GE-HOFMD fusion-based early fault feature extraction and diagnosis for rolling bearings

Abstract Currently, fault diagnosis of rolling bearings faces the challenge of accurately identifying early weak faults under strong noise interference. To address the difficulty of extracting early fault characteristic frequencies in noisy environments, this paper proposes a new method for extracting early fault characteristics of rolling bearings to improve the accuracy of fault characteristic extraction and diagnosis. First, the health status of rolling bearings is assessed using the generalized entropy (GE) index, and based on this assessment, abnormal signals are detected. Second, in response to the impact of parameters on the decomposition efficiency of feature mode decomposition (FMD), the Hippopotamus Optimization (HO) algorithm is used to optimize the filter length and the mode number of FMD. Finally, considering the periodic influence of fault signals and noise intensity, an evaluation index for HO-FMD is constructed by fusing the envelope Gini and ensemble kurtosis (EGEK) index. Experimental results on simulated signals and two full-life bearing datasets demonstrate that the proposed method can effectively extract early fault characteristic frequencies under strong noise conditions (down to −15 dB), significantly enhance fault feature clarity, and achieve earlier fault detection compared with traditional indicators and signal decomposition methods such as HOVMD and wavelet packet decomposition. These results verify the effectiveness and robustness of the proposed method for early fault diagnosis of rolling bearings.