多源验前信息下基于证据理论的ML-II融合方法

Fusion of Information of Multiple Sources Based on Evidence Theory and ML-II Theory

在复杂新型产品的可靠性评估中,试验的样本量通常较小,而Bayes方法在小样本统计推断中比经典统计方法更为适用。针对可靠性工程中遇到的验前信息的多源性问题,给出了一种基于证据理论的ML-II融合方法。在融合过程中考虑了信息源的可信性,减小了可信度低的信息源对评估结果的影响。以电子产品为例,算例实验表明了该方法在有"干扰信息源"存在的情况下的评估结果的有效性和合理性。

The sizes of experiment samples are usually small in the reliability evaluation of comprehensive new products. In this case, Bayesian statistical method is generally more suitable than the classical methods and has thus been widely used in the process of reliability assessment. The situation is that the prior information comes from different sources. A ML-II fusion method based on evidence theory is proposed. The confidence levels of the information sources are taken into consideration and the problem caused by the sources with low confidence can be avoided. Using electric products, numerical examples are given to illustrate the effectiveness of the proposed approach when there is ＂noisy information source＂.