Inverse Gaussian–based model with measurement errors for degradation analysis

To build more credible degradation models, it is necessary to consider measurement errors in degradation analysis. This article proposes an inverse Gaussian-based state space model with measurement errors that can capture the unit-to-unit variability of the degradation rate by incorporating a random effect. Then, the lifetime distribution and alarm probabilities are derived. Under the non-Gaussian assumptions, conventional parameter estimation algorithms cannot be applied directly. Therefore, an improved expectation–maximization algorithm that is combined with particle methods is developed to estimate parameters. Finally, this article concludes with a simulation study and two case applications to demonstrate the applicability and advantages of the proposed model.