Use this URL to cite or link to this record in EThOS:
Title: Ensuring the real-time behaviour of expert systems
Author: Goodwin, J.
Awarding Body: University of Wales Swansea
Current Institution: Swansea University
Date of Award: 1997
Availability of Full Text:
Access from EThOS:
As computers become more involved in the supervision and control of industrial processes, so the risk and potential damage that the failure of such systems can have increases. The use of Artificial Intelligence (AI)-based techniques in industrial applications is becoming increasingly popular, especially those based on expert systems, which attempt to reproduce a human's expertise. This thesis suggests that system verification must be carried out before such a system is allowed to control a real process, and that effective verification requires the creation of formal, mathematical models of system performance. These models need to cover both the functional and the temporal behaviour of the system. The focus of this study is the capturing of the real-time "knowledge" necessary to create the temporal model. Whereas much work has been done on requirements capture, this has largely focused on the functional domain. It is argued here that the temporal domain is of equal importance, and that a full temporal model, showing the interaction between the AI-based control system and the process being controlled, is mandatory. If an expert system is to be used within a real-time environment, then that expert system must meet the rigorous timing constraints imposed by such an environment. This work investigates such usage of expert systems, and suggests the criteria that they must meet. Of particular importance is that the inherent structure of the expert system, and its processing engine, must ensure predictable and timely performance. The work culminates in the development of a new designer's tool, called the Temporal Computer Aided Knowledge Engineering (T-CAKE) tool. Its aim is to permit a plant engineer with little, or no, knowledge of real-time systems theory to create a temporal model of the plant, and its control system, suitable for the rigorous and formal analysis needed to prove that the overall system can meet all the timing constraints imposed upon it. The use of T-CAKE is demonstrated via an expert-system application designed to control an experimental pulsed column rig. The pulsed column here forms part of a solvent-extraction process used in the nuclear industry to reprocess nuclear waste. The work demonstrates that it is possible to assist a designer in the creation and formal analysis of a temporal model of a plant and its interactions with an AI-based control system. However, it is concluded that despite the original goals, the task of interpreting the model is of such difficulty that it cannot be recommended that non-experts should attempt to create AI-based control systems for real-time systems.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available