Bearing incipient fault feature extraction using adaptive period matching enhanced sparse representation.

• Proposal of a methodology for estimating the period of faulty impulses. • Development of an APMESR algorithm for extracting incipient bearing fault features. • Selection of MODWPT as the linear transformation for APMESR. • Evaluation and comparison of period estimation methodologies and transformations. • Verification of APMESR's effectiveness through several simulations and experiments. Bearing incipient fault feature extraction is crucial and challenging throughout its life cycle. In this paper, an adaptive period matching enhanced sparse representation (APMESR) algorithm is developed to address this issue. First, a novel methodology for estimating the period of faulty impulses is proposed from the perspective of mining the periodicity-related numerical patterns. Second, the period estimation methodology is embedded in a sparse representation model to implement adaptive period matching to form APMESR, which is capable of achieving periodic sparsity. Third, maximal overlap discrete wavelet packet transform is selected as the linear transformation of APMESR for improving its ability to reduce noise and highlight periodic impulse signatures. Furthermore, evaluations and comparisons are conducted using simulations to demonstrate the validity and performance of the proposed period estimation methodology, linear transformation, and APMESR. Experimental results indicate that APMESR can effectively extract incipient bearing fault features and outperforms other well-advanced methods. [ABSTRACT FROM AUTHOR]