Multichannel multiscale increment entropy and its application in roller bearing fault diagnosis

As a new index to measure the complexity of time series, increment entropy, which takes into account the fluctuation directions and amplitude of time series, has better performance than the traditional entropy analysis methods such as sample entropy and permutation entropy. However, the increment entropy value of time series at single scale cannot completely reflect the dynamic change of time series. In this paper, the multichannel multiscale increment entropy (MMIE) is proposed by introducing the coarse-graining and multichannel analysis tools of time series to explore the complexity of multichannel time series over multiple different scales. MMIE considers the dynamic relationships between multichannel data and the relevant cross-channel variations that are often overlooked in single-channel analysis, thus achieving a full utilization of state information. Finally, an MMIE-based fault diagnosis method is proposed for roller bearing. The analysis results of the simulation signals and the measured data of roller bearings indicate that MMIE performs better than MMDE, MMPE, and MMSE approaches in time costing and fault identification rates.