A Projection-Based Support Vector Machine Algorithm for Induction Motors’ Bearing Fault Detection

This paper proposes a binary fault detection algorithm for detecting inner raceway bearing faults in a 4KW induction motor. The algorithm uses Support Vector Machine (SVM) and Projection Recurrent Neural Network (PRNN) techniques and is based on data collected experimentally at different speeds and load conditions. Time and frequency contents of the three-phase stator currents are analysed using Discrete Wavelet Transform (DWT), Power Spectral Density (PSD), and cepstrum analysis. A feature set is obtained using various statistical measures, and feature selection algorithms are used to select the most relevant features. The SVM is then trained using these features, and its optimisation problem is formulated as Constrained Nonlinear Programming (NCP). A PRNN is proposed to solve the NCP and obtain the optimal decision boundary of the SVM. The study demonstrates that the accuracy of the algorithm depends on the type of kernel function and the number of relevant features selected. The results suggest that the proposed algorithm is effective in detecting inner raceway bearing faults in induction motors.