Multiscale phase-space velocity diversity entropy and its application in marine engine cylinder liner wear fault diagnosis

Abstract

Diversity entropy and its multiscale variants have shown promise for nonlinear dynamic feature extraction in marine engine fault diagnosis. However, existing diversity-entropy-based methods mainly describe the geometric properties of reconstructed phase-space trajectories, while the dynamic evolution between adjacent trajectories is insufficiently considered, limiting their ability to capture wear-related changes in cylinder liner vibration signals. To address this issue, this article proposes Phase-Space Velocity Diversity Entropy (PVDE) and its multiscale extension, termed Multiscale Phase-Space Velocity Diversity Entropy (MPVDE), for marine engine cylinder liner wear fault diagnosis. Unlike conventional diversity entropy methods that focus on trajectory geometry, PVDE introduces velocity vectors between adjacent phase-space trajectories to characterize local evolution patterns. By calculating the similarity distribution of adjacent velocity vectors and combining it with multiscale coarse-graining, MPVDE extracts dynamic complexity information from multiple temporal scales. Simulation experiments based on the Tent map and noisy modulation signals show that PVDE has good complexity consistency, noise robustness, and computational efficiency compared with BDE, DE, and PE. Experimental validation under three cylinder liner conditions, including normal, slight wear, and moderate wear, shows that the MPVDE-based diagnostic method achieves the highest average accuracy of 98.06% among the compared multiscale entropy methods, demonstrating its effectiveness and practical potential for marine engine cylinder liner condition monitoring.