Designing a sensor placement for diagnosability using structural properties

Adequate sensor placement is a necessary condition for efficient fault diagnosis systems. Sensor placement for diagnosability consists in studying where to place sensors to satisfy diagnosability specifications. This paper presents a method based on a structural analysis that takes into account diagnosability specifications. It is based on the study of properties of structural matrices and on the Dulmage-Mendelshon decomposition. The proposed approach manages constraints modeled by linear or nonlinear equations, qualitative relations,... and it does not require the design of analytical redundancy relations (ARRs) or the design of testable subsystems (TSSs). Assuming that a component is modeled by one constraint and that the cost of the measurement of each variable is defined, the proposed approach yields the best sensor placement according to a diagnosability criteria. An application example dealing with a dynamic linear system is presented.