Multi-Channel Data-Driven Broken Rotor Bar Fault Diagnosis Using Sparse Envelope Spectral Analysis

Abstract

Precise and prompt diagnosis of broken rotor bar (BRB) faults is essential for upholding the reliability of induction motors. Spectral analysis of motor current is the primary technology for diagnosing BRB faults, but its effectiveness is hindered by limited sampling time. While employing multi-channel schemes can enhance data size/quality, existing methods still suffer from inherent low resolution and power supply masking effect, and also require impractically precise synchronization of monitoring for all phase currents. To address these limitations, this study introduces a new method called sparse envelope spectral analysis for multi-channel data-driven BRB fault diagnosis. Initially, the Hilbert transformation is applied to extract current envelopes from various channels, effectively mitigating masking effects caused by power supply components. Subsequently, sparse envelope spectra are recovered by maximizing the Bayesian joint posterior probability of multi-channel data. This is achieved by effectively capturing the arithmetic structure of fault sidebands, eliminating the synchronization requirement due to the phase-independence of the magnitude spectra. Additionally, a real-valued transformation is integrated into the sparse envelope spectral analysis to streamline computational complexity by circumventing time-consuming complex-valued operations. Both simulation and experimental results demonstrate the superiority of the proposed method, and the corresponding MATLAB codes have been made available online.