Fault Diagnosis of a Brushless DC Motor Using K-Nearest Neighbor Classification Technique with Discrete Wavelet Transform Feature Extraction

Diagnosis of motor faults plays an important role in the industry to implement preventive maintenance and avoid an unscheduled shutdown. A mechanical fault diagnostic system of a BLDC using the k-NN classification technique with DWT feature extraction is proposed in this study. Voltage signals from healthy and faulty BLDC motors were recorded. The voltage signals of the motors were then decomposed, and features were extracted using the DWT and divided into three data sets which are: training, validation, and testing. The k-NN prediction model was then trained, validated, and tested using a MATLAB environment. The diagnosis was run on 5 different Haar levels. Results showed that Level 1 produced the highest accuracy of 89.510% over the other 4 levels and was statistically verified using ANOVA Test.