An Improved Stator Winding Short-circuit Fault Diagnosis using AdaBoost Algorithm

Brushless DC (BLDC) motors, bearing the characteristics of permanent magnet synchronous machines, have gained immense popularity in industrial applications due to its excellent efficiency and ease in control over conventional DC motors. Stator related faults are the most common types of faults in BLDC motors while operating under a higher loading and complex condition. Conventional machine learning (ML) classifiers such as- Support Vector Machines (SVM), k-nearest neighbors (KNN), Naïve Bayes (NB) classifiers fail to obtain optimum accuracy when there are abrupt changes in health states of the BLDC motor. Especially the classification of weak health features ends up with an erroneous result. Boosting techniques allow adding weights to weak features and a better result in classification. This study presents an improved stator related fault classification of BLDC Motor using AdaBoost technique. Several parameters of AdaBoost algorithm are optimized and implemented along with Random Forest (RF) classifier in order to determine a developed algorithm for fault classification.