多台同型产品同步纠正可靠性增长试验所得到的数据是单失效数据,依据美军标/国军标和AMsAA-BISE模型得到的分析结果,该数据往往高于产品实际可靠性水平。针对这种情况,提出一种数据统计分析的新方法,即先采取多层Bayes估计法和加权最小二乘估计法,求得各个失效发生时间的期望值,替代单台AMSAA模型中的失效时间,再实现对产品可靠性增长试验数据的统计分析。通过对以上方法进行蒙特卡罗模拟,对比分析结果表明,利用新方法所得结果更接近产品的实际可靠性水平。
The values of reliability assessment based on the MIL-HDBK-189, GJB/Z77 and AMSAA-BISE model for the failure data would he greater than the actual values. Herein, the expec- tation of each time-to-failure was fist obtained by using the hierarchical Bayesian method and least squares estimation method. Then, the growth testing data was analyzed through a new model ob- tained by substituting the expectation for the time-to-failure in AMSAA model. The effects of the three statistical methods mentioned above were assessed based on the Monte Carlo method. The re- sults indicate the values of reliability assessment by applying the method proposed are closer to the ac tual values.