1.

0 Introduction
An integrated CAD 1.1 Design of heterogeneous objects
system for design of Recent developments in the field of structural
optimization have lead to techniques in which
heterogeneous objects geometry and material distribution are both
optimized simultaneously (Bendsoe and
Srinivas Bhashyam Kikuchi, 1988; Bendsoe et al., 1993;
Ki Hoon Shin and Cherkaev, 1994; Cherakaev and Kohn,
1997). These techniques result in
Debashish Dutta heterogeneous objects in which there is a
continuous material variation along with the
geometry. Such optimization techniques are
automated and output the final (optimal)
geometry and material distribution for a given
set of loading, domain specification and
boundary conditions. However, there are
The authors situations which require the designer to create
Srinivas Bhashyam, Ki Hoon Shin and Debashish heterogeneous objects. For example, a
Dutta are based in the Department of Mechanical number of emerging applications in various
Engineering at the University of Michigan, Ann Arbor, fields of engineering require materials to
Michigan, USA.
satisfy multiple objectives (Markworth and
Saunders, 1995; Tanaka et al., 1996; Kwon,
1994; Suh, 1999). Rather than abruptly
Keywords joining the different materials, each of which
CAD, Layered manufacturing satisfies a specific objective, it is advantageous
from the standpoint of maximizing structural
Abstract performance gradually to vary the material
composition from one material to the other in
Computer aided design systems are routinely used by
a controlled manner (Holt et al., 1993). The
designers for creating part geometries. Interfaces to
designer has the flexibility to tailor the
computer aided analysis and manufacturing are also material distribution to achieve a desired set
commonplace enabling the rapid fabrication of the of objectives. This design scenario is
designed part. Thus far, however, the focus was on elucidated with the following example.
objects with homogeneous interior. Two recent advances The specific example under consideration is
± use of functionally graded materials in parts, and a heterogeneous pressure vessel, cross section
layered manufacturing technology ± have brought to the shown in Figure 1 (Kwon, 1994). The
forefront the need for CAD systems to support the pressure vessel is subjected to a high
creation of geometry as well as the graded material temperature/pressure on the inside while the
inside. This paper reports on such a system. We describe outer surface is subjected to ambient
the need, the components and implementation of a CAD conditions. It is desirable to have ceramic on
system for creating heterogeneous objects. Two examples the inner surface of the pressure vessel due to
illustrate the use. its good high temperature properties while it
is also desirable to have metal away from the
Electronic access inner surface owing to its good mechanical
properties. Joining the two materials abruptly
The research register for this journal is available at
will lead to high stresses at the interface. The
[Link] composition of the metal is therefore
The current issue and full text archive of this journal is gradually increased in a controlled manner
available at starting with zero at the inner surface and
[Link] gradually increasing to unity on the outer

The authors gratefully acknowledge the financial

support received from ONR N00014-97-1-0245,
NSF grant MIP-9714951 and DARPA/NIST
grant 7NANB8H0068.
Rapid Prototyping Journal Received: 14 June 1999
Volume 6 . Number 2 . 2000 . pp. 119±135 Reviewed: 17 September 1999
# MCB University Press . ISSN 1355-2546 Accepted: 9 November 1999
119
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 1 Heterogeneous pressure vessel create the physical part in a very short
time(e.g. hours instead of days).
An attractive feature of LM, that is just
beginning to be harnessed, is the selective
deposition of material (on a layer) and the
changing of material between layers. Different
materials, in varying proportions, can be
deposited on a single layer to create graded
interiors (Mazumder et al., 1997; Merz et al.,
1994). Furthermore, it enables the direct
fabrication of multi-material structures since
embedding of prefabricated (electronic or
other) components is possible (e.g. a die of
H13 tool steel with interior copper cooling
surface. The designer has to create a material channels). While commercial LM systems of
composition function for the metal variation today are used for rapid prototyping
and further optimize it such that the radial applications, the fabrication of heterogeneous
and tangential stresses are minimized. objects represents the true potential of layered
To accomplish this task, the designer needs manufacturing.
to be provided with the necessary tools for
creating the heterogeneous object with an 1.3 Need for a heterogeneous CAD
initial ``guess'' for the material composition and system
geometry, representing it on a computer, CAD systems exist for designing geometry in
visualizing the heterogeneous object created, the form of constructing solid models from
performing analysis and modifying the material primitives, visualizing the object on a
composition (and geometry) and repeating the computer and discretization of the models for
cycle until an optimal design is obtained. While finite element analysis (FEA). Generative
such tools exist for geometry, to the best of the techniques, such as optimal design techniques
author's knowledge no such tools are available discussed in section 2.1, are beginning to
for iterative design or synthesis of material emerge that automate the design of material
composition. This paper describes the composition and geometry. Here the designer
development of an integrated Computer Aided plays a passive role in the design process. In
Design (CAD) system which enables the variant design, discussed in section 2.2,
design of geometry and material composition. heterogeneous objects (similar to the pressure
vessel example discussed earlier) are created
1.2 Fabrication of heterogeneous objects step-by-step by the designer. Such a
Layered manufacturing is a new method of heterogeneous CAD system would include
fabrication that has gained rapid acceptance tools for creating and modifying geometry and
in the last decade. In layered manufacturing material composition, computer
(LM), a part is built by depositing material, representation and visualization of the
layer-by-layer, under computer control. geometry and material variation, estimating
While, in conventional manufacturing (e.g. material properties required for analysis and
CNC machining), a part is fabricated by discretization of the heterogeneous object for
selectively removing material from a FEA. To the best of our knowledge such
workpiece, layered manufacturing is a 2.5D CAD systems do not yet exist. In this paper, a
process-stacking up layers. prototype heterogeneous CAD system is
In industry, layered manufacturing is often described that allows the designer to create
referred to as ``rapid prototyping'' reflecting and manipulate heterogeneous objects.
the most common use. LM is widely used for
the rapid fabrication of physical prototypes of
functional parts (important in the design 2.0 Paradigms for synthesis of
stage), patterns for molds, medical prototypes heterogeneous objects
(implants, bones), consumer products, etc.
The unique feature of LM is ``direct'' There are two paradigms for synthesis of
fabrication ± it does not involve tooling, heterogeneous objects, namely the generative
fixturing and other peripheral activities of approach and the variant approach. Each of
conventional manufacturing. Therefore, it is these approaches will be described briefly in
possible to start from a CAD model and the following sections.
120
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

2.1 Generative approach stresses and shielding of heat flow. In many

A flowchart illustrating the concept of cases the objectives that are required to be
generative approach is shown in Figure 2. In fulfilled are conflicting from a physical point
this approach, heterogeneous objects are of view. An example of a generative method
obtained as a result of optimal design for FGM is given in (Tanaka et al., 1996).
techniques. An example of such a design The specific application under consideration
technique is the homogenization design is a thermal barrier FGM. Rather than
method (HDM) (Bendsoe and Kikuchi, assuming a form for the material composition
1988) which outputs the optimal material function, it is obtained by solving the
density distribution for given loading and optimization problem. The objective is to
boundary conditions in a specified domain. A minimize the circumferential stresses with
reverse engineering paradigm is followed to respect to a given reference stress distribution
convert the outputs of such optimal design subject to the constraints that these stresses
techniques to heterogeneous solid models must lie within a desired band.
(Marsan and Dutta, 1996) which can then be Note, the generative approach limits the
used for analysis or layered manufacturing. role of the designer in the design process.
The optimal design system will output both Further, tools for optimal design are
geometry and material distribution limited and expensive, and not easily
information. The reverse engineering task is available.
now to fit a suitable material composition
function, F, to the available material 2.2 Variant approach
distribution data. Fitting surfaces to scattered In the variant approach, the designer's
data is a topic of current interest and several creativity, physical intuition and experience
algorithms exist. However, the focus has been can be effectively exploited. The role of the
on recreating the outer geometry. Only, designer is enhanced from merely interpreting
capturing the material density information the results of an automated system to actually
inside is a difficult problem, one that has not designing the geometry and material
received sufficient attention thus far. composition functions of the heterogeneous
Generative approaches can also be found in object, performing analysis, interpreting
synthesis of functionally gradient materials results, changing the design if the results are
(FGMs) (Markworth and Saunders, 1995; not satisfactory and going through several
Tanaka et al., 1996; Kwon, 1994). In FGMs, iterations of this design-analysis cycle until an
the (continuously varying) material optimal design is obtained. The design
composition is also designed along with process flowchart illustrating the variant
geometry, tailoring the physical properties for approach is shown in Figure 3.
a specific application or function. Tailoring in However, there is a lack of synthesis tools to
FGMs is usually carried out under multiple aid the designer in such an iterative design of
objectives such as reduction of thermal heterogeneous objects. The basic components

Figure 2 Design process flowchart for generative approach

121
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 3 Design process flowchart for variant approach addition to geometry. (Jackson et al., 1998)
have proposed a modeling scheme which
captures the object's geometry, topology and
material composition. Their approach is
based on subdividing the solid model into
sub-regions (finite elements) and associating
analytic composition blending functions
within each finite-element through the use of
barycentric Bernstein polynomials. Our
approach is described in (Kumar and Dutta,
1998; Kumar, 1998) and forms the basis of
our ongoing work in the design and
processing of heterogeneous objects for layer
manufacturing. We provide an overview of
the rm-object framework next.
3.1.1 Rm -object framework
The rm-object framework utilizes r-sets
(Requicha, 1980; Hoffmann, 1989), as the
basis of representing the geometry of
heterogeneous objects. Heterogeneous
objects are composed of different constituent
materials (called primary materials). At any
given point in the heterogeneous object, the
material is (in general) a combination of the
of a variant system for the design of
primary materials and is specified by the
heterogeneous objects include:
volume fraction of each of the primary
. heterogeneous solid modeling technique;
materials. The volume fraction of each of the
. library of material composition functions;
(n) primary materials can vary continuously.
. library of property estimation rules;
Therefore the material space will be
. a graphical user interface (GUI).
represented by Rn, with each dimension
In addition, a database of primary materials representing volume fractions of a primary
along with their material properties would material. The physical material space will be
serve as an auxiliary component. The GUI restricted due to the fact that the sum of the
would allow the designer to access these volume fractions of the primary materials
databases efficiently. Next, we describe each should be unity. Thus the physical material
component in detail as implemented in our space V can be defined as:
( )
prototype system. n
X
n
V ˆ  2 R j kk1  i ˆ 1 and i  0 …1†
iˆ1

3.0 Components of heterogeneous CAD where k
k1 denotes the L1-norm and vi is the
system volume fraction of the ith primary material. A
material composition function, F, can be
3.1 Heterogeneous solid modeler defined that maps a geometrical point in a
A heterogeneous solid modeler using heterogeneous object to a material point .
heterogeneous objects which can be The mapping is a one to one mapping from
represented and manipulated is at the core of E3 to Rn and is given by:
our integrated heterogeneous CAD system.
F : E 3 ! V j kF…x 2 E 3 †k1 ˆ 1
The need for going beyond geometry in solid
modelers has been recognized (Hoffmann and where F…x†  v…x† ˆ fvi …x†g …2†
Rossignac, 1996). New heterogeneous
modeling schemes have been recently Based on the above framework for modeling
proposed (Jackson et al., 1998; Kumar and geometry and material, a structure called
Dutta, 1998; Kumar, 1998) which enhance rm-set can be defined which captures both
the capability of existing solid modeling geometry and material information in a
techniques to capture variations in material in heterogeneous object. An rm-set is defined as
122
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

a subset (P, B) of T where P  E3 is an r-set functions have been identified from literature.
and B  V assigns material to the r-set P. An These composition functions along with
equivalent definition of rm-set is (P, F) where primary material combinations and intended
F(P) is the material function that provides a applications are shown in Table I (not
mapping between geometry and material. exhaustive). The designer can choose a
The material function F is required to be composition function from this database. For
continuous in the r-set P. Abrupt variations in this particular choice of composition
material composition, if present, are explicitly function, physical properties can be estimated
modeled through a collection of rm-sets called from the library of property estimation
rm-objects. An rm-object which is a finite models which forms the next component of
collection of rm-sets is defined as: synthesis tools. Next, analysis can be
S ˆ fPj 2 A; Bj  V g; where j is finite …3† performed using numerical analysis tools such
as finite element method (FEM) and the
The material function, F, in the rm-set is results can be evaluated.
defined over the entire E3, but is only valid If the results are not satisfactory or could be
within its corresponding geometry, P. In other optimized further, the coefficients in the
words, the geometry P in an rm-set provides a composition function can be altered and the
bound for validity of the associated material new design can be analyzed. Several iterations
composition function, F. The material of this design-analysis cycle can be repeated
composition function itself is defined in an until an optimal solution is obtained (e.g.
unbounded domain. To numerically evaluate stress levels are less than the allowable limit).
the material composition belonging to a Applications which use this approach for
rm-set at a point x, a point membership design of thermal barrier type FGMs are
classification (PMC) has to be done to available in the literature (Markworth and
determine if x lies within the geometry of the Saunders, 1995; Suh, 1999; Suresh and
rm-set. Only if the point lies within the Mortensen, 1997; Drake et al., 1993). For
geometry, the material function is evaluated. example, Markworth and Saunders (1995)
The rm-object is defined as the have considered the design of a metal/ceramic
mathematical model for representing FGM in a one-dimensional domain. One end
heterogeneous objects and we refer to this (ceramic) is exposed to a high temperature
model as a heterogeneous solid model and the other end (metal) to a cooler
(HSM). To manipulate heterogeneous temperature. The material composition in
objects, modeling operations on rm-objects between is optimized subject to criteria that
can be defined. Modeling operations include the thermal flux across the material is
regularized Boolean operations such as minimized, thermal stresses are minimized
regularized difference, regularized and are restricted below the yield stress of the
intersection and regularized union. For details composite material. They assume a parabolic
on Boolean operations on rm-sets and distribution for the volume fraction of the
rm-objects see (Kumar and Dutta, 1998; metal and optimize the coefficients of the
Kumar, 1998). These operations can be used parabolic function.
to create complex heterogeneous objects from Once the material composition function has
simple ``heterogeneous primitives'' akin to the been designed, physical properties need to be
manner in which complex objects are estimated for purposes of analysis. A library of
generated from simple primitive shapes in property estimation rules which are
constructive solid geometry (CSG). For microstructure dependent are necessary for
heterogeneous objects, a primitive geometry estimating properties.
with a material composition function defined
with respect to its local coordinate system 3.3 Library of property estimation rules
(i.e. an rm-set) will be the heterogeneous Analysis of heterogeneous objects calls for
primitive. methods by which effective properties of the
underlying multiphase structures can be
3.2 Library of material composition determined. This is a problem that falls in the
functions domain of micromechanics and data
The library of material composition functions characterizing the microstructure such as size
is a key component of the system. Towards and shape of the inclusion phase in the
this end, several material composition matrix, the orientation and spatial
123
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Table I Material composition functions from literature

Composition function(s)/
Property variations Application Primary materials
2
VA ˆ a0 ‡ a1 x ‡ a2 x Minimize thermomechanical stress, Material A: Metal
minimize heat flux in a thermal barrier Material B: Ceramic
FGM (Markworth and Saunders,1995)

xp
VA ˆ Minimization of residual stresses in Ni- Material A: Nickel (Ni)
t
Al2O3 FGM cylindrical specimens Material B: Alumina (Al2O3)
subjected to cooling from a fabrication
temperature (Drake et al., 1993;
Williamson et al., 1995)

    
1  2y
VA ˆ sin 1 ‡1 Reduction of thermal residual stresses Material A: Metal
2 2 h
near free edges of an interface between Material B: Ceramic
( 3 ) two different materials
1 2y
VA ˆ 1 ‡1
2 h

( 5 )
1 2y
VA ˆ 1 ‡1
2 h

r r0
VB ˆ Minimization of thermal stresses in Material A: Metal matrix
r0 ri
axis-symmetric bodies (such as composite (MMC)
 2
r ri cylinders) subjected to radial Material B: SiC particulates
VB ˆ
r0 ri temperature gradients and pressure
differentials (Kwon, 1994)
r2 ‡ 2r0 r ‡ …r2 2r0 r1 †
VB ˆ
…r0 rr †

r r0 2ri
VB ˆ
r0 ri

 x n
VB ˆ 1 Reduction of transient stresses in Material A: Stainless Steel
W
nuclear reactor liners subjected to high Material B: partially
heat flux environment (Holt et al., stabilized zirconia (PSZ)
1993)

 n
x x0
VB ˆ Thermal stress reduction in thermal Material A: Metal
x1 x0
barrier applications (Hirano and Material B: Ceramic
Wakashima, 1995)

B ˆ A e…jyj†
Improving bond strength and reducing No specific materials
interfacial stresses between bonded
… : shear modulus†
dissimilar materials in presence of
interfacial cracks (Holt et al., 1993)

ax
VB ˆ Thermal barrier coatings in gas turbine Material A: NiCrAlY
t
engines (Holt et al., 1993) Material B: ZrO2±Y2O3

124
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

distribution of the inclusions in the matrix are Table II Range of applicability of micromechanical
necessary (Suresh and Mortensen, 1997; models
Hirano and Wakashima, 1995; Weng et al., Volume fraction, VB Micromechanical models
1990; Nemat-Nasser and Hori, 1993).
VB < 0.1 Dilute approximation
Effective material properties include
VB > 0.9 method
mechanical properties: bulk and shear
modulus; thermal properties: coefficient of 0.1  VB  0.3 Kerner's equation, Mori-
thermal expansion and coefficient of thermal 0.7VB  0.9 Tanaka method, method of
conductivity; electrical properties: electric Tamura et al., self-
conductivity and dielectric constant; and consistent method, modulus
chemical properties such as diffusion constant based rules, Voigt and
(Kwon, 1994). Reuss models
In this paper, the scope is restricted to
determination of thermal and mechanical 0.3  VB  0.7 Fuzzy logic method
properties. A variety of microstructure based
property estimation models for estimation of
thermomechanical properties have been 4.0 Prototype heterogeneous CAD
proposed (Kwon, 1994; Suresh and system and its use
Mortensen, 1997; Hirano and Wakashima,
A prototype heterogeneous CAD system
1995; Markworth et al., 1995; Zuiker, 1995).
incorporating synthesis tools and the
These models (or methods) include rule of
heterogeneous solid modeler for design of
mixtures (Voigt-model) and inverse rule of
heterogeneous objects has been implemented.
mixtures (Reuss-model), Kerner's method,
A schematic representation of our
Mori-Tanaka method, method of Tamura et
architecture is shown in Figure 4. The
al., the self-consistent method, modulus based
synthesis process is enabled by the GUI which
rules (for coefficient of thermal expansion) and
permits the interaction of the user with the
dilute approximations method. A database of
heterogeneous CAD system. The
property estimation is generated consisting of
heterogeneous solid modeler using the rm-
all of the above available property estimation
object framework is implemented in C++
rules from literature.
For a summary of all these models and their using ACIS kernel (Spatial Technologies,
corresponding equations see Suresh and Inc) and the GUI is implemented using Motif
Mortensen (1997) and Zuiker (1995). All of and OpenGL libraries. The geometry and
the above models can be used when the material distribution functions are input by
microstructure is primarily an inclusion phase the user to the heterogeneous solid model. An
dispersed in a matrix phase. object model (OM) (Kumar et al., 1998) is
For intermediate volume fractions, where created from the heterogeneous solid model
the microstructure is an interconnected after appropriate pre-processing. This object
network of the individual phases, fuzzy logic model is the input for finite element analysis
techniques are used to interpolate properties (FEA).
(Tanaka et al., 1996; Hirano and Wakashima, The automatic conversion of an object
1995; Zuiker, 1995). The volume fraction model to a finite element input through
ranges over which different models are preprocessing algorithms is beyond the scope
applicable is summarized in Table II. of this work and in our system is performed
manually (semi-automatic).
3.4 Graphical user interface (GUI) In order to design a heterogeneous object,
The GUI is necessary for a high level the user performs the following steps :
interaction of the user with the heterogeneous (1) Select primitive geometry (e.g. box,
CAD system. The GUI for our system was cylinder, sphere, cone, torus etc.).
implemented using Motif and OpenGL (2) Select the primary materials for the
libraries. The details of the GUI are discussed primitive.
with respect to specific examples in section (3) Input material composition function
5.0. from the database for the primitive.
125
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 4 An integrated heterogeneous CAD system

(4) Represent the geometry and material 5.0 Synthesis examples

composition function as a heterogeneous
solid model. 5.1 Example 1
(5) Repeat steps 1 through 4 as necessary The first synthesis example is the
to construct all heterogeneous heterogeneous pressure vessel discussed in
primitives. section 1.1. The specific problem under
(6) Using Boolean operations, combine the consideration is a hollow cylinder which
heterogeneous object primitives (rm - contains a high temperature/pressure fluid on
sets) created to form an rm-object. the inside (Figure 1). The cylinder is thus
(7) Convert the heterogeneous solid model subjected to an internal pressure Pi
corresponding to the fluid pressure head and
to an object model using appropriate
a temperature Ti corresponding to the fluid
property estimation rule(s) from the
temperature. The outer surface is
database.
mechanically unloaded and is subjected to a
(8) Convert the object model to a finite
temperature To. The temperature gradient
element input file using appropriate pre-
coupled with the internal pressure leads to
processing algorithms (manual).
thermomechanical stresses developing in the
(9) Perform FE analysis and evaluate
cylinder. The geometry, boundary conditions
results.
and loading are all axisymmetric and hence all
(10) If results are not satisfactory modify
quantities vary only in the radial direction.
composition function variables and go to
The material on the inner surface is ceramic
step 8.
(Al2O3) which has good high temperature
Note, in step 10, a new composition function properties and the material on the outside is
can also be chosen if the results are not metal (Al alloy). The material composition in
satisfactory. Alternatively, as described, new between is varied to minimize circumferential
values for design variables can be used. stresses.
Guidelines need to be developed for the The synthesis process of the pressure vessel
modification of composition functions; in Figure 1 using the GUI is illustrated in
this is a topic we plan to investigate in the Figures 5-8 with respect to the above
future. example. The GUI consists of several levels of
The synthesis procedure utilizing the menu buttons. The first level contains File ,
heterogeneous CAD system for designing Create Rm set, Manage Rm set and Rm object
heterogeneous objects will be illustrated with buttons. Clicking on the File opens up the
examples in the next section. next level of menu by which .sat(ACIS
126
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 5 Create geometry

Figure 6 Create material information

format) and .rmo(newly defined file format clicking on geometry and choosing hollow
which consists of geometry + material cylinder from the pull down menu options for
information) can be retrieved or saved. geometry. The hollow cylinder is further
Clicking on Create Rm set button opens up a instantiated with dimensions.
pop-up menu consisting of options for The next step is to choose material
choosing geometry, material composition composition function and primary materials
function and property estimation rules. The from the library (Figure 6). This is
user can then choose geometry (primitive), accomplished by once again clicking on Create
material composition function and property Rm set in the main menu and choosing
estimation rule for the Rm set by clicking on Material Comp from the pull down menu
the corresponding options. options for Create Rm set. A new window
The first step in synthesizing the pressure opens up which contains options for choosing
vessel is to choose or create geometry (Figure primary materials, defining material
5). This is done by first clicking on Create Rm composition functions and their associated
set in the main menu and subsequently coordinate systems. The primary materials
127
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 7 Visualization of material distribution

Figure 8 Apply property estimation rules in this equation can be fixed based on the
conditions that Vm = 0 on the inner surface r
= Ri and Vm = 1 on the outer surface r = Ro.
The above equation can then be expressed in
terms of a single unknown coefficient a1 as :
 
a1 R0 Ri …Ri R0 † R2i
Vm …r† ˆ
R20 R2i
 
1 a1 …R0 Ri † 2
‡a1 r ‡ r …5†
R20 R2i

The second constant, a1 is allowed to vary

freely and will be optimized such that
are chosen as 2 in this case and entered in the circumferential stresses are minimized. The
box next to No. of primary materials. The coefficient a1 is chosen as 0.0 as an initial
primary materials are chosen as aluminum guess and the coefficients a0 and a2 which are
(Al) for mat 1 and ceramic (Al203) for mat 2 dependent on a1 are calculated in accordance
from the options for primary materials which with equation 5. The mathematical
show up in the pull down menu when the expression corresponding to equation 5 is
buttons next to mat 1 and mat 2 are clicked. A entered in the box corresponding to mat 1
parabolic polynomial defined in a cylindrical (Al).
local coordinate system (fixed to the center of After the analysis has been performed and
the cylinder) is chosen from Table I, row-1 as the results are evaluated, the coefficient (a1) is
the material composition function. changed to obtain a new material distribution.
Vm …r† ˆ a0 ‡ a1 r ‡ a2 r 2 …4† This iterative process is continued until an
optimal material composition function is
Here, Vm is the volume fraction of the metal, r obtained. Whenever coefficients are updated,
is the radial coordinate and a0, a1 and a2 are the user can visualize the material distribution
unknown coefficients. Two of the coefficients by clicking on the material distribution button
128
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

as shown in Figure 7. The Manage Rm set analysis. Symmetric boundary conditions are
menu can be used to modify attributes imposed on the truncated ends of the finite
(geometry, material, etc.) of an Rm set already element model. The finite element mesh is
created. shown in Figure 9. The domain is modeled
The next step is to choose property using 32 CPE8 elements along the radial
estimation models for various direction. Only four along the circumferential
thermomechanical properties. This is direction are used since quantities do not vary
accomplished by clicking on the Property in this direction. A distributed pressure load
Model button and choosing the appropriate of 50 N/mm2 was applied on the inner face of
model for each of the properties from the the cylinder. The cylinder is also subjected to
pull-down menu. The physical properties that a uniform temperature of 800K on the inner
are relevant for analysis are Young's modulus surface and 400K on the outer surface. Non-
and Poisson's ratio, coefficient of thermal homogeneous material properties are
expansion and coefficient of thermal modeled by grouping elements having the
conductivity. For the example under same radial coordinate into an element set
consideration, Kerner's equation is chosen for and assigning properties that are calculated at
Young's modulus and Poisson's ratio, Voigt a radial coordinate value corresponding to the
rule for thermal conductivity and modulus mid-point of these elements. There is a
based rule for thermal expansion coefficient. discrete jump in the property values from
This step is illustrated in Figure 8. elements in one element set to the elements in
The material composition function in the the adjacent element set.
heterogeneous object is transformed to a The resulting temperature and stress
physical property variation through the use of distributions for different shapes of the
property estimation rules. The heterogeneous parabolic material composition functions
solid model then gets converted to an object obtained by varying a1 are shown in Figures
model that contains information regarding 10 through 12 for a1 = ±0.018, ±0.01, 0.0,
physical attributes of the heterogeneous 0.02, 0.04. Values of a1 < ±0.019 and > 0.043
object. The primary materials under generated invalid volume fractions (> 1 or
consideration for analysis in this work are <0). The variation in a1 (and composition
Alumina (Al2O3) as the ceramic and function thereof) has negligible influence on
Aluminum (Al) as the metal. Mechanical and the temperature distribution as can be seen in
thermal properties for each of these individual Figure 10, while there is a significant
phases can be found in (Kwon, 1994). The influence on the radial and circumferential
object model generated herein can be used for stress distributions as evident in Figures 11
finite element analysis (FEA). This step is and 12. The magnitude of the circumferential
accomplished here by manually generating a stresses,  is much larger than the radial
finite element input file for the geometry and stresses, rr , and hence more detrimental to
material distribution of the heterogeneous failure. By systematically varying the values of
coefficient a1 and thus the material
object. Eventually, this step will be automated
composition function, the magnitude of the
by directly converting the object model into
an FEA input file through appropriate pre-
Figure 9 Finite element mesh
processing algorithms. The finite element
model can then be used for analysis.
5.1.1 Analysis
A linear elastic finite element analysis was
performed using ABAQUS (Hibbit, Karlsson
and Sorensen, inc). Only one quarter of the
domain is modeled because of symmetry of
geometry, loading and boundary conditions.
The domain was discretized using eight
noded plane strain elements (referred to as
CPE8 in the abaqus element library) for stress
analysis and eight noded two-dimensional
diffusive heat transfer elements (referred to as
DC2D8 in the abaqus library) for thermal
129
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 10 Temperature distribution

Figure 11 Circumferential stress distribution

Figure 12 Radial stress distribution

130
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

peak circumferential stresses can be reduced the CAD system in a manner similar to that
(Figure 11). Varying a1 in the parabolic described in example 1 is shown in Figure 13
material composition function also shifts the for different values of exponent n. The
location of the peak circumferential stress. A properties are estimated in an identical
minimum peak circumferential stress occurs fashion to that described in example 1. The
for a1 = ±0.018. A similar trend can be object model thus generated is used for finite
observed in the radial stress distribution element analysis.
(Figure 12). Thus, for a parabolic material
5.2.1 Analysis
composition function given by equation (5),
A linear elastic finite element analysis similar
coefficient a1 = ±0.018 results in an optimal
to that described in example 1 is performed
stress distribution. The other coefficients a0
using the ABAQUS finite element software.
and a2 can be deduced from equation (5).
Eight noded axisymmetric elements (called
The parabolic material composition function
CAX8R in the ABAQUS element library) are
was chosen and the coefficient a1 was varied.
used to discretized the piston head. Four layers
Other forms of material composition function
of elements are used to represent the steel
such as linear or cubic variations can also be
substrate, 20 layers are used to represent the
used in this variant synthesis process. The
graded layer and two layers are used to
cubic variation will provide additional
represent the ceramic coating. Each layer has
flexibility in controlling material composition
four elements.
by virtue of being able to vary two coefficients
The finite element mesh of the piston head
as opposed to one as in the case of the
is shown in Figure 14. The loading on the
parabolic function.
piston head is a uniform cooling of 200K
5.2 Example 2 from the fabrication temperature. No
The second synthesis example is a mechanical loads are applied. Axisymmetric
heterogeneous piston head (Suresh and boundary conditions are applied to nodes
Mortensen, 1997). A ``thick'' brittle coating of lying on the axis of the cylinder. The
zirconia is applied on to a steel substrate for translational degrees of freedom of the lower
protection against thermal exposure, most node lying on the axis are constrained.
environmental attack, contact failure and wear.
The coating is deposited on to the metal Figure 13 Heterogeneous solid models for different material compositions
substrate by a thermal spraying process. Direct
deposition of the coating will lead to interfacial
delamination due to high residual normal and
shear stresses generated from cooling (from
fabrication temperature).
The mechanical integrity of such a coating
is insured by having a graded layer in between
the zirconia coating and the substrate metal
(Figure 1).
An exponential material composition
function from Table I, row 2 is chosen:
zn
Vceramic ˆ …6†
t

It is defined with respect to a cylindrical

coordinate system attached to the center of
the piston head. In equation 6, t and n are
variables. Here, t is fixed and only n is varied
in the ``optimization'' process. Values of n =
0.2, 1.0, 2.0 and 3.0 are chosen to illustrate
the variant synthesis approach. The thickness
of the ceramic coating is chosen as 20mm, the
graded layer thickness is 80mm and the steel
substrate thickness is 40mm. The
heterogeneous solid model generated using
131
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

The resulting normal and shear stress 5.3 Layered manufacturing

distributions due to cooling are shown in In this paper, as mentioned earlier, we focus
Figures 15 and 16. The stresses are extracted on the integrated CAD system for
as a function of the z-location at mid points of heterogeneous objects modeling, and not the
elements whose boundary lie on the free edge fabrication process. However, we briefly
of the cylinder. The abrupt change in stresses describe the processing step using the
correspond to the ``ungraded'' or the situation heterogeneous model that links with a l
wherein the coating is applied directly ayered fabrication method. We consider the
on to the substrate metal. It can be observed example in Figure 7. In the same way, the
from these figures that minimum normal and piston head model can also be processed for
shear stresses are obtained for n = 1.0, the fabrication.
linear material composition gradient and The first step is the discretization of the
hence will be the conclusion of the design optimized model (Figure 17). This is
process. necessary to output data that is compatible

Figure 14 Finite element mesh

Figure 15 Normal stress distribution

132
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

Figure 16 Shear stress distribution

Figure 17 Discretization of Figure 7 Figure 18 Adaptive slicing of Figure 7

with the fabrication process resolution. While (Figure 18). Actual fabrication steps are
our rm-object models can be thought of as beyond the scope of this paper.
possessing infinite resolution, the fabrication
process in use is of a finite resolution. The
rm-object is therefore discretized into 6.0 Summary and conclusions
homogeneous lumps based on an acceptable
variability (designer specified) in material An integrated heterogeneous CAD system
composition. In this example, material which incorporates synthesis tools consisting
resolution of 0.1 was chosen in terms of of a library of material composition functions,
volume fraction, thus discretizing the object a library of property estimation rules and a
into 11 homogeneous lumps each of which graphical user interface was described. Such a
has a constant volume fraction of primary system would serve as a powerful tool for the
materials. designer easily and efficiently to perform
Finally, we make a slice file which is various tasks involved in design, analysis and
compatible with the fabrication process (e.g. manufacture of heterogeneous objects. Our
DMD) using adaptive slicing algorithm ongoing work involves integrating process
133
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

planning algorithms into the system such that References

once the heterogeneous object has been
designed, it can be fabricated using layered Bendsoe, M. and Kikuchi, N. (1988), ``Generating optimal
manufacturing techniques by interrogating topologies in structural design using a
homogenization method'', Computer Methods in
the heterogeneous solid model. Further, the Applied Mechanics and Engineering, Vol. 71,
solid builder module can also be incorporated pp. 197-224.
in the system for processing outputs from Bendsoe, M., Diaz, A. and Kikuchi, N. (1993), ``Topology
generative techniques or reverse engineering and generalized layout optimization of elastic
data from sources such as MRI or CT scans structures'', in Bendsoe, M.P. and MotaSoares, C.A.
(Eds), Topology Design of Structures, Kluwer
(biomedical applications) into heterogeneous
Academic Publishers, Dordrecht, pp. 159-206.
solid models which can be used for analysis or Cherkaev, A. (1994), ``Relaxation of problems of optimal
manufacturing applications. structural design'', International Journal of Solids
The use of the proposed system was and Structures, Vol. 31 No. 16, pp. 2251-80.
demonstrated by considering the synthesis of Cherakaev, A. and Kohn, R. (1997), Topics in the
sample heterogeneous objects, namely, a Mathematical Modeling of Composite Materials,
Birkhauser, Boston, MA.
thermal barrier type Functionally Graded Drake, J.T., Williamson, R.L. and Rabin, B.H. (1993),
Material and a heterogeneous piston head. ``Finite element analysis of thermal residual stresses
The synthesis process is carried out by first at graded ceramic-metal interfaces, Part II, Interface
choosing a parabolic material composition Optimization for Residual Stress Reduction'', Journal
function and an exponential function from of Applied Physics, Vol. 74 No. 2, pp. 1321-6.
Hirano, T. and Wakashima, K. (1995), ``Mathematical
the library of composition functions for the
modeling and design'', Special Issue on Functionally
thermal barrier FGM and the piston head, Gradient Materials, MRS Bulletin, pp. 40-2.
respectively. An initial guess was made for the Hoffmann, C.M. (1989), Geometric & Solid Modeling,
unconstrained coefficient, a1, in the parabolic Morgan Kaufmann Publishers, San Matio, CA.
composition function and exponent, n, in the Hoffmann, C.M. and Rossignac, J.R. (1996), ``A road map
to solid modeling'', IEEE Transactions on
exponential function. The material
Visualization and Computer Graphics, Vol. 2.
composition together with the geometry was Holt, B., Koizumi, M., Hirai, T. and Munir, Z. (Eds) (1993),
then represented as a heterogeneous solid ``Functionally gradient materials'', Ceramic
model in the rm-object framework. The Transactions, Vol. 34.
heterogeneous solid model was then Jackson, T.R., Patrikalakis, N.M., Sachs, E.M. and Cima,
converted into an object model by choosing M.J. (1998), ``Modeling and designing components
with locally controlled composition'', Proceedings of
appropriate property estimation models the Solid Freeform Fabrication Symposium, Austin,
chosen from the library of property estimation TX.
rules. Kumar, V. (1998), ``Solid modeling and algorithms for
A finite element input file required for heterogeneous objects'', PhD Thesis, Department of
analysis was generated manually from the Mechanical Engineering and Applied Mechanics,
University of Michigan, Ann Arbor, MI.
information on geometry and material
Kumar, V. and Dutta, D. (1998), ``An approach to
property distribution based on the object modeling and representation of heterogeneous
model. A proposed area of future work would objects'', ASME Journal of Mechanical Design, Vol.
be to automate this step by directly converting 120 No. 4, December, pp. 659-67.
the object model into a finite element input Kumar, V., Burns, D., Dutta, D. and Hoffman, C. (1998), ``A
framework for object modeling'', Computer Aided
file through appropriate preprocessing
Design, to appear, Department of Mechanical
algorithms. Engineering and Applied Mechanics, University of
While inclusion of materials is a departure Michigan, Ann Arbor Technical Report, UM-MEAM-
from geometry-only systems, the focus is still 98-07.
on representation. Future CAD systems can Kwon, P.Y. (1994), ``Macroscopic design and fabrication of
possibly provide synthesis environments for functionally gradient materials'', PhD Thesis,
Department of Mechanical Engineering, University
``functional'' designs that include the creation
of California, Berkeley, CA.
of geometry and material, not in an iterative Markworth, A.J. and Saunders, J.H. (1995), ``A model of
manner, but following sound design structure optimization for a functionally graded
principles such as the axiomatic design (Suh, material'', Materials Letters, Vol. 2, pp. 103-7.
1999). That is, systems that will allow the Markworth, A.J., Ramesh, K.S. and Parks, W.P. (1995),
``Modelling studies applied to functionally graded
mapping of functional requirements (FRs) of
materials ± review'', Journal of Material Science,
products to design parameters (DPs) and Vol. 30, pp. 2183-93.
then use the DPs to create the associated Marsan, A. and Dutta, D. (1996), ``Construction of a
geometry and material distribution. surface model and layered data from 3D
134
 An integrated CAD system for design of heterogeneous objects Rapid Prototyping Journal
Srinivas Bhashyam, Ki Hoon Shin and Debashish Dutta Volume 6 . Number 2 . 2000 . 119±135

homogenization output'', Journal of Mechanical Suresh, S. and Mortensen, A. (1997), ``Functionally graded
Design, Vol. 118 No. 3, pp. 412-18. metals and metal-ceramic composites:
Marsan, A. and Dutta, D. (1999), ``Computational thermomechanical behavior'', International
techniques for automatically tiling and skinning Materials Reviews, Vol. 42 No. 3, pp. 85-116.
branched objects'', Computer & Graphics, Vol. 23 Tanaka, K., Watanabe, H., Sugano, Y. and Poterasu, V.F.
No. 1.
(1996), ``A multicriterial material tailoring of a
Mazumder, J., Choi, J., Nagarathnam, K., Koch, J. and
hollow cylinder in functionally gradient materials:
Hetzner, D. (1997), ``The direct metal deposition of
H13 tool steel for 3-D components'', JOM, Vol. 49 scheme to global reduction of thermoelastic
No. 5, pp. 55-60. stresses'', Computer Methods in Applied Mechanics
Merz, R. ,Prinz, F.B., Ramaswami, K., Terk, M. and Weiss, and Engineering, Vol. 135, pp. 369-80.
L. (1994), ``Shape deposition manufacturing'', in Weng, G.J., Taya, M. and Abe, H. (Eds) (1990),
Proceedings of the Solid Freeform Fabrication Micromechanics and Inhomogeneity, Springer-
Symposium, The University of Texas at Austin, Verlag, Berlin.
pp. 1-8. Williamson, R.L., Rabin, B.H. and Byerly, G. E. (1995),
Nemat-Nasser, S. and Hori, M. (1993), Micromechanics: ``FEM study of the effects of interlayers and creep in
Overall Properties of Heterogeneous Materials, reducing residual stresses and strains in ceramic-
Elsevier Scientific Publishers, Amsterdam.
metal joints'', Composites Engineering, Vol. 5 No. 7,
Requicha, A. (1980), ``Representations for rigid solids:
theory, methods and systems'', Computing Surveys, pp. 851-63.
Vol. 12 No. 4. Zuiker, J.R. (1995), ``Functionally graded materials: choice
Suh, N.P. (1999), in Kals, H. and Houten, F. van (Eds), of micromechanics model and limitations in
Applications of Axiomatic Design, Integration of property variation'', Composites Engineering, Vol. 5
Process Knowledge into Design Support Systems. No. 7, pp. 807-19.

135