基于聚类和优序图法的事故再现结果融合方法

Fusion of Accident Reconstruction Results Using Clustering and Precedence Ordering Chart

科学合理地融合多个事故再现结果具有重要现实意义,针对现有事故再现结果融合方法不能处理结果中包含潜在风险区间的问题,采用聚类算法与优序图法相结合的方式对多个事故再现结果进行融合。首先提出基于聚类的基础融合法,其先离散多个事故再现结果,再对离散后的数据进行聚类得到类结果,最后整理聚类结果可得融合结果。基础融合法操作简单、高效但不能处理再现结果中的潜在风险区间。进而通过引入优序图法提出基于优序图法的改进融合法,所提改进方法先从基础融合法的结果中得到各聚类区间概率与各聚类中心之间的距离表,并根据距离表中的距离极值将距离分段;再按照正态分布原则可知,各聚类中心之间的距离十分接近,现遵循距离越小分值越高的前提给予每段距离一定分值,距离表可根据分值得到各聚类中心的权重计算表;然后对权重计算表使用优序图法计算出各聚类中心的权重即各聚类区间权重;最后结合各聚类区间概率及其对应的权重得到融合结果。数值算例与真实案例均表明:改进融合法不仅可高效融合再现结果,还能识别出潜在风险区间并降低其在融合过程中的影响,从而提升融合结果的客观性,得到更理想的符合正态分布的结果,为融合包含潜在风险区间的多个事故再现结果提供有效解决方案。

The scientific and reasonable fusion of multiple accident reconstruction results is important.Existing methods for fusing accident reconstruction results cannot account for potential risk intervals in the reconstruction results.Hence,a clustering algorithm and a precedence ordering chart are combined in this study to fuse multiple accident reconstruction results.First,a basic fusion method based on clustering,which first discretizes multiple accident reconstruction results and then clusters the discrete data to obtain the clustering results,is proposed.Subsequently,the clustering results are organized to obtain the fusion results.The basic fusion method,although simple and efficient,cannot account for potential risk intervals in the reconstruction results.Hence,an improved fusion method based on a precedence ordering chart is proposed.First,each cluster interval probability and the distance table between the final cluster centers are obtained using the basic fusion method,and the distance is segmented based on the extreme value of distance from the distance table.Subsequently,based on the principle of normal distribution,a higher score is assigned to each segment of the distance when the distance is smaller;consequently,the weight calculation table of each cluster center can be obtained based on the score.Next,the weight of each cluster center,namely,the weight of each cluster interval,is calculated using the precedence ordering chart.Finally,the fusion results are obtained by combining the probability of each cluster interval and its corresponding weights.Numerical examples and actual cases show that the improved fusion method can not only fuse the reconstruction results efficiently,but also identify potential risk intervals and reduce their effects during the fusion process,thus improving the objectivity of the fusion results,yielding more ideal fusion results that conform to the normal distribution,and providing an effective solution for fusing multiple reconstruction results when the reconstruction results contain potential risk intervals.