An Approach to Fault Diagnosis Using Meta-Heuristics: a New Variant of the Differential Evolution Algorithm

Abstract

This paper presents an application of meta-heuristics to fault diagnosis. The idea behind this application is to develop methods for fault diagnosis that should be robust, sensitive and with an adequate computational cost. Applications of meta-heuristics are possible based on the formulation of fault diagnosis as an optimization problem. The results indicate the suitability of the use of meta-heuristics for fault diagnosis. In particular, this study shows an application of meta-heuristic termed Differential Evolution to diagnosing a DC Motor benchmark. This allowed developing a new variant of Differential Evolution, namely, Differential Evolution with Particle Collision. This new algorithm was validated with some benchmark functions for continuous optimization, showing that it over-performed the behavior of Differential Evolution.