Information Theoretic Approach to Reliability Based Sensor Placement for Fault Detection and Diagnosis

Abstract

In sensor placement literature, reliability has been considered to be a point value, corresponding to a specific time. Thus, sensor placement obtained by maximizing reliability changes if time were changed. We propose a cumulative residual Kullback-Leibler divergence based sensor placement formulation for fault detection and diagnosis, which explicitly incorporates time. The formulation also provides an opportunity to the end-user to specify the target performance for sensor placement. Further, we use a greedy algorithm to solve proposed sensor placement design problem. We illustrate the applicability of the approach using a benchmark case study.