一种综合性能与寿命数据的Bayes-Bootstrap方法

对现代长寿命、高可靠产品进行可靠性评估,传统大样本寿命试验和统计推断方法存在很大困难,包括伪寿命数据方法在内的性能与寿命数据综合的可靠性评估方法也存在很多问题。提出了将性能可靠性建模与Bayes方法相结合的方式解决这一困难问题,其中对性能试验数据建立性能退化模型,并利用Bootstrap方法处理伪寿命数据导致的随机性误差和构造融合Bayes验前分布,然后利用Bayes验后分析进行可靠性统计推断。对随机斜率模型和长寿命卫星动量轮轴承组件润滑寿命的应用表明,提出的方法可以充分利用产品研制、使用过程中各种类型的信息,特别是性能试验信息,具有一定的适用性。

A Bayes-Bootstrap Method Synthesizing Performance and Life Data

For modern long-life high reliability products,classical large sample life tests and statistical methods have many difficulties.There are also many issues when using methods including pseudo-life data method to synthesize both performance and life test data to assess product reliability.In this paper,a method based on performance reliability modeling and Bayes method was provided to solve this difficult problem.Therein a performance degradation model was built using performance test data,and the random error of pseudo-life data would be settled by Bootstrap method,which was also applied when constructing Bayes prior distribution.Then product reliability could be evaluated using Bayes posterior analysis.The method has been applied for a stochastic slop model and to solve lubrication life of bearing of some long-life satellite momentum wheel,which shows that the method can make full use of various kinds of information collected during product development and operation,especially performance test data,and is relatively widely applicable.