Rolling Bearing Fault Diagnosis Using a Slope‐Weighted, Limited Penetrable Horizontal Visibility Graph and Related Indicator

The reliability and safety of mechanical systems rely heavily on fault diagnosis of rolling bearings. In comparison with traditional graph signals, the changes in the dynamic characteristics of vibration signals are more accurately reflected by the spectra obtained from graph signals converted by the weighted, limited penetrable horizontal visibility graph (WLPHVG). However, its capacity to record changes in the health status of bearings is significantly limited by the selection of the limited penetrable visibility distance, the choice of the weighting method, and the effectiveness of graph spectral feature enhancement and extraction. To address these issues, we developed a slope-weighted, limited penetrable horizontal visibility graph (SWLPHVG) by determining the limited penetrable visibility distance for WLPHVG and developing a slope-weighting approach. This improvement significantly increased the responsiveness of the WLPHVG graph spectrum to the dynamic characteristics of vibration signals. Finally, a fault feature indicator based on spectrum preprocessing, referred to as the graph spectral area of SWLPHVG, was proposed. This indicator revealed both the amplitude distribution and clustering characteristics of the graph spectrum while improving processing efficiency. Experimental results verified the superiority of these measures for fault diagnosis in rolling bearings.