摘要
为了将非成败型试验数据折算为成败型数据,在分析现有方法局限性的基础上提出多点优化法。该方法于原始置信分布上选取多个折算控制点,并取这些位置点上的误差平方和作为目标函数,进而利用优化算法求解获得折算后的成败型数据和置信分布。该方法没有现有方法的局限性,是一种普遍适用的折算方法。算例表明,该方法较其他现有方法,如二阶矩法,点估计下限法,两点法以及近似优化两点法等,在应用范围、折算精度和计算量方面具有明显的优势,是一种简单而行之有效的折算方法。
In order to convert non-binomial data to binomial data, multi-points optimization method is presented in this paper basing on the limitation analysis of the conversion methods. Multi-points optimization method used already, which selected several points as the control points in original confidence distribution. By this method, the sum of the square errors at these points was chose as the objective function. Using optimization algorithm, the equivalent binomial data of the original test data could be got. Then the reliability confidence distribution after conversion could also be obtained. Without the limitations, multi-points optimization method is generally applicable. In aspects of usage range, conversion precision and computational quantity, the multi-points optimization method have advantages over other conversion methods used already, such as two moments method, point estimation and bounded below method, two points method and approximately optimized two points method. The validity and the advantages of the method proposed in this paper are proved by the example.
出处
《强度与环境》
2010年第2期55-60,共6页
Structure & Environment Engineering
关键词
可靠性
置信分布
数据折算
优化算法
非成败型数据
reliability
confidence distribution
data conversion
optimization algorithm
non-binomial data
