115
Forum on Fuzziness
Applications of a Universal Expert System
in Industry *
M. Dohnal
Dept. of Chemical Engineering, Technical University of Brno,
Kravi Hora 2, 602 O0 Brno, Czechoslovakia
The theory of artificial intelligence has reached a level
which enables practical applications. As a consequence of this
several expert systems (corrosion, rubber science etc.) already
exist. However, a development of such expert systems is time-
consuming and requires highly sophisticated software. There-
fore a fuzzy based expert system SENECA is proposed. A fuzzy
simulation program is available so it is relatively easy to
develop the expert system based on a concept of fuzzy similar-
ity. An expert base of SENECA can absorb very heterogenous,
partially inconsistent and incomplete data of different range
and accuracy. The application of SENECA is demonstrated on
a hypothetical chemical reactor with two independent variables.
Keywords: expert system, artificial intelligence, fuzzy simula-
tion, fuzzy sets, ill-known system, engineering expe-
rience, linguistic quantification, SENECA, intelli-
gent user interface, corrosion, bio-engineering.
M. Dohnal was born in Brno,
Czechoslovakia, in 1943. He received
his Ing. and CSc degrees in Mechani-
cal Engineering from the Technical
University of Brno in 1965 and 1971,
respectively; and his DrSc degree in
System Engineering from the Techni-
cal University of Leuna-Merseburg,
GDR in 1983. In 1968 he joined the
Technical University of Brno, where
he is currently Associate Professor of
Control Engineering. From 1976 to
1977, he was with the Chemical En-
gineering Department at the University of Cambridge. From
1980-81, he was with the Department of System Engineering at
the Technical University of Leuna-Merseburg as a visiting
Associate Professor. His research interests include application
of Artificial Intelligence, and System and Chemical Engineer-
ing. He has written more than 100 papers on these subjects. He
was formerly employed as a Process Engineer with Chemopro-
jekt and Chepos, performing CAD and economical evaluation
of chemical and food plants.
* This is the second article in our series on "Fuzziness ". For
the first article see Vol. 6 Nr. 1 of the journal.
North-Holland
Computers in Industry 6 (1985) 115-121
1. Introduction
Primary information is such information which
is available at the very beginning of its analysis
(e.g. statistical [1], fuzzy [2]). It means that experi-
mental results (laboratory studies, pilot plants ex-
perimentation, full-scale operation) are not the
only source of primary information. Re-used in-
formation (e.g. literature) supplies primary infor-
mation provided that original experimental results
are not accessible directly.
Whether an engineer is concerned with design,
planning, operability or hazard analysis, operation
and control, predictions of the process behaviour
are needed. Therefore a suitable mathematical
model has to be developed. This is why primary
information is needed.
An information analysis of modelling confirms
that a limited extent of the primary information
set and the information loss during the formal
treatment (e.g. statistical analysis), are two main
bottlenecks on the way towards suitable models
[31.
2. Pre-Processing
A limited amount in a primary information set
and information loss during the formal treatment
can be significantly influenced by a suitable choice
of pre-processing mode. A pre-processing is the
formal treatment of primary data in order to in-
crease data adressability and instant applicability.
Addressability is a property of information items
to be stored within a given information structure
(data base). An instant applicability means a pos-
sibility to use the result of pre-processing for de-
sired purposes (optimization, hazard analysis etc.)
without any further formal treatment.
0166-3615/85/$3.30 © 1985 Elsevier Science Publishers B.V. (North-Holland)