Design and implementation of automotive model-based diagnosis using the Echidna constraint reasoning system
Afwarman Manaf · Summit (Simon Fraser University) · 1991
Developing engineering applications using expert systems has been an important driving force for research in expert system technology.The automotive diagnosis problem is one of the most challenging domain applications.Existing automotive diagnosis systems which rely solely on rule-based knowledge inherit many limitations.Model-based diagnosis is an area of active research which overcomes these limitations since it uses knowledge from first principles.This thesis provides a new model-based diagnosis system for automotive systems.Our architecture uses a deductive model as the main part and diagnostic rules to increase the speed of diagnosis processes.We use both abductive and deductive reasoning t o provide a more powerful diagnostic system.Both model-based techniques and heuristic knowledge are integrated in the system.We claim that our model makes a significant step towards the use of a generic model to solve problems in many applications.The model integrates both correct and faulty behaviour using first-order logic and has a high degree of readibility and modifiability.We also use constraint satisfaction techniques that have been shown t o increase efficiency in A1 applications.We have developed a working prototype of a model-based diagnostic system.The system generates the list of faulty components based on observational input supplied to it.We use the Echidna reasoning engine, a new expert system shell being developed in the SFU Expert System Laboratory.It is my pleasure to acknowledge the many people involved through the completion of this thesis.My senior-supervisor, Dr. Bill Havens, provided many hours, insight, discussion and encouragement during our weekly meetings.His support and help were always there when I needed them.Dr. John Jones helped me t o understand the other side of the issues I was tackling.He also patiently helped and corrected me in writing this thesis.Dr. Lou Hafer provided much encouragement.I especially want to acknowledge Charlie Hunter who deserves credit.He helped me in understanding the domain of my work and was very cooperative when I was building the model designed.I would like to thank to Stefan Joseph and people in McCarney Technolgy