Performance analysis of an evolutionary algorithm for fault detection in t-diagnosable multi-processor systems

Abstract

In this paper, we present a performance analysis of an evolutionary approach for fault identification in t-diagnosable systems, i.e. systems in which the set of t permanently faulty units could be unambiguously identified. The proposed algorithm is based on evolutionary/genetic computing and is shown to correctly identify the set of faulty units. Simulation results are more than encouraging and should stimulate future research.