Real-time Bearing fault detection using Intelligent Algorithm combined with Wavelet Transform

Abstract

The monitoring of bearings in electromechanical induction machines has become in the last decades a field where a lot of research is invested. We can understand this because of their responsibility for the defects of these machines and the enormous losses and expenses that they generate. However, if it is true that many methods have been proposed, it must be said that a large part of them, although based on the stator current, does not focus on the only defect of the bearing as we propose through this work. In this study, our goal is to analyze the current induced in a copper coil placed side by side with magnets placed in the inner ring of the bearing, by the Machine Current Signature Analysis method in order to detect from the genesis any defect of the bearing. To do this we will apply algorithms such as Cuckoo Search-Support Vector machine, Convolutional Neural Network, Kernel-Principal Component Analysis, Recurrent Neural Network, and Support Vector Machine on the data extracted from this induced current using the Wavelet Transform technique in order to determine among these algorithms, which one would allow to detect in real time and with a good precision the defect of bearing when it occurs during the operation of the machine that incorporates it. The whole with an aim of developing, in combination with the embedded electronics an autonomous electronic system of detection of the defect of bearing in real time in an effective way.