Knowledge acquisition for expert systems in fibre production
The aim of the study described in this thesis is to investigate the application of expert system technology to acrylic fibre production, with a particular emphasis on knowledge acquisition requirements. In doing so, it is intended to provide an understanding of the requirements and appropriate techniques for the effective application of expert systems in the process industries; The scope of the study is limited to process fault administration which involves detecting, diagnosing arid correcting abnormalities in process operation. A methodology is provided for the systematic development of expert systems within the defined area of application. An important phase in the development methodology is that of expert system specification. This involves an analysis of expert behaviour and the specification of expert system functionality: In order to assist system specification, generic knowledge types and human expert activities have been identified within the context of process fault administration. Knowledge acquisition is discussed in terms of the requirements during each phase of the proposed development methodology. A detailed review is given of the available techniques for knowledge acquisition and an assessment is presented of the most appropriate techniques to apply during each phase of the methodology. A new knowledge acquisition technique is described. The technique is designed to record knowledge of process operation and process fault diagnosis. It is based on a hierarchical decomposition of the process in terms of process objectives. Two complementary forms of knowledge representation are produced: a hierarchy diagram which shows the dependency relationships between individual process objectives and a task statement table which provides a more. detailed explanation of the objectives. Finally, three. case studies are described in which the techniques described in the thesis were applied and developed.