Fault Diagnosis for a Petri net class using reduced observable projections

This paper deals with fault diagnosis in discrete event systems modeled with partially observed Petri nets. The approach herein presented removes unnecessary information from the incidence matrix by building a reduced observable projection net using the observable complement subspace of the original net. Using this observable projection net and an off-line algorithm, a structural characterization of the diagnosability of Petri nets is proposed. The reduced observable projection net is also used to design a fault diagnoser. An illustrative example is presented to show the effectiveness of the results.