Bearing fault severity assessment using variable-step multiscale fusion Lempel-Ziv complexity

Abstract

Abstract Most research on bearing health condition monitoring is devoted to distinguish the defective condition from the healthy condition. In practice, the fault severity assessment is also critical for performing the prognostics and maintenance of bearings. Lempel-Ziv complexity (LZC) has been widely used for the bearing quantitative fault diagnosis. However, the original LZC extracts fault information only at the single scale and often fails to portray the fault features. Then, the multiscale LZC is proposed to more comprehensively extract the fault information. However, multiscale analysis would shorten the length of time series and lead to inaccurate calculation results as the scale factor increases. As such, this paper proposes a novel bearing fault severity assessment method using variable-step multiscale fusion Lempel-Ziv complexity (VSMFLZC) to facilitate the quantitative fault diagnosis of bearings. The variable step length strategy is developed in the proposed method to optimize the coarse-grained procedure. Then, Laplace score is applied to evaluate the features and weights at each scale to obtain the proposed VSMFLZC. By such a fusion algorithm, the sequence can be converted into a single but comprehensive evaluation indicator for the fault severity assessment. The experimental results indicate that the proposed method outperforms the original LZC and multiscale LZC, where the fault features can be more comprehensively extracted and the fault severity assessment of rolling bearing can be successfully realized.