AbstractIndustrial processes can often neither be modelled analytically nor identified experimentally with respect to all their nonlinear properties. For control engineering purposes, real processes are therefore approximated with models of relatively low complexity. Although such approximations are normally sufficient, it is often helpful to be aware of the variety of possible nonlinear effects to improve the design of control systems. Knowledge about these influences is frequently part of the industrial engineer's experience from working with the processes. At present, however, this experience can rarely be used systematically for control engineering purposes, since these process experts are in general not skilled in process modelling or control.
The subject of this work is the development of an approach that facilitates the systematic acquisition of the process expert's (or "area engineer's") experience in order to derive process models, which are possibly very nonlinear and/or multivariable or only partially defined, depending on the available information. This knowledge acquisition approach addresses therefore the above situation in that it serves as a new means of communicating the area engineer's process knowledge to the control experts. Furthermore, the area engineers themselves are able to use their experience in a systematic fashion for control engineering purposes. This latter aspect is addressed within a collaborative research project between the University of Glamorgan and the Fachhochschule Hannover, of which the work presented in this thesis forms the first part. Overall, the collaborative project aims at making Computer Aided Control System Design (CACSD) approaches accessible for engineers with little or no experience in control engineering.
A novel fuzzy hybrid approach for the simplified modelling of nonlinear multivariable dynamic processes is the first main contribution of this work. It builds on partial information about the global system behaviour in the form of locally valid single variable transfer functions. This important contribution forms an integral part of the knowledge acquisition approach to process modelling, the overall subject of this work. The systematic and structured knowledge acquisition procedure as such is the second major contribution. It facilitates the build-up, storage and flexible re-use of knowledge about static and dynamic system properties. In particular the above mentioned fuzzy hybrid models can be automatically generated through the prototype implementation of this knowledge engineering approach, without requiring any experience in modelling or fuzzy logic from the user. The third main contribution is a new modelling approach based on descriptive attributes referring to the process behaviour. This approach, which forms also an integral part of the knowledge acquisition procedure, facilitates the use of particularly abstract and not quantifiable information.
|Date of Award||Mar 1997|