Bayesian Reliability Inference for Non‐Linear Degradation Accelerated by Non‐Arrhenius Mechanism With Application on O‐Rings

Accelerated degradation testing (ADT) is an effective approach for assessing the reliability of polymeric materials, particularly rubber O‐rings. However, degradation mechanisms may vary across temperatures, leading to non‐Arrhenius behavior, as exemplified by the 1986 Space Shuttle Challenger disaster. Existing non‐Arrhenius models capture temperature‐dependent mechanism transitions but often rely on fixed combinations of mechanisms or piecewise approximations, thereby limiting flexibility and accuracy. We propose a novel non‐Arrhenius model that smoothly adjusts the contributions of high‐ and low‐temperature processes, offering a continuous representation of mechanism transitions. Embedded within a stochastic process framework, the model is applicable to ADT data and supports Bayesian reliability evaluation using power priors to integrate historical information. Validation with paired historical and current high‐temperature ADT datasets of nitrile rubber O‐rings shows that our method outperforms existing models in term of deviance information criterion (DIC) and fitting accuracy, providing a robust tool for mission‐critical reliability assessment.