Estudo comparativo de ferramentas estatísticas em falhas de máquinas elétricas

Abstract

This work presents a comparison between methodologies used to detect and classify faults in tree-phase induction motors driven by frequency inverters. The applied method was the analysis of the current signals with stator, rotor and bearing faults, besides current signals of healthy motor. Such analysis was accomplished thought the use of Fast Fourier Transform, Wavelet Transform and Principal Component Analysis as signals processing techniques. In order to identify the presence of motor faults and to classify their severity two intelligent networks were used: Multi-layer Perceptron and Fuzzy ARTmap.