RUL prediction based on a two-phase random volatility inverse Gaussian process

To address the phased-depended and random characteristics exhibited by products along their degradation process, this paper proposes a novel degradation modeling and reliability analysis method based on a two-phase Random Volatility Inverse Gaussian (RV-IG) process. Firstly, the Schwarz Information Criterion (SIC) is employed to identify change points in the product’s degradation trajectory. Secondly, parameter estimation is conducted through a hybrid approach of Maximum Likelihood Estimation (MLE) and Gibbs sampling algorithm. Then, to tackle the analytical challenges of reliability assessment for the two-phase random IG degradation process, statistical inference of the remaining useful life (RUL) for the constructed model is performed using convolution formulas and Bayesian algorithms. Finally, the proposed model is validated through simulation and a real-world case study involving train wheels degradation. The results demonstrate that the constructed model outperforms, traditional single-phased and fixed effects degradation models in terms of reliability predictions accuracy, thereby better aligning with engineering practice needs.