ECAI-2000
home page
Workshop on
Knowledge-Based Systems for
Model-Based Engineering
22.08.2000
Call for discussion
The papers presented at this workshop illustrate,
to a certain extent, the state-of-the-art of knowledge based engineering
today. Many of them contain descriptions of research prototypes or of successfully
applied industrial cases in the field of model based diagnosis or engineering
design. Most of them concentrate on a specific problem (or problem class)
in the broad field of engineering applications, and follow a special approach
to model and solve this problem.
In order to provide a common discussion
platform that goes beyond very specific application or technical interests,
the workshop organizing committee asked the authors and the other workshop
participants to consider in their presentations / discussions the relation
between their past experiences and the way they see the future developments
in the field of knowledge-based engineering. The introduction “Knowledge-Based
Engineering: Where we are – Where we go?” by Rüdiger Klein (to
be found in the Workshop Notes) is meant to stimulate these discussions
and to suggest some common discussion topics.
Rüdiger Klein
Knowledge-Based Engineering: Where we are
– Where we go?
Information technologies have been of growing
importance in this context over the last decades. Currently, this process
is gaining dramatically speed and power in all areas. The main reason is
the “internet revolution”. The advent of internet and intranet technologies
changes the way information is exchanged and used: inside companies as
well as in their outer business. E-business and b2b commerce are only the
newest buzz words. They complete the more “old fashioned” (just a few years
old) slogans as concurrent engineering, computer supported cooperative
work, knowledge management, and such “dinos” as CAD, CAE, CAM, PDM, etc.
They
all will be influenced substantially by the new-coming technologies.
What is exchanged in this way is information
- but what is necessary is to share, communicate and use knowledge.
With current information technologies, the meaning of data is mainly left
to the human users. They are the (main) “sources” and “processors” of knowledge.
When computers are used to support this information processing this is
done mainly in a pre-defined and explicitly coded way. Automated knowledge
processing by knowledge based systems is still more an academic issue than
one of daily practice - though in the meantime a number of industrial knowledge
based systems have been implemented. Knowledge based systems in engineering
practice are still more an exception than the normal case. Such systems
are mostly restricted to relatively small and isolated applications, and
they need a large effort to get built and maintained.
This has to change – this is our confession
– if the “internet revolution” shall be mastered effectively and efficiently.
Knowledge
processing has to become an integrated part of information technologies.
Only in this way the new complexity of information in the internet
age can be successfully managed. This is especially true in engineering
with its complex and manifold tasks: engineering knowledge management,
requirements engineering, specification-driven design, design-to-X, RAMS-engineering
(reliability, availability, maintainability, safety), etc.
Knowledge based technologies have two main
aspects in engineering (and not only there):
-
knowledge modelling, and
-
problem solving.
Knowledge modelling is important as “backbone”
and “glue” of the different information systems used. The information processed
by these systems has explicitly to be modelled including the underlying
assumptions. This is essential for system interoperability, information
integration, intelligent information retrieval, knowledge warehouses, etc.
Ontologies emerged in recent years as a main field of knowledge modelling
especially dedicated to this task
But knowledge modelling has a second focus
– directly related to problem solving by knowledge based technologies.
Knowledge based systems can be seen as two separated but interacting and
interdependent components: the applied problem solving methods and the
corresponding knowledge base. Both have to correspond to each other: the
knowledge base has to be modelled according to the problem solving methods,
and the problem solving methods have to chosen according to the task at
hand (the problem solving competence) and the applied modelling assumptions
in the knowledge base.
Currently, this process of knowledge modelling
and knowledge based system design is a complicated and laborious process.
It needs special skills in knowledge based technologies as well as in the
intended domain. Due to missing comprehensive domain models and powerful
generic problem solving methods, various compromises and short cuts have
to be made during modelling (which normally requires a deeper understanding
of the domain). The resulting knowledge based systems are typically hard
to reuse or to extend.
Let’s describe the way out of this dilemma
as a vision of knowledge based engineering:
-
We have a library of comprehensive, generic,
application independent domain models including explicitly described
meta models, underlying assumptions, etc.; and
-
We have libraries of problem solving methods
each with a well formulated problem solving competence, and an explicit
description of the assumptions the corresponding knowledge base has to
observe.
Both knowledge libraries can be accessed through
the internet using intelligent retrieval, and knowledge based systems can
be configured automatically reusing such building blocks from the libraries.
In this way, much less effort will be
needed in order to build much more complex knowledge based systems.
Visions are valuable: because they allow
us to dream, they help us to get motivated, and they guide our search to
turn them into a reality. Visions become a danger if one forgets the difference
between dream and reality.
Both roles of this vision provide the
ground for our workshop: which are the challenges on our way to knowledge
based engineering, and what is realistic today or can become real on this
way within the near future.
Of course, the vision described above
is not restricted to engineering. But engineering has some special aspects.
They make it challenging to discuss knowledge modelling and problem solving
issues especially related to this domain. These aspects can be summarised
as follows:
-
Engineering knowledge tends to be quite complex:
physical laws, technical principles, geometric shapes and relations, norms
and directives, functions, structures, and behaviours etc. have to be modelled
with all their facets and interdependencies.
-
Though quite complex, engineering knowledge
is well structured and has a relatively clear meaning.
-
Knowledge based technologies will be important
in engineering – but they are not everything. To the contrary: key to the
success of knowledge based techniques in engineering is the integration
of these techniques with conventional information processing techniques:
geometric modelling, numerical simulation, etc. What is essential is that
this integration can not be a one way street: both sides have to move towards
each other.
-
Engineering knowledge is dealing to a large
extent with “object level” knowledge: system, component, and function descriptions,
physical laws, behaviours, etc. But due to its complexity, this engineering
knowledge is also highly problem solving related: how to attack
a given task, which way to choose, how to manage conflicts, etc. Engineering
methodology
is an essential and indispensable part of engineering knowledge because
this is the only way to manage the complexity in this domain.
The papers presented at this workshop are
typical examples of the state of the art of knowledge based engineering
today. They contain descriptions of successfully applied industrial cases
or of academic research systems. They deal with model based diagnosis or
with engineering design. Each of them concentrates on a special problem
or problem class in the broad field of engineering applications, and follows
a special approach to model this problem and the problem solving. There
is a large amount of similar research activities and results described
elsewhere dealing with special problems in engineering and special approaches
to them.
In order to provide a common platform
for the workshop, we suggest to focus the discussion onto the following
issues related to the vision of knowledge based engineering outlined above:
-
What were the main modelling principles
applied in each of the described systems: on the domain knowledge level
as well as on the side of problem solving methods? On which assumptions
are these models based?
-
What changes and/or extensions in the applied
modelling approaches would be necessary in order to come to more generic
models? Is that realistic? Do human engineers apply generic models
in order to solve their problems? If yes – how do these models look like;
if no – why not?
-
What should an engineering ontology
(or a system of engineering ontologies) look like in order to support generic
knowledge models in engineering domains?
-
How is problem solving controlled in engineering?
If there are more than one problem solving method applied in a system,
how is their interplay managed and controlled?
Of course, it is not an easy task to find
comprehensive answers to these questions. Probably, this will not be achievable
at this workshop alone. It will need time and much more substantial research
efforts. But we should find answers to these questions – for the benefit
of knowledge based engineering.
This is my personal view. Also counter
positions are, of course, very welcome – if anybody does not believe
in the visionary approach formulated above, or anybody has concrete experiences
in contradiction to what has been suggested.
May this workshop provide a good and stimulating
starting point for these discussions.
Download this position paper: PDF
(23 kB)
This page is
maintained by mst@dbai.tuwien.ac.at
and was last modified on 19/07/2000.