Study on the Optimal Feature Number in the Induction Motor Fault Diagnosis Based on Support Vector Machine Using Current and Vibration Signals

Number of features used in induction motor faults diagnosis based on machine learning model affect the accuracy of the diagnosis. Not all statistical features can accurately represent the faults of the motor. Too many statistical features may lead to a high computational load and its complexity. The aim of this paper is to study the optimal number of feature used in diagnosing inductionInduction motor faults based on Support Vector Machine (SVM). In this study, current and vibration signal data from an inductionInduction motor operating at a speed of 1499 RPM were employed. Data preprocessing was performed using the Variational Mode Decomposition (VMD) method. Subsequently, statistical features in time domain are evaluated in stages, starting with the use of a total of 4 features and increasing to a total of 8 features. Feature selection was then carried out using Principal Component Analysis (PCA) method. To diagnose the conditions of the inductionInduction motor, various kernel types are examined in the SVM classification models. The results show that the variation in the number of features in the current signal consistently produces the same and high classification accuracies. Meanwhile, the vibration signal demonstrated optimal results when utilizing 6 features and the Fine Gaussian Kernel function.