Content uploaded by Raivo Sell
Author content
All content in this area was uploaded by Raivo Sell on Jan 02, 2014
Content may be subject to copyright.
Proceedings of the ASME 2009 International Design Engineering Technical Conferences &
Design Automation Conference
IDETC/DAC 2009
August 30-September 2, 2009, San Diego, USA
DETC2009-87776
OPAS: ONTOLOGY PROCESSING FOR ASSISTED SYNTHESIS
OF CONCEPTUAL DESIGN SOLUTIONS
Franc¸ois Christophe∗
Department of Engineering Design and Production
Helsinki University of Technology
and IRCCyN Laboratory - Ecole Centrale de Nantes
P.O. Box 4100 (Otakaari 4)
FI-02015 TKK, Finland
Email: francois.christophe@tkk.fi
Raivo Sell
Department of Mechatronics
Tallinn University of Technology
Ehitajate tee 5
EE-19086 Tallinn, Estonia
Email: raivo.sell@ttu.ee
Alain Bernard
IRCCyN Laboratory - Ecole Centrale de Nantes
1, rue de la No ¨
e, BP 92101
44321 Nantes Cedex 3, France
Email: alain.bernard@irccyn.ec-nantes.fr
Eric Coatan´
ea
Helsinki University of Technology
P.O. Box 4100 (Otakaari 4)
FI-02015 TKK, Finland
Email: eric.coatanea@tkk.fi
ABSTRACT
This article focuses on a key phase of the conceptual design,
the synthesis of structural concepts of solution. Several authors
have described this phase of Engineering Design. The Function-
Behavior-Structure (FBS) is one of these models. This study
is based on the combined use of a modified version of Gero’s
FBS model and the latest developments of modeling languages
for systems engineering. System Modeling Language (SysML) is
a general-purpose graphical modeling language for specifying,
analyzing, designing, and verifying complex systems. Our de-
velopment shows how SysML types of diagrams match with our
updated vision of the FBS model of conceptual design. The ob-
jective of this paper is to present the possibility to use artificial
intelligence tools as members of the design team for supporting
the synthesis process. The common point of expert systems de-
veloped during last decades for the synthesis of conceptual solu-
tions is that their knowledge bases were application dependent.
∗Address all correspondence to this author.
Latest research in the field of Ontology showed the possibility
to build knowledge representations in a reusable and shareable
manner. This allows the construction of knowledge representa-
tion for engineering in a more generic manner and dynamic map-
ping of the ontology layers. We present here how processing on
ontology allows the synthesis of conceptual solutions.
KEYWORDS: Conceptual Design, FBS model, Systems
Engineering, SysML, Knowledge Representation, Ontology, Tax-
onomy, Semantic web standards
1 INTRODUCTION
Researchers in the field of engineering indicate that an
important part of the future properties of a product is being
strongly constrained during the design phase and more specif-
ically during the conceptual stage of design. Decisions made
at the conceptual design stage have important influence on
1 Copyright c
2009 by ASME
all of the future performance factors of the final service or
product [1] [2]. A poorly conceived design concept can never
be compensated by a good detailed design. Methodologies
have been developed in order to assist the different phases of
conceptual design [3] [4]. We base our research on a particularly
interesting model of conceptual design on the knowledge repre-
sentation viewpoint, Gero’s Function-Behavior-Structure model
(FBS) [5]. The development of models for product design as a
systematic process suggested the possibility to build computer
applications enabling the semi-automatic conception of systems.
The synthesis of design solutions is a stage of the design
process in connection with creativity and innovation. For
many designers and scientists, creativity is seen as a rare,
somehow magic, skill and thus, is perceived as highly valuable.
Nevertheless, besides being strongly connected with individual
psychology, creativity also have strong relations with more con-
crete aspects [4] such as knowledge and information. Therefore,
the use of knowledge representation and reasoning about this
knowledge is needed for assisting designers in their tasks [6].
In most practical cases, engineers and designers focus on few
known solutions that they will try to improve because of the
lack of time before decision. Therefore, the use of computer
power for assisting the synthesis of solutions would allow
searching the design space more thoroughly. Additionally,
because of the tremendous growth of complexity in systems,
design often has to be collective and concurrent. Both of these
characteristics require to systemize the design reasoning and
to define a sufficiently abstract language in order to acquire
generic applicability to different industries and practices. Recent
researches in knowledge representation have organised the
necessary knowledge for engineering design into upper level
ontologies. This representation enables reasoning and assisting
designers in the exploration of the design space during the
synthesis of conceptual design solutions.
In this publication we will present our research on expanding
designers’ knowledge on the possible solutions available during
the synthesis of conceptual solutions. First of all we will present
the state-of-the-art of systems engineering in conceptual design
as well as the state-of-the-art in knowledge representation and the
recent developments of ontology for engineering design. Sec-
ondly, we will present our model of conceptual design and our
contribution in the organisation of a mid-level ontology for the
synthesis stage of conceptual design. Our main concern is to cre-
ate reusable knowledge which remains application independent.
Additionally, we will show how this knowledge organisation al-
lows taking engineering from document-based to model-driven
engineering by the use of data interchange standards. In the third
section we will present the results of our research and the de-
veloped application in further details. We will then discuss the
novelty of our method and tool as well as its limitations and con-
clude by summarizing our findings and give the possible ways to
our future research.
2 STATE-OF-THE-ART
In this section, we will first present conceptual design in
the scope of systems engineering and secondly we will have an
overview of existing expert systems and latest developments of
ontologies for conceptual design.
2.1 Design considerations
2.1.1 Conceptual Design and Models Conceptual
design mainly consists of three phases: the refinement of the de-
sign problem, the synthesis of conceptual solutions and the eval-
uation of these solutions. Lots of fruitful work has been done
in respect to the evaluation phase of conceptual design and our
research group has also tackled this issue according to the new
environmental needs [7]. The purpose of this study is to bring
our focus upfront by tackling the synthesis stage of conceptual
design. In fact, in order to enable the evaluation of concepts one
must previously have searched the design space for most pos-
sible solutions available. Through the years, researchers have
provided guidelines, methods and models for engineering de-
sign and precisely for our purpose, for conceptual design. These
models became very precise and suggest a systematic applica-
tion of them for a successful conception. This systematic aspect
brought the idea that an artificial intelligence application could,
as well, learn how to generate design solutions. For that pur-
pose, researchers started to build representations of the necessary
knowledge needed to be able to conceive a solution to a design
problem. One of these representations is Gero’s FBS model [5].
Figure 1 presents this model. In this model:
Frepresents a set of functional variables, the necessary knowl-
edge in order to be able to explain what the system should
do,
Be is the expected behavior of the system, e.g. the set of vari-
ables showing how the system should work,
Sis the set of variable representing the physical structure of the
system,
Bs is the set of variables enabling the representation of the ef-
fective behavior of the system, e.g. its “actual” behavior
Drepresents the variables contained in the documentation given
for more detailed design.
The different stages of conceptual design are represented by
eight fundamental processes in Gero’s FBS model:
1. Formulation: transforms the design problem, expressed in
function (F), into behavior (Be) that is expected to enable
this function.
2. Synthesis: transforms the expected behavior (Be) into a so-
lution structure (S) that is intended to exhibit this desired
2 Copyright c
2009 by ASME
Figure 1. GERO’S FBS MODEL
behavior.
3. Analysis: derives the “actual” behavior (Bs) from the syn-
thesized structure (S).
4. Evaluation: compares the behavior derived from structure
(Bs) with the expected behavior to prepare the decision if
the design solution is to be accepted.
5. Documentation: produces the design description (D) for
constructing or manufacturing the product.
6. Reformulation type 1: addresses changes in the design state
space in terms of structure variables or ranges of values for
them.
7. Reformulation type 2: addresses changes in the design state
space in terms of behavior variables or ranges of values for
them.
8. Reformulation type 3: addresses changes in the design state
space in terms of function variables or ranges of values for
them.
This model has been the reference for building expert sys-
tems based on this triplet of knowledge. Nevertheless, it nowa-
days needs some updating, for instance due to model-driven sys-
tems engineering, process 5 can be automatically generated and
the documentation has become less important than the models
themselves.
2.1.2 Model-driven design and systems engineer-
ing language Last significant developments in systems engi-
neering is the new modeling language - SysML (System Model-
ing Language) derived from UML 2.0 (Unified Modeling Lan-
guage). SysML 1.1 specification [8] was published last year
by the Object Management Group (OMG) and is composed by
many industry leading corporations and organizations. This new
language brings closer software design concept and product de-
sign, enabling the natural synergy of multidisciplinary design of
products (e.g. software, mechanics, electronics and others) at the
very beginning of the design process and continuing to support
it through the design lifetime. SysML specification is defined by
using UML 2.0 specification techniques. These techniques are
used to achieve the following goals.
1. Correctness
2. Precision
3. Conciseness
4. Consistency
5. Understandability
SysML concept is similar to software design techniques, but ex-
pands it in several ways. As engineering design, mechatronic
product design is not a pure technical problem anymore. It has
become a rather complex activity, which needs to also involve
artifacts, people, environment, market, in addition to hardware
and software components. In order to reach an understanding,
all these aspects have to be modeled in the same methodical way.
SysML tries to provide the generic language and environment to
support complex systems engineering design process.
In general, SysML diagrams are divided into three main
groups as shown Figure 2.
Figure 2. SYSML DIAGRAM TYPES
Depending on the design concept some diagrams can be
used for different purposes. A widely used example is Use Case
diagram which describes the system functions or services and
can be successfully used for clarifying system requirements as
well as main activities. It is also common to regroup SysML dia-
grams into four main pillars where parametric diagram becomes
a separate group. Figure 3 shows the generic diagrams in four
pillars.
By this brief presentation of SysML, we have shown that the
diagrams provided by the language fit the description of Function
(with Use Case, Block Definition and Internal Block diagrams),
Behavior (with State-Machine, Activity and Sequence diagrams)
and Structure (components, Packages and Internal Block dia-
grams). Nevertheless, SysML provides more than Gero’s FBS
because it contains Requirement and Parametric diagrams. Addi-
tionally, one of the strength of SysML is that it is a modeling lan-
guage, and, as such, engineering becomes now model centered in
contrast with document centered. First of all, documents can now
be generated automatically but, most of all, model based engi-
neering allows the verification and validation of the coherency
3 Copyright c
2009 by ASME
Figure 3. THE FOUR PILLARS OF SYSML
of each phase of systems engineering. This coherency is in-
sured and maintained by the coherency between each models of a
project. This coherency can even be verified nowadays by com-
puter tools called model checkers. Additionally, model-driven
engeineering enables the use of design patterns, stereotypes, spe-
cific profiles and libraries for reuse which was not possible with
document based engineering.
2.2 Knowledge representation considerations
FBS framework has seen the development of several differ-
ent Computer Aided Conceptual Design tools. This chapter will
make a brief state-of-the-art of these expert systems developed
during the second part of the 90’s. In a second part we will
present the latest researches of Artificial Intelligence in the field
of ontology and the development of standards for knowledge
interchange. The third part of this chapter will then present
the latest tools and ontology developed for engineering design
which introduces our problematic.
One of the first modeler for designers using FBS knowledge
representation was Umeda’s FBS Modeler in 1995 [9]. This
modeler uses both functional and behavioral representations.
In addition to these two bases, FBS Modeler uses Qualitative
Reasoning from Qualitative Physics and as such, the system
becomes able to derive unwanted and unexpected behaviors
of the future product. Additionally, this modeler includes a
module for adding functional redundancy to a part of the future
product. EFDEX, Engineering Functional Design Expert, also
involves ontology in its inner base [10]. This Expert System
has especially put an accent on the functional decomposition
of systems. In fact, a function as such can be decomposed
into functions from a lower focus with finer granularity. This
knowledge base includes different types of decomposition
and logical way of decomposing functions. This system has
particularly developed this decomposition part and a set of
domain-specific rules for decomposition is included. This set
of rules allows the system to automatically decompose main
functions of a system into different functional architectures.
Another relevant work using the FBS knowledge base is from
Tian et al. [11]. In their research, they extended the FBS frame-
work into T-F-B-S where T stands for Topology. Therefore,
they have included a topological insight on the future product in
order to show some of its important geometrical features. Their
system, as the others, integrates a functional representation, and,
additionally, a knowledge base of existing topological drawings
of mechanisms. This system also integrates an inference engine
in order to advise designers and derive his structures to simplify
them and create different concepts. This brief overview - several
other software environments for designing have been developed
on the same principles [12] - has shown interesting points on
the development of assisting tools for designer. First of all, all
of these applications are based on strong use of Knowledge
Representations and Knowledge bases. In our viewpoint, this
is an unavoidable feature of such kind of environment. In fact,
in order to be able to advise designers during their work, a
system should acquire similar knowledge about engineering.
Secondly, an ontological description is, most probably, the most
suitable concept to allow deduction on. For example, we’ve seen
above that most of the systems developed contained inference
engines in order to deduce, from its knowledge and designers
ideas, either errors of conception or new concepts of solution.
Nevertheless, these applications showed limitations in terms of
design of more complex systems involving multidisciplinary
fields because of their specificity in the field of mechanical
engineering. Moreover, due to this specificity each knowledge
base was difficultly reusable and strongly application-dependent.
Recent developments in Artificial Intelligence and the field of
ontology allowed the production of knowledge bases being inde-
pendent from application or service using these representations.
In the beginning of the 90’s, researchers in Artificial Intelli-
gence started to put their interest toward the notion of Ontology,
commonly used in Metaphysics, in order to formalize knowl-
edge. The main goal of this field is to “represent” what “exists”.
Within this context, they have defined an ontology as an artifact
enabling the representation of the existing using a formal and
consensual vocabulary. One of the first definitions of ontology,
commonly admitted in Artificial Intelligence, was published by
Gruber [13] as the explicit specification of a conceptualization.
This definition has then been refined by R. Studer [14] as the
formal and explicit specification of a shared conceptualization:
formal because an ontology needs to be read by machines,
which excludes natural language,
explicit because its definition explains the concepts being used
and their constraints of use,
conceptualization is the abstract model of a phenomenon from
4 Copyright c
2009 by ASME
the real world with identification of the key concepts of this
phenomenon,
shared because an ontology is not an individual’s property but
represents a consensus accepted by a community of users.
Practically, an ontology gives means to express concepts of
a domain by organizing them hierarchically and by defining their
semantic properties in a language of representation of knowl-
edge helping the computer applications using this language to
share a consensual view on that domain. Defining concepts and
link them together with semantic relations corresponds to the first
level of an ontology, the conceptual model, inspired of semantic
networks and moreover of conceptual graphs from Sowa [15].
The semantic web community is using the idea of ontology in
order to express the semantic content of web pages in order to
enable their exploitation by computer agents and not only by hu-
man users [16]. In fact, Tim Berners-Lee presents the semantic
web as an extension of the current one, in which information is
given well-defined meaning, better enabling computers and peo-
ple to work in cooperation [17]. In order to apply this vision
practically, an architecture composed of a set of languages has
been defined. This architecture is generally represented in the
form of a pyramid. Each layer sits on results defined at the lower
layers, thereby, each layer is progressively more specialized and
more complex than the previous one. Additionally, each level is
independent of the upper levels so as to be developed and oper-
ational in an autonomous manner according to the developments
of upper levels. Figure 4 shows this pyramid of languages.
Figure 4. PYRAMID OF SEMANTIC WEB LANGUAGES IN 2006
(FROM BERNERS-LEE)
Our interest here is focused on the description of concepts
used at the conceptual design of systems. Therefore we place
our activity on the first level of ontology, which is the level
of description of the conceptual model. We so concentrate
on OWL and RDF languages of the pyramid Figure 4. RDF
enables the description of metadata in the form of triplets
(resource, propriety, value) as specified in RDF schemes.
Ontology Web Language, OWL, allows the description of
more complex ontologies because it defines classes, attributes,
relations and axioms. The biggest interest of using owl is that
it can then be combined with rule-based languages allowing
reasoning about resources and inferring new knowledge about
them. Nevertheless, first of all, we need to have an appropriate
level of description of our domain and therefore define the
mid-level ontology of conceptual design in the sense of systems
engineering. The following chapter will present what exists in
terms of shared ontologies of conceptual design.
Researchers in engineering design have started to work on
the necessary knowledge for the design of systems such as me-
chanical systems and multi-disciplinary systems like mecha-
tronic systems. Kitamura, who has been involved in the early
development of frameworks for conceptual design as stated ear-
lier, developed a project called FOCUS, standing for a Functional
Ontology for Categorization, Utilization and Systematization of
functional knowledge. This project represents all the knowledge
required for the functional representation of systems as shown
Figure 5. This figure shows the organisation of different on-
tologies representing different knowledge about functions. The
project contains an ontology of device and function, a functional
concept ontology, descriptions of functional models of concrete
devices and the of function achievement and a reference ontology
of function. This work provides a very good insight on the de-
velopment of functional architectures for describing the features
of a device [18].
Figure 5. LAYERS OF FUNCTIONAL ONTOLOGIES IN THE FOCUS
PROJECT (FROM KITAMURA ET AL.)
The work of Chakrabarti’s team also makes reference in the
automatic synthesis of mechanical systems [19]. Their com-
puter application called FuncSION helps synthesizing topolog-
ical structures of mechanical objects which can then be trans-
formed into spatial structures embodying the functions. The
testing of earlier version of FuncSION was reported in Artifi-
cial Intelligence in Design in 1996 [20] and it was found that
5 Copyright c
2009 by ASME
the amount of generated solutions and their variety were always
larger than those of the designers. The interest of our research in
FuncSION is that it is also ontology based even though it is dedi-
cated to mechanical products, further readings will show how we
provide ontologies for systems engineering. In terms of systems
engineering, Tudorache recently published ontologies represent-
ing conceptual design knowledge in her doctorate thesis [21].
Her set of ontologies of engineering represents three ontologies
dedicated to requirements, components and systems, and con-
straints. We will focus on her description of components and
their organisation within a system as our aim is the synthesis
of solutions. Figure 6 represents the taxonomy of components.
Components may only be composite or atomic and compos-
ite components may contain other components whereas atomic
components may not. Tudorache ontologies are published under
Prot´
eg´
e1server and are usable under General Public License.
Figure 7 shows an example of the hierarchy of components and
their relationships.
Figure 6. TUDORACHE’S TAXONOMY OF COMPONENTS
This approach fits systems engineering demands and
provides a good basis for our research. Nevertheless, for her,
a component can as well be functional or structural whereas it
is important for us to reify 2the concept of function even if its
representation has typically the structure of a component, e.g. a
black-box.
This state-of-the-art sets two research problems. How to as-
sist engineers in the use of semantic tools? How to integrate
ontology search with SysML semantics? We will address these
questions in the following section.
3 CONTRIBUTION
In this section, we will first present our vision of con-
ceptual design through a model adapted from Gero’s FBS in
order to fit nowadays engineering. In light with this new
Requirement-Function-Behavior-Structure model, RFBS model,
1http://protegewiki.stanford.edu/index.php/Engineering ontologies
2to reify is to transform an abstract thing into a concrete one [15]
Figure 7. CLASS HIERARCHY OF COMPONENTS (FROM TUDO-
RACHE)
we will shortly present the general framework for intelligent con-
ceptual design. This framework is a meta-model of Conceptual
Design addressed for Systems Engineering. Secondly, we will
present the mid-level ontology that we have developed and its
integration with other ontologies of engineering and the connec-
tions with lower taxonomies.
3.1 RFBS Model for Model-driven Systems Engineer-
ing
As shown previously, Gero’s FBS model contains few points
which needed to be updated according to nowadays knowledge
about systems engineering. Apart from the fact, that this previ-
ous model did not include a requirement phase and that the docu-
mentation phase has now became secondary compared to models
themselves, a major point needs to be reanalyzed at the scale of
the synthesis phase of conceptual design. In fact, Gero stated
that the only possible link between function and structure was
through the expression of behavior. We argue here that it is pos-
sible to create “embryos” of structures out of functions only. We
call these preliminary structures generic or abstract structures.
As abstract classes in object oriented programming, their aim is
only to encapsulate each atomic function of the system into one
or more of the six families of organs from our prior works [22]
and in agreement with the bond graph theory. Therefore, we pro-
pose this RFBS model shown Figure 8. This model represents
the conceptual design process in a practical way. It corresponds
to the way we formulate conceptual solutions using SysML and
our computer tool: OPAS, guide for the synthesis of conceptual
solutions. This model will then allow us to notice the strengths
6 Copyright c
2009 by ASME
and limitations of the method.
Ris the set of constraints and performance variables required
Frepresents a set of functional variables, the necessary knowl-
edge in order to be able to explain what the system should
do according to requirements, thus F is derived from R
Be is the expected behavior of the system, e.g. the set of vari-
ables showing how the system should work, Be is set ac-
cording to Requirements and Functions,
GS is the representation of Generic Structure, e.g. abstract
classes encapsulating function and intriseque attributes, GS
is derived from F,
Sis the set of variable representing the physical structure of the
system, S is specializes GS according to Be.
Bs is the set of variables enabling the representation of the ef-
fective behavior of the system, e.g. its “actual” behavior,
Drepresents the continuation to a more detailed phase of sys-
tems engineering.
Figure 8. RFBS MODEL FOR SYSTEMS ENGINEERING
We have defined seven types of processes involved during
conceptual design:
1. Formulation type 1 (process 1): transforms the design
problem, expressed in requirements (R), into functions
(F) that the system should provide.
Formulation type 2 (process 1’): transforms the design
problem, expressed in function (F) and requirements
(R), into behavior (Be) that is expected to enable this
function with the performance set by the requirements.
2. Pre-synthesis: transforms the functional architecture of the
system (F) into a generic structure (GS) using abstract or-
gans.
3. Synthesis: specializes GS according to the expected behav-
ior (Be) into a solution structure (S) that is intended to ex-
hibit this desired behavior.
4. Analysis: derives the “actual” behavior (Bs) from the syn-
thesized structure (S).
5. Evaluation: compares the behavior derived from structure
(Bs) with the expected behavior to prepare the decision if
the design solution is to be accepted.
6. Detailing: prepares all drawn models for the detailed design
phase (from work classes into technology involvement)
7. Reformulation type 1 (process 7): addresses changes
in the design state space in terms of structure variables
or ranges of values for them.
Reformulation type 2 (process 7’): addresses changes
in the design state space in terms of abstract organs
or generic structure variables or ranges of values for
them.
Reformulation type 3 (process 7”): addresses changes
in the design state space in terms of function variables
or ranges of values for them (this reformulation in-
duces automatically changes in the expected behavior)
Reformulation type 4 (process 7”’): addresses changes
in the design state space in terms of requirement vari-
ables or ranges of values for them (this reformulation
involves discussion with the client for finding agree-
ment)
As explained section 2, SysML diagrams can be mapped
to some extent with each state of the RFBS model. Figure 9
represents the general framework of conceptual design of intel-
ligent systems engineering and allows us to present how SysML
diagrams fit with the different states of our RFBS model and are
used for modeling a systems through the entire conceptual design
process.
Figure 9. FRAMEWORK OF CONCEPTUAL DESIGN
This framework Figure 9, derived from our previous work
[23] [24], positions our research on tackling the entire conceptual
design process from the refinement of ill-defined problems to the
evaluation of the ecological impact of the product through its life-
cycle. The first stage of conceptual design involves an important
part of problem solving issues: the refinement of ill-defined prob-
7 Copyright c
2009 by ASME
lems with the refinement of verbal requirements into formal re-
quirements represented with SysML requirement diagrams. This
type of diagram represents requirements hierarchically and in or-
der of importance. Requirements diagram correspond to state
R of the RFBS model. Concurrently, the verbal requirements
are analyzed in order to extract the expected functionalities of
the future product. This analysis allows the creation of use case
diagrams presenting the system services and its surrounding en-
vironment. More detailed insight into the needed functionali-
ties enables the creation of functional architectures developing
different functional structures represented as “black boxes” with
SysML block definition diagrams and functional blocks [25] as
represented Figure 10.
Figure 10. FUNCTIONAL BLOCK SYSML REPRESENTATION [25]
Following the refinement phase of conceptual design, the
synthesis of concepts phase has the aim to transform functional
models into structural models enabling the system’s organisa-
tion in sub-systems and components. This phase is the core
of our research and positions OPAS, our application for auto-
matic synthesis within the global framework. OPAS can take
both functional architecture diagrams and use cases diagrams and
produces generic solutions using an online semantic atlas (SA)
based on contextual closure between verbs. OPAS produces in
a first step the generic solutions in the form of block definition
diagrams and secondly specializes these solutions into physical
components in the form of block definition diagrams and internal
block diagrams.
The evaluation phase of conceptual design is done with the
help of behavioral diagrams of different types. Activity dia-
grams, sequence diagrams and state-machine diagrams are de-
rived from both requirements and functional architectures and
represent the behavior expected from the system. Similarly, the
same set of diagrams represent the structural behavior of the sys-
tem and are derived by designers from the block diagrams rep-
resentations. Additionally, systems engineers represent the con-
nections between variables involved in the structure of the system
with parametric diagrams. The evaluation of structural concepts
of solution is then enbaled through the comparison of these two
sets of behavioral diagrams.
Probably the most interesting point in using SysML com-
pared to previous models is the use of profiles proposing man-
ners for developing systems as well as the use of design patterns
providing methods to answer known problems. Additionally,
SysML gives the possibility to use several libraries of existing
subsystems as well as libraries for the simulation of the system’s
behavior.
This section proposes an overall model of conceptual design,
in the following section we will now center the subject on the
synthesis phase and present the diverse elements used to enable
the automatic synthesis of concepts.
3.2 A reusable distributed ontology of conceptual de-
sign
In this section, we place our focus on the explanation of pro-
cesses two and three of our model: pre-synthesis and synthesis.
The main point of this section is to show what we consider as the
necessary knowledge needed for a machine to be able to interact
with humans at the same level of understanding than them. In
this part we provide an ontology of the needed concepts during
the synthesis of conceptual design solutions. Figure 11 shows
the concepts contained in this ontology and their relationships.
Figure 11. MID-LEVEL ONTOLOGY DESCRIBING THE SYNTHESIS
OF DESIGN SOLUTIONS
This ontology can be integrated with the work of Tudorache.
It is a frame-based ontology developed in OWL/RDF(S) with
Prot´
eg´
e3, an ontology editor. It is planned to publish this on-
tology and make it available for the community. This chapter
presents the core concepts of this ontology. As the synthesis of
solutions starts from functional architectures, it is important to
have knowledge about function. A function is, from our view-
point, represented by a verb and is a connection between an ini-
tial situation and a final situation. Hirtz have established a taxon-
3http://protege.stanford.edu/
8 Copyright c
2009 by ASME
omy of standard functions [26] and according to empirical stud-
ies, 94% of the functions could be described using the functional
basis proposed by Hirtz et al. [27]. These studies are based on
a total of 207 descriptions of functions representing various sub-
assemblies of an aero-engine and an aircraft design taken from
two different companies. Our approach is to use this taxonomy
at its full potential but with leaving freedom of vocabulary to
designers. Therefore, we will transform the functional descrip-
tion made by designers with their own vocabulary into a stan-
dard functional description. In order to do this standardization
we send requests to an online semantic atlas [28] based on the
concept of contexonyms. After this step, the next is to associate
each standard function of the lowest level of description to one
or more of the six families of abstract organs [22]. These organs
can be then specialized according to the type of energy they in-
volve. In fact, we also use the taxonomy of fields from Hirtz
in order to determine the type of variables involved in the in-
put/output of physical components. Our ontology is a mid-level
ontology and integrates Hirtz taxonomies of functions and fields
of energies. The concept of standard function has the elements of
the functional taxonomy as instances of our ontology. Similarly,
the concepts of energy and variables integrate the taxonomies of
generalized variables according to fields of energies.
4 APPLICATION EXAMPLE
In this section we will first present the application we devel-
oped around our ontology of synthesis of conceptual solutions
and, in a second time, we will show the results we obtain on the
study of the conceptual design of a mobile platform.
4.1 Software presentation
Our application is conceived on the fact that its meta-data
should be inter-operable with any SysML modeler. As SysML
“standards” for interoperability are XMI/AP233, we developed
OPAS to be compliant with them. In the same way, our ontol-
ogy and the taxonomies under it were exported in XMI format
in order to simplify the query system which is now only done
by parsing files. Figure 12 shows the overall interoperability of
OPAS and its input and output files.
Figure 12. INTEROPERABILITY OF OPAS WITH SYSML MODELERS
The structure of OPAS is shown Figure 13. The main class
asks for services from XMIParser, DBHandler and AtlasHandler
classes. XMIParser is in charge of receiving the functional archi-
tecture in the form of Use Case diagram from a SysML modeler
and, in the end, of sending the synthesized structures in the form
of Internal Block diagrams. DBHandler is in charge of sending
requests to the ontology for getting the instances of functions,
abstract organs and physical components in return. AtlasHandler
sends requests to the online semantic atlas in order to find cor-
respondence between designers’ vocabulary and standard func-
tions or functions and abstract organs.
Figure 13. OPAS CLASS DIAGRAM
4.2 Example
We base our study on the design of a mechatronic product: a
mobile robot platform. The entry point of the program is the Use
Case diagram shown Figure 14. In fact, the program receives it
as an XMI file which will be parsed in order to find the functions
to be fulfilled by the future system.
For the example, we will here only focus on the function
“to manipulate an object”. Looking up in Hirtz’s taxonomy of
functions, the program finds out that manipulate is not a standard
function. Thus, it sends a request on the semantic atlas 4online
to find corresponding verbs. Figure 15 shows the list of verbs
returned.
The verbs in the list belonging to the taxonomy of standard
functional verbs are “to control” and “to guide”. We stop itera-
tions at the first list where a verb belonging to the taxonomy is
found in order to avoid combinatorial explosion. Then, the pro-
gram suggests both verbs as standard verbs for “manipulate” as
well as their logical composition “guide and control” and “guide
or control”. Manipulate will be replaced by the leaves of Hirtz’s
classification containing respectively control and guide. Figure
16 shows these leaves corresponding to the verb manipulate.
The next phase of the program is to find the abstract organs
corresponding to this set of verbs. There are six abstract classes
4http://dico.isc.cnrs.fr/dico/en/chercher
9 Copyright c
2009 by ASME
Figure 14. USE CASE DIAGRAM OF A SERVICE ROBOT
Figure 15. LIST OF VERBS CORRESPONDING TO ’MANIPULATE’
Figure 16. STANDARD VERBS FOR MANIPULATE
of organs: Storage, Source, Dissipation, Conversion, Distribu-
tion and Calculation. The program sends parallel requests to the
semantic atlas with the six abstract classes verbs and the stan-
dard verbs “control” and “guide”. The results are shown Figure
17 with arrows representing the sense of the request. This fig-
ure shows the different semantic paths towards abstract organs.
Pursuing the algorithm further would provide irrelevant seman-
tic paths linking, for example, control with dissipation organs.
Therefore we propose a heuristic for setting the stop point of this
algorithm, which is that the program will stop sending request
on the rank after the first paths are found. For instance, in our
example, the rank of the first solution found is 1 because only
one verb separate guide and control from distribute or calculate.
Thus we stop our algorithm on the paths found at rank 2 which
provides us the path toward conversion organs.
Figure 17. MAPPING FUNCTIONS WITH GENERIC ORGANS
We derive from Figure 17 that, in order to “guide and con-
trol” an object with a robot, the generic organs to be used are
a combination of distribution, conversion and calculation or-
gans. Further, in order to specify these abstract organs, the pro-
gram needs input on the chosen functioning energies and the in-
puts/outputs of this component. As the aim is to manipulate an
object with the robot, inputs and outputs of the subsystem will
be, for the control part, the position and orientation of the object
in space relatively to the robot, and, for the guidance part, the
rotation, translation or allowance of degree of freedom orders.
Figure 18. MODELS FOR AN OBJECT MANIPULATOR
10 Copyright c
2009 by ASME
Figure 18 shows the models organised with one structural
choice of combination of the organs proposed previously. This
combination of organs proposes different structural possibilities
and, in further developments our aim will be to propose phys-
ical components “off-the-shelf” such as motors, which are spe-
cialized from converters components, processors, or mechanical
joints. This will again expand the tree of candidate solutions.
5 CONCLUSION
In this communication, we state that it is possible to enable
the automatic synthesis of conceptual design solutions of mul-
tidisciplinary systems such as robotic systems. We have shown
that it is possible to deduce physical components from the func-
tional description of a system. This process involves the use of
modern modeling language, SysML and a representation of the
necessary knowledge for synthesizing solutions, an ontology of
the synthesis in conceptual design. We have developed this on-
tology and it is coherent with other ontologies of engineering
design. We are able to suggest to designers the use of different
components enabling the same function.
Additionnally, we have developed a new model of
conceptual design based on Requirements-Function-Behavior-
Structures in order to perceive the forces and limitations of our
method. It appears that a hybridation of pre-synthesis and pre-
evaluation would be possible at the stage of generic structures,
in that sense, it would be possible to select embryos of solutions
more than others and this way we started to set heuristics for
cutting branches in the tree of possible solutions. Nevertheless,
besides the semantically irrelevant solutions, we chose, for the
moment, not to make any preliminary evaluation of the embryos
of solution because it is important for us to show to the design-
ers the variety of available solutions. Then, during the evaluation
phase, and according to multi-objective criteria of performance,
designers will be able to make their choice within a full range of
competitive concepts. In another hand, we are actually research-
ing on the possibility to derive the generic structures also partly
from the expected behavior. As such, does not an abstract organ
also contain intrinsic attributes in the form of generalized efforts
or flows?
Nevertheless, it is important to note that the publication of
our ontology will allow further development in the field knowl-
edge representation for engineering design. This fact, could al-
low a better interoperability between teams of designers as it
gives a conceptualization of the knowledge, concepts and their
relationships involved during conceptual design.
This paper has shown a review of latest improvements,
methods, models and languages in the field of Engineering De-
sign and more precisely conceptual design. Additionally, we
have shown here the improvements done thanks to model-driven
systems engineering over document-based engineering. We have
presented the developments of our computer application, OPAS,
a guide for designers during the synthesis of conceptual solu-
tions. Nevertheless, it is necessary for us to focus in the future
on empirical studies about the number of solutions synthesized,
first with small systems. Future work will be done on the com-
bination of our synthesis method with bottom-up methods and
design patterns in the aim for the automatic synthesis of bigger
and more complex solutions.
ACKNOWLEDGMENTS
The work of Franc¸ois Christophe is supported by a Doctoral
Scholarship from University of Technology (TKK). Franc¸ois
Christophe and ´
Eric Coatan´
ea are working for TKK HybLab
Project founded by MIDE foundation.
The work of Raivo Sell is supported by the Estonian Science
Foundation (grant No. 7542).
REFERENCES
[1] Lotter B., Manufacturing Assembly Handbook, Butter-
worths, Boston, 1986.
[2] Hsu W., Woon I.M.Y., Current Research in the Conceptual
Design of Mechanical Products, Computer-Aided Design,
Vol. 30, No. 5, pp. 377-389, 1998.
[3] Pahl G., Beitz W., Engineering design: a systematic ap-
proach, London: Springer, 1984.
[4] Altshuller G., Creativity as an exact science, Gordon &
Breach, Luxembourg, 1984.
[5] Gero JS., Design prototypes: A knowledge representation
scheme for design, AI Magazine 11(4): 26-36, 1990.
[6] Simon H. A., The Sciences of the Artificial, The MIT press,
1996.
[7] Coatan´
ea E., Alizon F., Christophe F., Yannou B., Selecting
technology alternatives for product families through tech-
nological coverage and functional verification, , In Pro-
ceedings of ASME design engineering technical confer-
ences DETC 2008, New-York, USA, 2008.
[8] Object Management Group,
http://www.omg.org/spec/SysML/1.1/, december, 2008.
[9] Umeda Y., Ishii M., Yoshioka M. and Tomiyama T., Sup-
porting Conceptual design based on the function-behavior-
state modeler., Artificial Intelligence for Engineering De-
sign, Analysis and Manufacturing, 10(4), pp. 275-288,
1996.
[10] Zhang W. Y., Tor S. B., Britton G. A. and Deng Y.-M.,
EFDEX: A Knowledge-Based Expert System for Functional
Design of Engineering Systems, Engineering with comput-
ers, Springer London, Vol. 17, No. 4, December, 2001.
[11] Tian Y.L., Zou H.J., Guo W.Z., An integrated knowl-
edge representation model for the computer-aided concep-
tual design of mechanisms, Int. J. Adv. Manuf. Technol.,
Springer-Verlag London Limited, May, 2006.
11 Copyright c
2009 by ASME
[12] Kitamura Y., Mizoguchi R., Ontology-based description of
functional design knowledge and its use in a functional way
server, Expert Systems with Applications, 24, 153-166,
2003.
[13] Gruber T. R., Toward principles for the design of ontologies
used for knowledge sharing. Presented at the Padua work-
shop on Formal Ontology, March 1993, later published in
International Journal of Human-Computer Studies, Vol. 43,
Issues 4-5, November 1995, pp. 907-928.
[14] Studer R., Benjamins V. R. and Fensel D., Knowledge engi-
neering: Principles and methods, Data & Knowledge En-
gineering Volume 25, Issues 1-2, Pages 161-197, March,
1998.
[15] Sowa J.F., Knowledge Representation: Logical, Philosoph-
ical, and Computational Foundations Pacific Grove, CA:
Brooks/Cole, xiv+594 pp; hardbound, ISBN 0-534-94965-
7, 2000.
[16] Berners-Lee T., Relational Databases on the Semantic Web,
http://www. w3.org/Design Issues/RDB-RDF.html, 1998.
[17] Berners-Lee T., Hendler J. and Lassila O.,
The Semantic Web, Scientific American,
http://www.sciam.com/article.cfm?articleID=00048144 -
10D2-1C70-84A9809EC588EF21, May 2001.
[18] Kitamura Y., Takafuji S., Mizoguchi R., Towards A Refer-
ence Ontology for Functional Knowledge Interoperability,
In Proc. of ASME IDETC/CIE 2007, DETC2007-35373,
2007.
[19] Liu Y.C., Chakrabarti A., and Bligh T.P., Further Develop-
ments of FuncSION , Artificial Intelligence in Design’00,
J.S. Gero (Ed.), pp- 499-519, Kluwer Academic Publish-
ers, 2000.
[20] Chakrabarti A. and Tang M.X. Generating Conceptual So-
lutions on FuncSION: Evolution of a Functional Synthe-
siser, Artificial Intelligence in Design ’96, J.S. Gero and F.
Sudweeks (Eds.), pp-603-622, Kluwer Academic Publish-
ers, 1996.
[21] Tudorache T., Employing Ontologies for an Improved De-
velopment Process in Collaborative Engineering, Doctoral
thesis, Berlin, November, 2006.
[22] Coatan´
ea E., Conceptual modeling of life-cycle design,
Doctoral thesis, Helsinki, 2005.
[23] Christophe F., Sell R., Coatan ´
ea E., Conceptual design
framework supported by dimensional analysis and Sys-
tem Modelling Language, Estonian Journal of Engineering,
14(4), pp.303 - 316, 2008.
[24] Sell R., Coatan´
ea E., Christophe F., Important Aspects of
Early Design in Mechatronic, Proceedings of the 6th inter-
national conference of DAAAM Baltic industrial engineer-
ing, pp.177 - 182, Kttner R. (Ed.), Tallinn, Estonia, 24-26th
april, 2008.
[25] Chiron F. and Kouiss K., Design of IEC 61131-3 Function
Blocks using SysML, Mediterranean Conference on Con-
trol and Automation, Athens, Greece, July 27-29, 2007.
[26] Hirtz J., Stone R.B., McAdams D.A., Szykman S, Wood
K.L, A Functional Basis for Engineering Design: Reconcil-
ing and Evolving Previous Efforts, Research in Engineering
Design, 13:65-82, 2002.
[27] Saeema A., Encouraging reuse of design knowledge: a
method to index knowledge, Design Studies, Volume 26,
Issue 6, Pages 565-592, Elsevier, November, 2005.
[28] Ji H., Ploux S. and Wehrli E., Lexical Knowledge Repre-
sentation with Contexonyms, In Proceedings of the 9th MT
summit, 194-201, 2003.
12 Copyright c
2009 by ASME
