Diagnosing of BLDC Motor Faults based on LSSVM Model and Vibration Signal

Abstract

A BLDC motor is commonly used as the driver of an electric vehicle. So that this part becomes a critical component in the electric vehicle system. Any faults in the motor can cause the vehicle to not operate. Early detection of motor faults can avoid sudden motor failure. This paper aims to diagnose the possible faults in a BLDC motor using the least squares support vector (LSSVM) model. In this paper, the motor in normal condition and the motor with bearing, unbalance, and stator faults are examined. The vibration signals are measured from the BLDC motor operating at 430 rpm. The signals are captured at a 20 kHz sampling rate. The signals are smoothed using a moving average filter. The feature selection is based on the ability to segregate the different fault conditions through visual observation. The kurtosis and frequency centre value features are selected as fault predictors. The diagnosis process is performed by the classification of motor conditions using the LSSVM model. The model is built from the training data. The result shows that the LSSVM model performs very well in diagnosing BLDC motor faults. The diagnosis accuracy is 100%, both for training and testing data.