加速退化试验可靠性评估的两步法与一步法比较研究

The accelerated degradation testing (ADT) under constant stress is an effective means for reliability assessment. It extrapolates product reliability under normal stress conditions by analyzing degradation data at elevated stress levels. Common approaches for handling degradation data include the one-step and two-step methods. The one-step method entails model parameter estimation and subsequent inference by maximizing the log-likelihood function based on degradation data. On the other hand, the two-step method first estimates pseudo-lifetimes for each sample, which is then converted into accelerated life testing (ALT) analysis. With the advancement in computational capabilities, the one-step method has become computationally more tractable, and previous research has compared both one-step and two-step methods in non-accelerated contexts. However, literature comparing these two methods in the context of ADT remains scarce. This study aims to systematically compare these two methods for constant stress ADT systems, providing more accurate and efficient guidance for selecting reliability assessment methods suitable for ADT. In this paper, we introduce criteria for distinguishing between overestimation and underestimation of true lifetimes using pseudo-lifetimes and compare the performance of the one-step and two-step methods in estimating the mean time to failure (MTTF) through numerical simulations under linear degradation conditions. Furthermore, we apply these methods to a set of classic datasets, and the results align with the simulation findings. In summary, the simulation results indicate that, for various sample sizes and numbers of observations, the one-step method demonstrates higher accuracy in product MTTF assessment compared to the two-step method based on pseudo-lifetimes with different distributions. This advantage is particularly pronounced in small sample scenarios.