Set theoretic based neural-fuzzy motor fault detector

Abstract

The usual motor incipient fault detection procedures require engineers and researchers to devote a significant amount of time and energy to investigate the motor system they are working with. This paper presents a set theoretic approach that provides a systematic way to formulate and incorporate information into the motor fault detection framework. Based on this set theoretic formulation, a heuristically constrained neural/fuzzy system is then used to learn the exact input/output relation of the fault detection process for a specific motor using measured data. This system is able to provide updated membership functions of the sets which better describe the fault detection problem. To illustrate their proposed methodology, a three-phase induction motor exposed to changing external factors is used for the detection of a friction fault.