Adaptive design and dynamic optimization for accelerated life testing with non-destructive one-shot devices: A sequential information maximization approach

This paper proposes the Recursive Information Trace Optimization (RITO) framework for adaptive constant-stress accelerated life testing (CSALT) of non-destructive one-shot devices. Conventional static designs fix inspection times a priori, resulting in suboptimal information extraction. RITO introduces a dynamic two-phase process: after initial fixed inspections, it recursively optimizes subsequent inspection times via trace maximization of the Fisher information matrix using real-time parameter estimates. This facilitates adaptive adjustments that approach theoretical information bounds, validated by a detailed Monte Carlo simulation study. The framework implements log-location-scale lifetime distributions (demonstrated via Weibull model) with log-linear stress-life relationships. RITO achieves near-optimal estimation efficiency with low computational needs, operable on standard PCs without specialized hardware and memory. A case study demonstrates practical advantages in reliability estimation under accelerated stresses. The proposed approach provides a computationally tractable, resource-efficient solution for precise reliability assessment.