Data-driven open-circuit fault diagnosis for PMSM drives: feature extraction via normalized current space vector sorting

Open-circuit faults in three-phase voltage source inverters can cause unbalanced currents, high torque ripple, and excessive core losses in permanent magnet synchronous machines. To effectively diagnose these faults, a novel fault feature extraction method based on normalized current space vector sorting is proposed. This approach transforms the three-phase currents under different fault conditions into unique fault patterns, ensuring a clear distinction between fault modes. Moreover, this method is robust to variations in torque and speed, ensuring reliable diagnosis across diverse operating conditions. To enhance fault detection capabilities, a one-dimensional convolutional neural network (1D-CNN) is designed to capture both local and global features effectively. The model is initially pre-trained using simulated data, with data augmentation applied to improve robustness and facilitate the learning of domain-invariant features. Simulation and experimental results validate the superiority of the proposed method over existing open-circuit fault diagnosis techniques across multiple evaluation metrics. Additionally, the proposed method's simple network architecture, fast inference times, high diagnostic accuracy and strong robustness make it a practical and efficient solution for real-world fault diagnosis applications.