A dual-objective optimized reweighted overlapping group sparse framework integrating frequency slice function for robust bearing fault diagnosis
Chaoyong Ma, W. Zhang, Lili Meng, Miaorui Yang, Kun Zhang, Yonggang Xu
Abstract
Robust detection of incipient bearing faults remains a significant challenge in industrial environments characterized by strong background noise. This paper introduces a dual-objective optimized reweighted overlapping group sparse shrinkage (ROGSS) framework combined with frequency slice function (FSF) to enhance weak fault feature extraction. The proposed approach begins by constructing frequency slice wavelet transform (FSWT) to achieve decoupled frequency-domain representations of signal components. Subsequently, a reweighted overlapping group sparse shrinkage model is designed to suppress noise through frequency-band energy redistribution. To overcome multi-parameter optimization challenges, a dual-criteria evaluation framework integrating harmonic spectral kurtosis (HSK) and correlation Theil index (CTI) is developed, with a multi-objective particle swarm optimization (MOPSO) algorithm employed for Pareto-optimal parameter selection. Validation through simulation and experimental case studies demonstrates that the proposed method significantly enhances fault-related impulse features while effectively suppressing noise, achieving 15–20% improvement in signal-to-noise ratio (SNR) and 30% reduction in root mean square error (RMSE) compared to state-of-the-art techniques. The approach establishes a novel spectral analysis framework for weak fault detection, providing superior performance in feature enhancement and diagnostic reliability under low signal-to-noise ratio conditions.