Diagnóstico de Falhas em Motores de Indução Trifásicos com Aplicação de Sistemas de Informação e Redes Neurais Artificiais em Sinais Submetidos à Transformação de Clarke

Abstract

This study presents a system for short-circuit fault diagnosis in three-phase induction motors directly connected to the grid. The methodology is based on delayed mutual information in order to extract relevant characteristics relating electrical current signals in α−β orthogonal reference system. Those data are submitted to a multilayer perceptron artificial neural network which performs the pattern classification. Tests considering several operation conditions validate the robustness and accuracy of the proposed methodology.