A Systematic Analysis of Measurement Selection Algorithms for Fault Isolation in Dynamic Systems

Abstract

Fault detection and isolation in dynamic physical systems based on analysis of transient system behavior involves monitoring for deviations and linking observed deviations to changes in system parameters. Components linked to the changed parameter values may be faulty. Functional redundancy is exploited in choosing measurements to uniquely isolate all faults. Measurement selection algorithms determine the minimum set of system variables that can uniquely identify all single faults in the system. This paper analyzes a set of measurement selection algorithms in the context of transient analysis in dynamic systems. Results show that an A*-search algorithm provides good results while considerably reducing the computational complexity of the measurement selection process.