains the displacements at
the opposite face where the impact load is applied. Case II (abbreviated as OTii -
optimized topology for case II) obtains the displacements in three lines: at the top
(where the impact load is applied), in the middle (where there is an interface
between the passive and the active regions) and at the bottom (similarly to Case I).
Figure 14 shows the optimization process evolution for the AISI 1020 (the worst
material for impact application as demonstrated in the previous section). The blue

Fig. 14 Topology optimization process at different times (AISI 1020, Case I)

232 F.J. Ramírez-Gil et al.

Fig. 15 Topology optimized circular plate with his graded function (AISI 1020, Case I)

color represents voids zones, while the red color indicates zones with base material.
For this optimization process, a constraint in volume is imposed, and the conver-
gence is achieved when this constraint is reached. The volume constraint is
equivalent to a weight constraint since the density is constant. The final topology
obtained by the TOM is given in Fig. 15. The comparison in terms of energy
dissipation between the solid plate, the heuristic graded plate (hole distribution 2—
HD2–) and optimized version are shown in Fig. 16.

Fig. 16 Energy dissipation between solid, heuristic graded and optimized graded plates (AISI
1020, Case I)
Optimization of Functionally Graded Materials … 233

Fig. 17 Topology optimized circular plate with his graded function (AISI 1020, Case II)

However, the optimization results are different if Case II (see Fig. 13b) for the
ESLs are considered as shown in Fig. 17. Figure 18 presents the comparison in
terms of dissipated energy between solid, heuristic graded and topology optimized
plate. Table 3 shows the numerical comparison for the Cases I and II with respect to
solid and heuristic graded plate by using the AISI 1020 steel. It can be seen that
with optimization approach, the graded plates outperform (Case I) or equally

Fig. 18 Energy dissipation between solid, heuristic graded and optimized graded plates (AISI
1020, case II)
234 F.J. Ramírez-Gil et al.

Table 3 Dissipated energy comparison for three topology cases by using AISI 1020
Material ratio Weight reduction (%) Energy increase (%)
Solid/heuristic hole distribution 2 (HD2) 18.8 52.8
Solid/optimized topology case I (OTi) 18.8 22.0
Solid/optimized topology case II (OTii) 18.8 52.8
HD2/OTi 0 −20.2
HD2/OTii 0 −0.013

Fig. 19 Topology optimized circular plate with his graded function for Mars 300. a Case I and,
b Case II of loads application

perform (Case II) when compared against the plate with heuristic holes distribution
2; however, both types of graded plates have a poor performance compared with
solid plate.
The same study as that conducted above is followed; however, changing the
material. In this case, Mars 300 steel is considered. The optimized topologies for
cases I and II are given in Fig. 19. The numerical comparison between solid,
heuristic graded and optimized graded plates are shown in Table 4. Since, in the
non-optimized approach, holes distribution 1 (HD1) gives the best results for Mars
300 steel, this hole distribution is considered here for the comparisons. The opti-
mized topology obtained for Case I improves the energy absorption, indeed, out-
performing the solid and heuristic graded plate (HD1). However, the OTii case only

Table 4 Dissipated energy Material Weight reduction Energy increase

comparison for three topology ratio (%) (%)
cases by using Mars 300
Solid/HD1 14 29.2
Solid/OTi 14 −4.9
Solid/OTii 14 25.4
HD1/OTi 0 −26.4
HD1/OTii 0 −3.0
Optimization of Functionally Graded Materials … 235

Fig. 20 Topology optimized circular plate for AISI 1020 by reducing. a 10 %, b 20 %, c 35 %

and, d 50 % of total weight

outperforms the plate with HD1 with a small percentage of difference; and it
performs poorly against the solid plate.
Finally, an experiment is conducted by varying the total volume to be removed
from the optimized graded plate. The topologies obtained for several percentage of
weight reduction are shown in Fig. 20 by using AISI 1020 steel. It is observed that
the dissipated energy through the plate decrease when the material removed is
increased in a non-linear relation as shown in Table 5.

Table 5 Dissipated energy comparison between solid and optimized plate considering four
weight constraints by using AISI 1020
Material ratio Energy increase (%)
Solid/optimized plate (weight reduction: 10 %) 5.9
Solid/optimized plate (weight reduction: 20 %) 17.6
Solid/optimized plate (weight reduction: 35 %) 113.5
Solid/optimized plate (weight reduction: 50 %) 132.6
236 F.J. Ramírez-Gil et al.

6 Conclusions of the Chapter

This chapter presents the concept of functionally graded materials and topology
optimization methods. Both concepts are combined to produce lightweight struc-
tures for low-velocity impact applications. The form that we use to reduce the weight
in the structures considered (circular plates) is by inserting holes into the material.
The addition of holes to circular plates under low-velocity impacts is evaluated in
two ways, by using heuristic and systematic approaches. However, the introduction
of holes to the plate tends to reduce its capacity for absorbing the impact energy.
Nevertheless, the porous material outperforms the solid plate once: by using the
topology optimization method and a high quality steel for impact applications (Mars
300). Additionally, the optimized approach minimizes the impact energy that crosses
the plate when compared with the heuristic holes distributions. These results are
promising for getting lightweight structures with improvements in their performance
against impact loads.

Acknowledgments The first author acknowledges financial support from COLCIENCIAS by the
scholarship “Becas de Colciencias, Doctorado en Colombia, 567—2012”. The third author
acknowledges the financial support of CNPq (National Council for Research and Development),
under grants 304121/2013-4.

References

1. Zukas, J.A.: High Velocity Impact Dynamics. Wiley, New York (1990)
2. Goicolea, J.M.: Estructuras sometidas a impacto. In: Car, E., López, F., y Oller, S. (eds.)
Estructuras sometidas a acciones dinámicas, pp. 535–567. Centro Internacional de Métodos
Numéricos en Ingeniería (CIMNE), Barcelona (2000)
3. Rosenberg, Z., Dekel, E.: Terminal Ballistics. Springer, Berlin (2012)
4. Anderson, C.E.: A review of computational ceramic armor modeling. In: Prokurat, L.,
Wereszczak, A., Lara-Curzio, E. (eds.) Advances in Ceramic Armor II: Ceramic Engineering
and Science Proceedings, vol. 27, no. 7, pp. 1–18. Wiley, Hoboken, NJ, USA (2008)
5. McCauley, J.W., D’Andrea, G., Cho, K., Burkins, M.S., Dowding, R.J., Gooch, W.A.: Status
Report on SPS TiB2/TiB/Ti Functionally Graded Materials (FGMs) for Armor, U.S. Army
Research Laboratory (2006)
6. Koizumi, M.: FGM activities in Japan. Compos. B Eng. 28(1–2), 1–4 (1997)
7. Markworth, A.J., Ramesh, K.S., Parks, W.P.: Modelling studies applied to functionally graded
materials. J. Mater. Sci. 30(9), 2183–2193 (1995)
8. Huang, J., Fadel, G.M., Blouin, V.Y., Grujicic, M.: Bi-objective optimization design of
functionally gradient materials. Mater. Des. 23(7), 657–666 (2002)
9. Birman, V., Byrd, L.W.: Modeling and analysis of functionally graded materials and
structures. Appl. Mech. Rev. 60(5), 195 (2007)
10. Obata, Y., Noda, N.: Optimum material design for functionally gradient material plate. Arch.
Appl. Mech. 66(8), 581–589 (1996)
11. Cardoso, E.L., Fonseca, J.S.O.: Complexity control in the topology optimization of continuum
structures. J. Braz. Soc. Mech. Sci. Eng. 25(3), 293–301 (2003)
12. Bendsøe, M.P., Sigmund, O.: Topology Optimization: Theory, Methods, and Applications.
Springer, Berlin (2003)
Optimization of Functionally Graded Materials … 237

13. Eschenauer, H.A.: Topology optimization of continuum structures: a review. Appl. Mech.
Rev. 54(4), 331–389 (2001)
14. Michell, A.G.M.: LVIII. The limits of economy of material in frame-structures. Philos. Mag.
Ser. 6 8(47), 589–597 (1904)
15. Bendsøe, M.P., Kikuchi, N.: Generating optimal topologies in structural design using a
homogenization method. Comput. Methods Appl. Mech. Eng. 71(2), 197–224 (1988)
16. Sigmund, O.: Topology optimization: a tool for the tailoring of structures and materials.
Philos. Trans. Roy. Soc. Lond. Ser. A Math. Phys. Eng. Sci. 358(1765), 211–227 (2000)
17. Sigmund, O.: A 99 line topology optimization code written in Matlab. Struct. Multi. Optim.
21(2), 120–127 (2001)
18. Bendsøe, M.P.: Optimal shape design as a material distribution problem. Struct. Optim. 1(4),
193–202 (1989)
19. Bendsøe, M.P., Sigmund, O.: Material interpolation schemes in topology optimization. Arch.
Appl. Mech. 69(9), 635–654 (1999)
20. Rozvany, G.I.N.: A critical review of established methods of structural topology optimization.
Struct. Multi. Optim. 37(3), 217–237 (2008)
21. Dadalau, A., Hafla, A., Verl, A.: A new adaptive penalization scheme for topology
optimization. Prod. Eng. Res. Devel. 3(4–5), 427–434 (2009)
22. Sigmund, O., Petersson, J.: Numerical instabilities in topology optimization: a survey on
procedures dealing with checkerboards, mesh-dependencies and local minima. Struct. Multi.
Optim. 16(1), 68–75 (1998)
23. Hassani, B., Hinton, E.: A review of homogenization and topology optimization III—topology
optimization using optimality criteria. Comput. Struct. 69(6), 739–756 (1998)
24. Tovar, A., Patel, N.M., Niebur, G.L., Sen, M., Renaud, J.E.: Topology optimization using a
hybrid cellular automaton method with local control rules. J. Mech. Des. 128(6), 1205–1216
(2006)
25. Lamberti, L., Pappalettere, C.: Move limits definition in structural optimization with sequential
linear programming. Part I: optimization algorithm. Comput. Struct. 81(4), 197–213 (2003)
26. Zhang, X., Zhang, H.: Optimal design of functionally graded foam material under impact
loading. Int. J. Mech. Sci. 68, 199–211 (2013)
27. Paulino, G.H., Silva, E.C.N.: Design of functionally graded structures using topology
optimization. Mater. Sci. Forum 492, 435–440 (2005)
28. Kim, J.-H., Paulino, G.H.: Isoparametric graded finite elements for nonhomogeneous isotropic
and orthotropic materials. J. Appl. Mech. 69(4), 502–514 (2002)
29. Matsui, K., Terada, K.: Continuous approximation of material distribution for topology
optimization. Int. J. Numer. Meth. Eng. 59(14), 1925–1944 (2004)
30. Vatanabe, S.L., Rubio, W.M., Silva, E.C.N.: Modeling of functionally graded materials. In:
Comprehensive Materials Processing, pp. 261–282. Elsevier (2014)
31. Cook, R.D., Plesha, M.E., Malkus, D.S., Witt, R.J.: Concepts and Applications of Finite
Element Analysis, 4th edn. Wiley, New York (2002)
32. Patel, N.M., Tovar, A., Kang, B.-S., Renaud, J.E.: Crashworthiness design using topology
optimization. J. Mech. Des. 131(6), 061013 (2009)
33. Xie, Y.M., Steven, G.P.: Evolutionary Structural Optimization. Springer, London (1997)
34. Tovar, A., Patel, N.M., Kaushik, A.K., Renaud, J.E.: Optimality conditions of the hybrid
cellular automata for structural optimization. AIAA J. 45(3), 673–683 (2007)
35. Evans, A.G., Hutchinson, J.W., Fleck, N.A., Ashby, M.F., Wadley, H.N.G.: The topological
design of multifunctional cellular metals. Prog. Mater Sci. 46(3–4), 309–327 (2001)
36. Choi, W.S., Park, G.J.: Transformation of dynamic loads into equivalent static loads based on
modal analysis. Int. J. Numer. Meth. Eng. 46(1), 29–43 (1999)
37. Kang, B.S., Choi, W.S., Park, G.J.: Structural optimization under equivalent static loads
transformed from dynamic loads based on displacement. Comput. Struct. 79(2), 145–154
(2001)