Diagnosis of Bearing Failure in BLDC Motors with Machine Learning Method

BLDC motor failures, which are widely used in industry, can cause work disruption and occupational accidents. The aim of this study is to provide early diagnosis of bearing failures in BLDC motors, thus increasing the efficiency of the motors and preventing unplanned outages. Traditional fault detection methods are usually based on periodic maintenance and manual inspections, which can lead to high costs and production losses due to delayed or inadequate interventions. In this study, a machine learning based system has been developed to classify bearing failures. A large data set consisting of 10,000 samples using vibration, noise, and speed data was evaluated with 33 different algorithms in the Matlab program. In the test phase, 15% of the data set was used for validation with data that the model had never seen before. The highest accuracy rate was obtained in Quadratic Discriminant Algorithm (QDA) with 96.7%. Although other algorithms also showed high success, it was determined that QDA was the method that provided the best discrimination between classes. The continued success of the model on the never-seen-before data used in the test phase shows that the method performs reliably and stably in a wide range of applications.