基于时序模型的加速退化数据可靠性评估

提出利用加速退化数据对高可靠长寿命产品进行可靠性评估与寿命预测的新方法.从时间序列的角度出发,研究了退化轨迹的一般建模方法；进而结合加速模型,利用整体推断方法给出了模型参数的极大似然估计.并以对数正态分布和Weibull分布为例,利用信仰推断方法给出了可靠度置信区间.通过有效利用加速退化试验各状态下的试验信息,结合时序模型对退化轨迹曲线自拟合性强的优点,提高了产品可靠性评估与寿命预测的稳健性,为基于产品加速退化数据进行可靠性评估提供了一条新的技术途径.最后通过实例验证了该方法的有效性.

Reliability assessment using accelerated degradation data based on time series model

A new method was proposed to evaluate reliability and predict lifetime using accelerated degradation data for high-reliability and long-life products. A general modeling approach based on time series for degradation path was analyzed. By introducing accelerated equations into analysis of accelerated degradation data, the calculation formula for reliability index and the evaluation algorithm for model parameters based on integral inference were presented. And taking the log-normal and Weibull distribution for example, the reliability confidence interval was given by fiducial inference. With the effective use of degradation data in different states as well as the advantages of strong self-adjustment with time series model, the robustness of product reliability assessment and lifespan prediction is improved. Moreover, based on accelerated degradation data, a novel technical way is provided to valuate reliability and predict lifetime. Finally, an example was presented to show the validity of this method.