Optimal sensor network for fault diagnosis using structural analysis

Abstract

This paper addresses the problem of instrumentation for fault detection and isolation in structured systems. The design of sensor networks is achieved by using structural analysis. This problem of sensor network design is formulated as nonlinear binary optimization problem. If the FDI problem is not solvable for an instrumented system, the strategy determines the additional sensors to be used to guarantee faults diagnosability. In addition, it is shown how to design a sensor network for non-instrumented systems while optimizing the number of sensors, the cost or the reliability of the network.