基于狄氏先验分布的机电产品主要失效模式贝叶斯分析

Bayesian Analysis Based on Dirichlet Prior Distribution of Dominant Failure Modes for Electromechanical Products

对于具有多种失效模式的复杂机电产品,主要失效模式分析是产品可靠性设计和试验中非常重要的环节。但目前常用的基于频率分析的定量分析方法,在小样本量场合并不是最优方法。针对这一问题,提出了基于Dirichlet先验分布的Bayes分析方法,实现了小样本量场合的主要失效模式定量分析。在此基础上采用Monte Carlo仿真分析方法的性质,并通过应用算例验证方法的有效性。结果表明,在小样本量场合Bayes方法比频率分析方法的精度更高,能更有效地找出产品的主要失效模式。

For complex electromechanical products with several failure modes,the analysis of dominant failure modes is a key process during reliability design and test.But the frequency analysis,which is commonly used for quantitative analysis at present,is not the best method when data is not plenty.For this problem,Bayesian method based on Dirichlet prior distribution is proposed for quantitative analysis of dominant failure modes in the case of small sample size.Monte Carlo simulation is performed to check the proposed methods and an example is provided.The results show that,Bayesian method is more suitable than the frequency analysis method,and is more effective to find out the dominant failure modes when sample size is small.