Fault Diagnosis of Foreign Object Insertion and Asymmetric Load in BLDC Motors Using FFT-Based Signal Analysis and CNN Model

This study proposes a CNN (Convolutional Neural Network)-based fault diagnosis method for real-time detection of BLDC (Brushless DC) motor faults. A cost-effective experimental system using an STM32 microcontroller was constructed, and current data were collected under three conditions: normal operation, asymmetric load due to rotor imbalance, and foreign substance insertion in the bearing. The collected current signals were analyzed using FFT (Fast Fourier Transform), and a CNN model was employed to classify fault types, especially for cases where frequency-based analysis alone was insufficient, such as in the presence of foreign substances. The proposed model achieved an average classification accuracy of 91.8%, demonstrating particularly high performance in detecting normal and asymmetric conditions. These results suggest that the proposed method can contribute to improving the reliability and maintainability of BLDC motor systems in practical applications.