A Methodology for Optimum Sensor Localization/Selection in Fault Diagnosis

This paper introduces a methodology for deciding the type, number, and location of sensors required to monitor accurately and robustly fault indications or signatures in a critical military or industrial system. Sensor localization/selection is an integral component in the design of modern condition based maintenance systems. Unfortunately, it has received minimal attention in the past, and most of the published work focuses on qualitative analysis tools. In our approach, a fault detectability metric is defined quantitatively, which expresses the capability of a sensor to detect a fault; a novel graph-based technique, called quantified-directed-model, is called upon to model fault propagation from one component or subsystem to the next of a complex large-scale system; and an appropriate figure of merit is defined in order to maximize fault detectability and minimize the required number of sensors while achieving optimum sensor placement. The proposed sensor localization/selection strategy is integrated into a diagnostic/prognostic system architecture running in realtime with actual or simulated data. The performance of the proposed strategy is tested and validated with a five-tank system.