Summary and applicability of analog fault detection/isolation techniques

Abstract

A survey of common analog fault diagnostic techniques is provided. The focus is on providing information to facilitate the use of viable techniques suited for a problem definition. The level of understanding of the analog system is correlated with the type of techniques best suited for the diagnostic system. Reviewed are techniques from Control Theory, Probability Theory, Computational Expert Systems, and Computational Artificial intelligence. The Control Theory techniques are suited best for system diagnostic problems where the system model is accurately understood. Techniques that employ Probability Theory are valuable for addressing uncertainty that arises in diagnostics of systems affected by noise. Computational Expert Systems address the problem of diagnostics by creating a data situations where the diagnostic process is accurately understood. Computational Artificial Intelligence techniques are presented as being best suited for systems where little reliable knowledge is known. The analysis does not preclude the use of any technique, but rather addresses efficiency of application.