基于不确定性测度与支持度的测试性验前信息融合方法

针对目前验前信息融合方法存在主观性和单一性的缺陷,引入验前分布的不确定性测度和支持度,提出了基于不确定性测度与支持度相结合的验前信息加权融合方法.首先,利用β分布对测试性验前信息的不确定性进行描述,运用不同的验前信息确定验前分布超参数;然后,引入不确定性测度和支持度作为验前信息加权因子,设计了相应的融合算法;最后利用混合验前分布,结合测试性现场试验数据,计算了Bayes融合评估的验后结果.该方法不仅考虑了验前分布本身对测试性参数的真实情况描述的接近程度,还考虑了不同来源的验前信息之间的支持程度.实例分析表明,在同一置信水平下,基于该方法的测试性评估置信下限相对于传统方法的测试性评估置信下限大约提高了0.7%. 

Testability prior information fusion method based on uncertainty measure and supporting degree

To overcome the shortcoming of subjectivity and unicity in those fusion methods, the uncertainty measure and supporting degree of prior distribution was introduced; and the prior information weighting fusion was proposed on the basis of the uncertainty measure combined with the supporting degree. First, the uncertainty of testability prior information was described using β distribution and the prior distribution hyper-parameter was determined by different prior information. Then, the uncertainty measure and supporting degree was proposed as the weight coefficient of prior information, and the corresponding fusion algorithm was designed. Finally, using the mixed prior distribution and combining with the testability field test data, the posterior results of Bayes fusion evaluation were caculated. This method considered both the proximity of the measure parameter to the real situation, and also the level of supporting degree between different prior information. The example shows that the lower limit of confidence estimation based on the proposed method increases 0.7% than the traditional method at the same confidence.