Inverse Gaussian process models for degradation analysis: A Bayesian perspective

Abstract

This paper conducts a Bayesian analysis of inverse Gaussian process models for degradation modeling and inference. Novel features of the Bayesian analysis are the natural manners for incorporating subjective information, pooling of random effects information among product population, and a straightforward way of coping with evolving data sets for on-line prediction. A general Bayesian framework is proposed for degradation analysis with inverse Gaussian process models. A simple inverse Gaussian process model and three inverse Gaussian process models with random effects are investigated using Bayesian method. In addition, a comprehensive sensitivity analysis of prior distributions and sample sizes is carried out through simulation. Finally, a classic example is presented to demonstrate the applicability of the Bayesian method for degradation analysis with the inverse Gaussian process models.