基于置信分布和K-L信息距离的成败型产品可靠性分析

Bayesian analysis for reliability of success-failure products based on confidence distribution and K-L information distance

针对成败型小子样产品可靠性试验现场数据少、具有多源先验信息的问题,基于K-L信息距离理论和试验数据置信分布方法,提出一种多源信息融合方法。利用K-L信息距离度量参数分布之间距离,从而确定不同权重融合多源先验分布,进而根据Bayes方法得到更加合理的验后分布。示例分析及对比验证了本方法的可行性和合理性。

To solve the problems of lacking real samples and multi-source prior information in reliability test for small samples of success-failure products,a method of multi-source information fusion was proposed based on K-L in-formation distance theory and confidence distribution method.The distances between parameter distributions were meas-ured by K-L information distance,and multi-source prior distributions were fused according to the weights.Further,the more rational posterior distributions were obtained by using the Bayesian method.Example analysis and comparison ver-ify the feasibility and rationality of the method.