‘black boxes’, since the applied methods are poorly
documented and the source code is inexpressive.
This has a negative effect on the maintainability and
the extendibility of such applications.
In this paper an ontology, a structure of the
existing knowledge categories and their relations, is
proposed to support the implementation of a second
generation KBE systems in industry. The ontology
for engineering knowledge is geared towards routine
design processes. It entails a categorisation of
knowledge elements into product versus process
knowledge and domain-specific versus problem-
specific knowledge. The proposed ontology aims to
achieve better use of intellectual human resources as
tangible asset: it enables the reuse of domain
knowledge across multiple design problems as well
as sharing problem related knowledge across
multiple application fields. Repositories of
knowledge based on the proposed ontology provide
a platform to rapidly develop design models for
KBE applications, reusing the knowledge already
captured and formalised.
2 ENGINEERING DESIGN
Engineering design can be considered a deliberate
search problem in a solution space for artefacts that
satisfy functional needs within a set of constraints.
The solution space or design space can be
considered a range of available components and a
set of relations between those components in order
to form artefacts. Although the solution space can
encompass an infinite number of solutions, in
general only a small number of artefacts form
feasible and satisfying, not to mention optimal,
solutions. Such a solution to the design problem
entails a collection of components and their
relations, that together provide a complete
specification of the system that delivers the
requested functions and satisfies the constraints
(Chandrasekaran, 1990).
2.1 Design Problem Categorisation
Design problem solving can be divided into routine
and non-routine design activities, based on the
identification and availability of the knowledge
involved in the design process. Brown and
Chandrasekaran identified three classes of design,
related to the level of ‘routineness’ (Brown, 1989):
For the first class, Class 1 Design, neither all
possible decompositions of the artefact nor the
approach to solve the design problem is
known in advance.
For Class 2 Design, the possible
configurations and components are known,
however the problem solving strategy is not.
For the third class, Class 3 Design, all possible
configurations, components and design
variables are known. Furthermore, the
problem solving approach is acknowledged,
resulting in the availability of so-called design
plans.
The three abovementioned classes of design
correspond to the general acknowledged categories
of design into routine design, innovative design and
creative design as introduced by Gero (Gero, 1993).
Here, routine design concern designs that fit within
the space of previous solutions. Therefore in routine
design the components and their variables, the
constraints for those variables and the type of
requirements are known in advance. Innovative
designs are based on known design options, however
the applicable range of values for the variables is
extended. Creative design involves the definition of
new components, variables or relations between
components. Here, neither all components nor the
problem solving strategy is known in advance.
Innovative and creative design together make up the
non-routine design problems. Typically, a
development process for a new product will
encompass routine, innovative and creative design
activities. The remainder of this paper will focus on
routine design problems.
2.2 Routine Design Problems
Judged by the type of components and the assembly
of the artefact, different dimensions of routine
design problems can be distinguished (Wielinga,
1997). Figure 1 summarizes the most common types
of routine design problems. It should not be
considered an attempt to provide a comprehensive
overview of all routine design problems.
Starting with the most basic form of routine
design, verification problems aim to confirm the
validity of a synthesized artefact, where both the
assembly as well as the components are predefined.
Assignment problems deal with matching sets of
resources (components in product design) with a
fixed collection of subjects, defined as the assembly
outline or skeleton. Example problems are the
assignment of airplanes to terminal gates or the
assigned queuing of travellers at airport security
gates. Lay-out design or scheduling problems also
involve fixed sets of components, but the outline nor
composition of the artefact is known in advance. An
example is the arrangement of machines for a
factory layout. Parametric design assumes a
KEOD 2009 - International Conference on Knowledge Engineering and Ontology Development
262