Abstract
In developing knowledge-based fuzzy systems it is
possible to utilize several different development
processes. In this thesis the incremental development
process, where systems are analyzed, designed, installed
and tested in parts, is seen as the most suitable one.
Building a knowledge-based system is a human-oriented
interactive process, that is disturbed by biases,
mistakes, misunderstandings, and different levels of
expertise. Therefore the acquired information should be
validated before using it in any installed expert
(knowledge-based) system. In this thesis we present
methods to validate acquired information by combining
heuristic knowledge and measured knowledge (data). Fuzzy
models are built using heuristic knowledge and data is
used to finetune parameters of the models. The validity
of acquired knowledge can be estimated with analyzing the
tuning results.
An essential point of this thesis is that fuzzy models,
developed in the knowledge acquisition phase, can be
utilized later during the development work. The main
areas concerned are model-based control and diagnostic
systems. By using experimental results we present methods
to construct model-based control systems and fault
diagnosis systems by using the incremental development
process, fuzzy models, modular architecture, and advanced
techniques. The results show that incremental development
process can be applied also when model-based systems are
built. In fault diagnosis, fuzzy modeling gives the
possibilities to increase the level of adaptivity and
robustness in the future, after reliable tuning and
learning methods have fulfilled demands. In this thesis
we present new methods in applying fuzzy logic and fuzzy
model-based systems in developing fault diagnosis
systems.
More intelligent control systems are seen as one
important factor in the future. When the complexity of
machines and processes increases, manual operation of
them becomes difficult, even impossible. This has to be
taken into account in developing control systems to
ensure the high availability of target systems. Our
solution, presented in this research, is to use the
information produced by a fault diagnosis system in
adapting the control system. This makes the control
system more robust, and increases the availability of a
target system. The solution is based on using modular
architecture, in which meta-rule modules are exploited to
produce the adaptivity needed. The solution is formed so
that fault diagnosis information can be installed later,
after experiences of using the target system have been
gained.
Original language | English |
---|---|
Qualification | Doctor Degree |
Awarding Institution |
|
Place of Publication | Espoo |
Publisher | |
Print ISBNs | 951-38-5367-5 |
Publication status | Published - 1999 |
MoE publication type | G5 Doctoral dissertation (article) |
Keywords
- fuzzy logic
- knowledge engineering
- model-based fault diagnosis
- computer programs
- embedded systems