利用相似产品信息的成败型产品Bayes可靠性评估

Bayes comprehensive assessment of reliability for binomial products by using test information of similar product

在成败型产品的可靠性评估中,为提高估计精度,经常利用历史数据来确定先验分布.但在工程实际中,历史样本和样本本质上属于不同的总体,这对可靠性评估结果有着显著影响.为此,采用相似系统分析确定历史样本和样本的相似程度,将其归纳为继承因子;然后,根据历史样本信息确定产品可靠性的历史后验,基于无信息先验得到产品可靠性的更新后验;最后通过继承因子,综合历史后验和更新后验,得到产品可靠性的融合后验,并在此基础上进行可靠性推断.该方法不仅充分利用了相似产品信息,而且突出了产品的独有特性.

For reliability assessment of binomial products, the history sample and the current sample often follow different populations, which have remarkable effect on reliability assessment. The similarity degree between the history sample and the current sample were studied by similar system analysis, and their similarity degree was denoted by inheritance factor. Then, the history posterior was obtained from the history sample and the innovation posterior was given by the non-informative prior, which reflect the history information of similar products and the current test information respectively; the complex posterior was obtained from the history posterior and the innovation posterior via inheritance factor, and statistical inference was made on the basis of the complex posterior. This method makes full use of information of similar products and emphasizes the particular characteristic of the evaluated product.