Reference: C. Wood and P. Clark. FADES: An Expert System for Fault Analysis and Diagnosis. TIRM 87-024, Turing Institute, 1987.
Abstract: As the power and complexity of modern processing systems increases, both the urgency and difficulty of accurate, rapid fault detection in such systems rises also. Such systems need to be even more carefully designed so that operators can quickly, easily and unambiguously locate faults and take corrective action, in order to avert possible catastrophies.
This report presents FADES, a fast and accurate designer's tool for fault diagnosis and providing corrective advice in a simulated complex fluid system. It aims to help designers improve the "human window" of transparant communication between controllers and system. FADES incorporates a qualitative model of the fluid system, in whihc examples of sets of failures can be simulated and their effects propogated through the system by a set of behaviour rules.
After synthesizing examples of failures during fault analysis, a rule set for diagnosing failure type is generated by an induction module, based on the AQR induction system and extended to predict disjuncts of failures where a single failure cannot be uniquely determined from observable attributes. Such rules indicate to designers where bottlenecks of complex or ambiguous fault detection exist, and can also be incorporated in an advisor module for providing corrective advice for operators in the event of fluid system failure.