Individualized inverse Gaussian process-based reliability modeling and optimal degradation test design for rubber V-belts

Abstract

This study presents an individualized inverse Gaussian process-based reliability modeling and optimal degradation test design for rubber V-belts, addressing the high cost, long duration, and data scarcity of traditional reliability demonstration testing (RDT). Degradation experiments on five B-type V-belts monitored slip ratio, tension force, and primer crack depth every 12 h, yielding lifetimes between 190 and 209 h (mean: 197 h). A stochastic degradation model incorporating individual heterogeneity was developed, with parameters estimated via Bayesian inference using Markov Chain Monte Carlo. By minimizing the asymptotic variance of the 0.1-quantile lifetime under cost constraints, an optimal RDT scheme was derived. Results show that when the asymptotic variance is below 0.0035, the relative error between estimated and actual lifetimes remained below 1.52%. The optimal plan—four samples, 18 measurements, and a 10-h interval—achieves high evaluation accuracy at a total cost of 7696 CNY. This work advances existing methods by explicitly modeling individual variability, justifying the inverse Gaussian process choice through empirical validation, and offering a practical, cost-efficient RDT framework extensible to other degradation-prone products.