An Autoregressive-Based Motor Current Signature Analysis Approach for Fault Diagnosis of Electric Motor-Driven Mechanisms

Maintenance strategies such as condition-based maintenance and predictive maintenance have become key concepts in Industry 4.0, increasing the importance of online condition monitoring of electromechanical systems. Motor current signature analysis (MCSA) is a noninvasive alternative to vibration analysis for fault diagnosis of mechanical systems driven by electric motors. This paper presents a novel data-driven MCSA approach using autoregressive (AR) spectral estimation. A discrete wavelet transform (DWT) is applied to the motor currents to isolate noise and disturbances, and AR spectral estimation is then applied to selected wavelet details to extract features for fault diagnosis. A reference AR power spectral density (PSD) is estimated from healthy data and continually compared to new data via the symmetric Itakura–Saito spectral distance (SISSD), which serves as a health indicator. The method is validated on two real datasets: an in-house experimental setup with simulated imbalance faults and a public dataset with bearing faults. Results demonstrate the effectiveness of the proposed approach for both fault detection and isolation.