AI-powered acoustic analysis for non-intrusive fault detection in BLDC motors of electric vehicles

The necessity for stable and efficient Brushless DC (BLDC) motors has increased due to the quick uptake of electric vehicles (EVs). Due to their compact size, high efficiency, and low maintenance requirements, BLDC motors are ideal for propulsion applications. To support safe, efficient operation and enable predictive maintenance in EVs, BLDC motors must operate reliably. Traditional fault detection techniques frequently depend on intrusive, sensor-based temperature, vibration, and current monitoring. Despite their effectiveness, these methods raise integration difficulties, system complexity, and cost, particularly in sealed or small motor designs. This paper presents an acoustic-based non-intrusive fault detection approach that utilizes artificial intelligence (AI) and acoustic signal processing. Motor noises under several states of normal operation, bearing fault, and propeller fault are captured and analyzed to extract essential acoustic properties, including Zero Crossing Rate (ZCR), Root Mean Square (RMS), Spectral Centroid, and Mel-Frequency Cepstral Coefficients (MFCCs). A dataset comprising the retrieved features is assessed using a number of supervised machine learning (ML) models. Comparative results show that the Voting Classifier (VC) achieves the best performance among the evaluated models, with approximately 98% classification accuracy on the test dataset. The results demonstrate the feasibility of using acoustic signals for non-intrusive BLDC motor condition monitoring under controlled laboratory conditions.