A Subspace-aided Fault Diagnosis Approach for Electrical Drive Systems

For the safe operation of electrical drive systems, a data-driven fault diagnosis approach is proposed by subspace-aided signal reconstruction in this paper. Specifically, based on the subspace identification method (SIM), the frequency estimation of the signal is reformulated into the eigenvalue identification, which can be directly extracted from the time-domain signal. The subspace frequency estimation approach can avoid the problem of spectrum leakage. Under the known frequency estimation, the amplitude and phase estimation are adopted by the least squares (LS) method. The signal is reconstructed for fault detection by estimating frequency, phase, and amplitude. The fault detection scheme is designed to be helpful for the implementation of fault diagnosis in the periodic signal. Eventually, the effectiveness of the proposed approach is verified.