Classificação de Falhas em Motor de Indução Utilizando Curvas Principais

Abstract

: Electric motors are very important machines in any industrial plant, due to your large range of uses and robustness. Evaluate the conditions of this machines is crucial to ensure the operation with security and quality. This paper presents a methodology of fault classification using vibration analysis, the vibration signals were measured by a 3-axis accelerometer linked to un Arduino microcontroller. The front bearing was evaluated in good conditions and in two failures situations in its races. The feature extraction was realized by Higher-Order Statistics using cumulants of 2nd and 4th orders with zero lag and the classification was made using Principal Curves. Principal Curves are obtained to each motor condition and realized the classification by measure of the distances of each event to the curve. Classification results were obtained with hits rates above 95%.