An Embedded System for Stator Short-Circuit Diagnosis in Three-Phase Induction Motors Using Information Theory and Artificial Neural Networks

Abstract

This study presents an embedded system in hardware based on mutual information measurements and artificial neural networks for the stator winding short-circuit diagnosis of three-phase induction motors (TIMs) with a line-connected sinusoidal power supply. The methodology employs an information theory measure to extract the most relevant characteristics of the current signals of TIM phases A and B. These data are presented to a multilayer perceptron neural network that performs the pattern classification. Experimental tests with different machine operating conditions validate the robustness and efficiency of the proposed methodology.