Acoustic Fault Detection in Brushless Direct Current Motors Using Machine Learning for Predictive Maintenance

Performance maintenance of Brushless Direct Current (BLDC) motors is crucial due to their widespread applications, ranging from household appliances like fans to critical systems in unmanned aerial vehicles (UAVs). As these motors are integral to various operations, ensuring their reliability and detecting faults in an early stage is essential to prevent failures and maintain optimal performance. Numerous studies have been conducted to develop methods for fault detection in BLDC motors, focusing on different aspects such as vibration analysis, current monitoring, and sound analysis. This paper proposes a reliable approach for detecting faults by developing a system that identifies bearing and propeller faults based on the audio analysis of the operational sound of the BLDC motor. Acoustic features were meticulously extracted from the motor's operation sound, and five different machine learning models were trained and evaluated for their effectiveness in fault detection. Among these models, the K-Nearest Neighbors (KNN) and Extreme Gradient Boosting (XGBoost) models achieved a remarkable accuracy of 100%, outperforming the other models. Furthermore, the study identified the top five acoustic features that significantly contributed to the models' high performance, providing valuable insights for further improvements in fault detection systems. This research offers a robust and efficient solution for maintaining the performance of BLDC motors.