MONET Workshop on Model-Based Systems at 16th European Conference on Artificial Intelligence, August 22-26, 2004, Valencia, Spain
Adequate sensors are a necessary condition for fault diagnosability. Sensor placement for diagnosis task is to study where to put the sensors so that they are the minimal set to diagnose certain faults. This paper presents a method of sensor placement based on diagnosability analysis using the simulation model in the CAD environment. The fault signature matrix is determined by the projections of different operation modes on observable variables. The minimal sensor set for detecting faults and for discriminating the faults can be computed from the fault signature matrix. We also consider that values of exogenous variables are a condition for diagnosability. By introducing the concept of virtual sensors, faults can be detectable/ discriminable based on their signatures on virtual sensors. The advantages of this approach are that not only the minimal sensor set but also the conditions of causal scopes are obtained and the procedure is fully automated.
MONET Workshop on Model-Based Systems at 16th European Conference on Artificial Intelligence [Proceedings].