Failure Lifetime Evaluation Based on Accelerated Generalized Wiener Degradation Process Models with Random Diffusion Coefficients

This paper proposes a modeling framework for nonlinear degradation under constant-stress accelerated degradation testing (CSADT) to predict failure lifetime. The proposed employs a generalized Wiener process to characterize degradation, wherein the drift coefficient is stress-dependent and the heterogeneity in the diffusion coefficient is explicitly modeled. Random effects are introduced to capture volatility variability across degradation trajectories, and model parameters are estimated via the expectation-maximization (EM) algorithm. Using the law of total probability, the probability density function (PDF) and reliability function of failure lifetime under normal operating conditions are derived. The proposed model is validated using crack propagation simulation data and experimental wear scar width data from an alloy product. The results demonstrate that the proposed model improves prediction accuracy for failure lifetime and reliability, highlighting its potential utility in engineering applications.