Fault severity assessment for rotating machinery via improved Lempel–Ziv complexity based on variable-step multiscale analysis and equiprobable space partitioning

Abstract

Fault severity assessment based on vibration measurements for rotating machinery is critical since it can reduce downtime and guarantee the reliability of the equipment. Lempel–Ziv complexity (LZC) has been widely used for the fault severity assessment. However, LZC is of a single-scale analysis and 0-1 encoding, which cannot fully explore the features of vibration signals measured from rotating machinery. This paper, thus, proposes an improved LZC based on the variable-step multiscale analysis (VSMA) and equiprobable space partitioning (ESP) strategies to fully explore features of vibrations of rotating machinery. The VSMA is proposed to overcome the drawback that the single-scale analysis fails to comprehensively uncover features and solve the problem that the traditional multiscale analysis significantly reduces the length of sequences. With the VSMA, a string of time series under different scales can be generated. The ESP is developed to transform the time series into symbolic series, with the capability of reserving the features of vibration signals and being more robust against noise, particularly for non-stationary signals. Then, the ESP based variable-step multiscale LZCs (i.e. ESP-VSMLZCs) are obtained. To fuse the obtained ESP-VSMLZCs and obtain a comprehensive indicator for fault severity assessment, Laplacian score weighting is adopted. As such, a single ESP based variable-step multiscale fusion LZC indicator can be obtained. The proposed indicator is verified by simulated data from a bearing dynamic model and experimental data measured from rotating machinery.