OBJECT ORIENTED VOXEL BASED

STOCHASTIC MODELING OF COMPLEX MATERIALS

MANUEL IGNACIO BALAGUERA1 Fundación Universitaria Konrad Lorenz

Abstract
In the present paper, an object oriented stochastic approach is proposed for the construction of
synthetic, computational models of complex materials. The conventional approach to model and
study materials mechanics will be outlined, indicating its limitations to deal with complex
heterogeneous materials. The proposed object oriented integrative modeling will be explained
emphasizing in its advantages compared to continuum mechanics to deal with complex materials.
Finally, the stochastic assembly of complex materials synthetic samples is described and the
architecture of the 3M2S (Multiphysics Materials Modeling and Simulation System) is shown,
indicating further work based on 3M2S.

1. Introduction
The limited availability of natural resources and the high diversity of specific properties and behaviors
imposed to the materials used in high tech industry and health, has motivated an increased interest in
the design and development of smart materials: materials with a high adaptability and able to respond
in a controlled fashion to changes in the physicochemical environment and to external signals.
Those smart materials are complex systems, since they are composed by a high diversity of objects,
some of them with a certain degree of autonomy (cells, nanodevices, nanoparticles). In addition, those
materials are highly heterogeneous and anisotropic in their physical (mechanical, thermal,
electromagnetic, optical) and chemical properties.
As a result, classical mathematical models for those physical properties and behaviors aren't able to
represent such complexity in order to allow the execution of simulation experiments that give the
necessary information to take low uncertainty (and risk) design decisions about smart materials.
In the present paper, an object oriented stochastic approach is proposed for the construction of
synthetic, computational models of complex materials. By using object oriented methods it is possible
to create a digital sample of a complex material that allow to keep record of the individual states and
dynamics for each component, in this case a voxel. In addition, thanks to object oriented methods it
is possible to define autonomous components (cellular automata) which contains individually a set of
"programs" (methods) allowing their adaptation to the surrounding environment and to respond to
external signals.
The assembling of a macroscopic sample of the complex material is made by a stochastic process
choosing at random from a base set of different structure voxels chosen in agreement with probability
distributions built from data obtained using real materials samples.

1
Postal Address: Facultad de Matemáticas e Ingeniería, Fundación Universitaria Konrad Lorenz, Carrera 9 Bis
No. 62 – 43 Bogotá DC, COLOMBIA.
In the present paper the basic principles and components of the proposed methodology will be
explained and in a next publication, the results of validated simulations experiments will be
accounted.

2. Review of the analytical approaches, their background and limitations

Due to the unlimited diversity of materials, physics classifies them in a small set of abstract material
bodies categories: particles, rigid solids, deformable solids, fluids, viscoelastic bodies and some
additional specialized categories.
Actually, this is an artificial classification because depending upon the spatial and temporary scales
of observation and on the range of applied forces, any body may be classified in any category.
When a given set of forces or a force field is applied to an elastic body, there is a “volumetric
distribution of force” in the body which is characterized by a stress field and a strain field. Those
fields together represent the complete mechanical state of the body.
From a general perspective, stress and strain are mathematically represented by one tensor
respectively. A tensor is a 3x3 matrix, where each row is a vector that quantify the tensor in a specific
address, generally normal to a symmetry plane of the body.
In order to illustrate what tensors are and how to represent them, let us consider what occurs with a
hollow cylinder with a thick wall.

Figure 2.1 Axis for stress tensor in a hollow cylinder with a thick wall

As shown in figure 2.1, the cylinder has three symmetry directions: radial, circular and axial. Each
direction corresponds to one point cylindrical coordinate: r,  and z, respectively and to a deformation
mode: volumetric (inflation): change in “r”, torsion: change in “” and tension (traction, axial) change
in “z”.
 Figure 2.2 Stress tensor components

As illustrated by figure 2.2, the stress tensor is the matrix composition of three stress vectors, each
one corresponding to a plane of symmetry:

̂ ̂ where ∙ ̂, ∙ , ∙ ̂ (2.1.a)

̂ ̂ where ∙ ̂, ∙ , ∙ ̂ (2.1.b)

̂ ̂ where ∙ ̂, ∙ , ∙ ̂ (2.1.c)

Where holds for normal stresses and the elements are used to represent shear stress.

Finally, the complete stress tensor is:

(2.2)

In the conventional approach (continuum mechanics and its numerical extension, the finite element
method) the mechanical behavior of materials is "governed" by Hooke's tensor law:

(2.3)

Where is the Young's (elastic) modulus tensor for the material and is the strain tensor.

In continuum mechanics, the hardest case is for orthotropic2 materials whose physical properties
parameters have spatial change in agreement with rules expressed by tensor functions of position
(tensor fields). That restrictions or work hypothesis fit very well for crystalline materials, even for
glasses where the uniformity of randomness guarantees ergodicity3 and, as a consequence, the use of
statistical physics.
For a complex material sample, the fact that some of its components may be autonomous objects
inhibit the possibility to express its states probabilities only as dependent upon its energy and, in this

2
An orthotropic material allows change in physical properties depending upon each of three orthogonal axis.
3
The probability associated with a given ergodic system state depends only of the system's energy (the value
of its Hamiltonian function) and its averages on time coincide with averages over its possible configurations.
way there is breaking of ergodicity and as a consequence the inadequacy of statistical mechanics
principles (mainly the principle of equipartition) in the development of models for such a complex
material. In the next section the main aspects of Object oriented integrative modeling of Complex
Systems will be explained and justified.

3. Object oriented integrative modeling of a complex material sample

Thanks to the main features of object orientation: classification, encapsulation, inheritance and
polymorphism, it is possible to automatically create and assemble a huge number of component
objects coming from a diversity of classes whose definition may encapsulate specific and particular
properties and sets of behavior rules.
In object oriented integrative modeling the starting point is the analysis of the system to be modeled
beginning with a hierarchical decomposition into different levels of structural categories, represented
by the concept of "class" in the object oriented approximation.

Figure 3.1.a

Figure 3.1.b
Both, figures 3.1.a and 3.1.b are UML (Unified Modeling Language) Class Diagrams. A class
diagram displays the different categories of objects, classes, composing a complex system and its
relations, in this case, composition relations:
In 3.1.a, the class diagram corresponds to the planned structure for 3M2S (Multiphysics Materials
Modeling Simulation System), in development by the author, that will allow to realize simulation
experiments about the multiphysics response (mechanical, thermal, fluid diffusion and chemical
reactions) of a complex material.
In agreement with the 3.1.a class diagram, in order to create the class experiment, it will be necessary
to create three composing objects, each one belonging to one of the three classes: physical system,
simulation and interface. Notice that the double tipped arrows indicate an array of components.
The physical system will be the composition of a solid structure (porous), a fluid phase (liquid or gas)
and a thermal field, defined at each point (node) of the solid structure and at each point of the fluid
phase.
The simulation component of the experiment will be composed by a simulation scenario which
defines the external agents acting on the material sample: arrays of mechanical loads, fluid and heat
sources. In addition, the simulation class include as components all the boundary and initial conditions
for the experiment.
The simulation class is also equipped with a set of mathematical engines necessary to process the
time evolution of each component and the complete system: a Monte Carlo (Metropolis) engine used
to simulate all stochastic processes involved in the experiment, a finite element method engine used
to simulate the mechanical behavior of the solid phase and, if necessary, the heat diffusion. Depending
upon the type of fluid phase, the behavior of each fluid phase will be simulated by using the most
convenient fluid dynamics engine chosen between a Lattice Boltzmann engine and a Molecular
Dynamics engine.
Finally, regarding the 3.1.a experiment class diagram, it is necessary to include a user interface
composed by an input interface for information source specification and a simulation output interface
which may be a set of 2D xy like plots or a 3D structure and fields visualizator.
The figure 3.1.b presents the class diagram representing the hierarchical structure of the material
sample model which is visually represented in figure 3.2. In agreement with the diagram, the objects
composing a material sample, from bottom level to top level belong to: 1. Node class, 2. Edge class,
3. Face class, 4. Voxel class and 5. Solid class. Those classes are hierarchically arranged as linked
lists: an edge is an one dimensional object defined by a list of two nodes, a face is an array of edges
forming a 2D polygonal region (generally quadrilateral or triangular), a voxel is a closed 3D region
limited by e polyhedron, it is useful to use quadrilateral prisms or tetrahedrons, and, finally the solid
is an array of voxels.
This hierarchical composition allows us: 1. To individually handle and keep record of the state of
those diverse components in agreement with physical principles, 2. Associate different structure
components with respective physical objects: the solid is a structure formed just by edges which gives
account of the mechanical behavior of the solid, fluids and vacuums occupy and diffuse across voxels,
and thermal fields are temperatures associated with each voxel. 3. Detect special events as face
collisions and associate them with fracture. 4. Build up different field visualizations.
 Figure 3.1 Hierarchical structure to voxelize a complex material sample

as an example, Figure 3.1 illustrates a hierarchical observational decomposition of a cylindrical

sample used to implement a stochastic model allowing to keep track of the state and evolution of all
components in each hierarchical level. As generally the experimental samples have a regular shape
(a cylinder or a rectangular prism) the sample assembling process may be organized by hypothetic
substructures: here, for easy navigability, the lists indexes are arranged as: "i" for slices, "j" for rings,
"k" for (circular) sectors, as to allow the use the algorithm shown in code 1:
Input sample height and radius
Define and calculate step size for slices
for i=0 to nSlices
Define and calculate step size for rings
for j=0 to nRings
Define and calculate step size for sectors
for k=0 to nSectors
Create node list //*each node is an object
Create edge list //* each edge is an object
Create face list //* each face is an object
Create voxel list //* each voxel is an object
Code 1: pseudocode algorithm for sample model assembly

In the creation of the different objects composing the model a stochastic association between
mechanical, thermal, fluid and chemical properties with model objects can be established:
mechanical with edge object list, thermal, fluid and chemical with faces ad/or voxels.
 4. Conclusions and Further work
1. In the present work, object oriented methods for modeling complex materials has been presented
and compared with continuum mechanics, remarking its advantages.
2. The hierarchical decomposition of the material sample has been presented and explained by using
class diagrams, which facilitates code implementation and maintenance as allow to keep
individual record and control of each component object.
3. It has been presented an algorithm for the sample assembly in the case of regular solid samples,
a cylinder here. This algorithm together with appropriate probability distributions for material
phases and components and a dataset coming from real samples characterization will allow to
reproduce the high space heterogeneity of a complex material.
As a second phase of the project, the 3M2S (Multiphysics Material Modeling and Simulation System)
will be implemented and validated by comparing simulation experiments results with corresponding
experimental data.

Acknowledgement
The author would like to express his gratitude with Fundación Universitaria Konrad Lorenz for all
the given support and to Professor Carlos Díez for reviewing the project and this paper and giving his
valuable and helpful comments.

References

********** IN CONSTRUCTION ********