A Novel Particle Swarm Optimization Enhanced Smoothed Variable-Weight Grey Model for Gearbox Wear Prediction

Abstract

This study presents a gearbox wear prediction method by introducing a particle swarm optimization (PSO) enhanced smoothed variable-weight grey model (PSO-SVWGM(1, N)). The PSO-SVWGM(1, N) model improves upon conventional grey models by incorporating a smoothing generation factor with variable weights, optimized through PSO. Applied to oil debris analysis data from controlled gearbox wear experiments, the model achieved a remarkably low mean relative percentage error of 0.67% in predicting wear growth rates. The PSO-SVWGM(1,N) model can analyze intricate relationships between multiple wear-related elements enabling precise wear location identification, and successfully pinpointing the gear pump as the primary source of iron-related wear.  This research advances gearbox condition monitoring and demonstrates the powerful synergy between grey system theory and swarm intelligence in solving complex engineering problems. The findings have far-reaching implications for predictive maintenance strategies across various industries, promising more targeted interventions, extended gearbox lifespans, and reduced operational costs.

Keywords: Gearboxes, Wear prediction, Smoothed multi-variable grey model, Particle Swarm Optimization, Oil debris analysis, predictive maintenance