以B(O,O)作为元件可靠性的先验分布的二项串联系统可靠性的贝叶斯置信限的讨论

DISCUSSION ON BAYES' CONFIDENCE BOUND FOR BINOMIAL SERIES SYSTEM RELIABILITY USING B (0,0) AS THE PRIOR DISTRIBUTION OF COMPONENT RELIABILITY

本文对以B(0,0) 作为元件可靠性的先验分布的二项串联系统可靠性的贝叶斯置信限进行了讨论,提出: 1° 用梅林变换法得到的贝叶斯精确解与矩拟合法得到的贝叶斯近似解,两者比较接近。 2° 矩拟合法得到的结果,相对于经典精确解是“冒进”的,经模拟计算,从覆盖系统可靠性真值P~*的百分比来看,也是冒进的。

In this paper, we discussed the Bayes' confidence bound for binomial series system reliability using B (0,0) as the prior distribution of component reliability. We point out:1. The Bayes' exact solution obtained by Merlin transformation method and the Bayes' approaximate solution obtained by moment fitting method are with a very little difference, only.2. The results obtained from moment fitting method is non-conservative than the classical exact solution. Through simulating computation, evaluating from the proportion that overlaying the true value of system reliability, p, they are non-conservative also.