基于解释图的概率多规划识别事故分析方法研究

基于多规划识别在行为分析的多目标性、过程性与结果解释性方面的优势，提出了一种改进的概率多规划识别事故分析法，并将其应用于井下电机车撞、压人事故行为分析。通过构建事故规划库和基本事件解释图，找出顶层事故致因的所有解释集合，并定义解释集合重要度对解释集合路径进行重要程度排序，从而找到导致顶层事件发生的最佳解释集合；通过临界重要度系数验算，对解释集合重要度的排序结果进行论证，两者计算结果基本一致且后者优势明显，说明了基于解释图的多规划识别算法应用于事故分析的可行性和优越性；最后指出了此算法对处理逻辑结构复杂的大样本事故分析存在的不足及拟解决办法。

Study on accident analysis method with probabilistic multiple plan recognition based on interpretation map

Based on the advantages in the aspects of multi objectives, procedure and results explanatory of multiple plan recognition in behavior analysis, an improved accident analysis method with probabilistic multiple plan recognition was proposed and applied in the behavior analysis of human knock and pinning accidents caused by underground electric locomotive. All the interpretation sets of top accident cause were found out through establishing the accident plan library and the interpretation map of basic events, and the importance of interpretation set was defined to carry out the important degree sorting on the interpretation set path, so as to find out the best interpretation set causing the occurrence of top event. The sorting results about the importance of interpretation set were demonstrated through the checking on critical importance coefficient. It showed that the calculation results of them were basically consistent, and the latter had the more obvious advantage, which illustrates the feasibility and superiority of applying multiple plan recognition algorithm based on interpretation map in accident analysis. Finally, the shortcomings of this algorithm when processing large sample accident analysis with complex logical structure and the proposed solutions were pointed out.