Amplitude-based multiscale reverse dispersion entropy: a novel approach to bearing fault diagnosis
Hao Song,Yong Lv,Rui Yuan,Xingkai Yang,Gangbing Song
DOI: https://doi.org/10.1177/14759217231219649
2024-01-20
Structural Health Monitoring
Abstract:Structural Health Monitoring, Ahead of Print. The multiscale fluctuation dispersion entropy algorithm (MFDE) is widely used to extract the characteristics from a variety of complex nonlinear signals, including bearing signals, due to its excellent performance to quantify the uncertainties of complex nonlinear systems. However, limited by the classification number and coarse-graining process, the periodic impulses generated by the defect point cannot be effectively detected by MFDE, restraining the characterization abilities of entropy features and resulting in undesirable diagnosis results for bearing faults. To overcome the disadvantages of MFDE, an amplitude-based multiscale dispersion entropy (AMDE) is proposed in this paper. The AMDE utilizes the phase scale factor to calculate multiple groups of amplitude difference series that contain different amplitude information. As such, the amplitude compression caused by the large-scale factor in traditional coarse-graining process is avoided, and the calculated entropy features not only characterize the irregularity of the whole signal but also reflect the changes of the impulse components. Afterwards, the perception range and the sensibility of AMDE are expanded and enhanced for amplitude variation, and the coarse-graining process and Gaussian reference are used to obtain multi-dimensional reversed entropy features. Combining those steps, the amplitude-based multiscale reverse dispersion entropy (AMRDE) algorithm is proposed. Finally, the capability of the proposed algorithm to track the amplitude variation and fluctuation is successfully demonstrated by analyzing noisy signals and amplitude-modulated signal. Meanwhile, the features extracted from bearing signals demonstrated that it is effective to use AMRDE to represent the health conditions of rolling bearing. Therefore, the entropy metric calculated by AMRDE can be the useful indicator in the fields of mechanical equipment fault diagnosis, structural health monitoring, and so on.
engineering, multidisciplinary,instruments & instrumentation