A0198
Title: Multivariate degradation modeling with inverse Gaussian processes
Authors: Guanqi Fang - Zhejiang Gongshang University (China) [presenting]
Abstract: Many engineering products have more than one failure mode, and the evolution of each mode can be monitored by measuring a performance characteristic (PC). It is found that multi-dimensional degradation processes have often been observed in engineering practice. A novel multivariate degradation model built upon inverse Gaussian processes is introduced. The model is able to account for 1) the stochastic nature of each individual PC, 2) the heterogeneity among different units, and, more importantly, 3) any possible dependence among these PCs. Along with the model, some mathematically tractable properties are discussed, including the joint and conditional distribution functions that could facilitate future degradation prediction and lifetime estimation. In addition, a statistical inference method and model validation tools are provided. Finally, the proposed methodology is demonstrated using simulation studies and illustrative examples.
