A Comprehensive Framework for Model Based Diagnosis

This paper reviews our work on monitoring prediction and fault isolation methods for complex dynamic sys tems a ected by abrupt faults The key to this work has been our ability to model the transient behavior in response to these faults in a qualitative framework where the predicted transient e ects of hypothesized faults are captured in the form of signatures that spec ify future behavior for the fault with higher order time derivatives The dynamic e ects of faults are analyzed by a progressive monitoring scheme that avoids direct measurement of second and higher order derivatives from real signal values which are unreliable when sig nals are noisy However generating qualitative charac teristics from real time varying signals is still a chal lenging problem More recently we have been investi gating statistical techniques for reliable change detec tion and labeling in noisy signals We discuss some of these techniques and study the tradeo between speed of detection and sensitivity of analysis The integrated framework for monitoring prediction and fault isola tion is being tested on real data generated from an automobile engine test bed