多源不确定信息下车人碰撞事故高可靠再现方法研究

There are multi-source uncertainties in vehicle-pedestrian accidents because of the influence of many factors, such as limited information. The two most important kinds are the uncertainty in model selection and the uncertainty in traces, under such condition, how to enhance the reliability of the reconstructed result is a necessary and important problem in the accident reconstruction field. So, a method is proposed to solve such problem, which is to validate traces firstly, and then to calculate the calculation uncertainty, a final conclusion can be given by result fusion methods finally. Considering that the injury is not concerned and there is no cross-validation methods for dealing with the multi-uncertain traces problem in trace validation, the relationship between different parts of the human body will be given according to accident cases data and simulation data firstly, and then an model for validating traces will be given according to the relationship, methods for cross- validating traces based on cluster analysis theory will be given finally. Considering that it will introduce new uncertainty in calculation uncertainty analysis because of there are too many methods, a method for analyzing the calculation uncertainty based on the evidence theory will be given. Considering that there is no method in result fusion, methods for fusing the accident reconstruction results based on the evidence theory and the multiple attribute decision making theory will be given. And all methods’ accuracy will be improved by introducing the sub-interval technique, while their efficiency will be improved by optimization modeling. The research will provide as much methods for reconstructing an accident effectively as possible.

受信息局限等诸多因素影响，事故再现中面临模型选择、痕迹等多源不确定性问题，如何高可靠再现事故亟待研究。为此，本项目以车人碰撞事故为研究对象，通过先验证痕迹、再分析结果计算不确定性、最后融合结果的方法，实现将模型选择、痕迹等多源不确定信息客观反映到结果中去，以此提升事故再现的可靠性。针对痕迹验证中较少关注人体损伤且缺少多不确定痕迹交叉验证方法的问题，通过案例及仿真数据给出人体高相关部位损伤之间关系且据此建立痕迹验证模型，并提出基于聚类分析的痕迹交叉验证方法；针对计算不确定性分析中因方法过多而导致选择困难的问题，提出基于证据理论的面向痕迹的计算不确定性分析方法；针对结果融合中方法极少的问题，提出基于证据与多属性决策等理论的结果融合方法。同时引入子区间技术以提高各方法计算精度、建立多层嵌套优化模型并求解以提升各方法计算效率。研究成果将为高可靠再现车人碰撞事故提供更多实用、高效的方法。

项目紧密围绕“多源不确定信息下车人碰撞事故高可靠再现方法”开展研究，具体从人体各部位损伤之间关系及其痕迹验证建模、面向痕迹的事故再现结果计算不确定性分析方法、基于聚类分析的事故现场痕迹交叉验证方法、多模型下事故再现结果融合方法4个方面开展研究。在人体损伤方面，探讨了车头参数对人地碰撞损伤的影响，分析了车人碰撞事故中人体损伤来源及相关性，建立了人体头胸部损伤相关性模型，提出了系列防护人地碰撞损伤的方法，提炼了行人在碰撞事故中的应急反应，评估了考虑行人应急反应的行人自动紧急制动系统效果。在不确定性分析方法方面，总结了区间痕迹下事故再现结果不确定性问题的求解策略，提出了用于事故再现结果不确定性分析的实验设计方法及面向案例的事故再现结果不确定性分析方法。在痕迹交叉验证方法方面，提出了基于熵值法的多模型事故再现异常结果识别与融合法、一种复杂不确定条件下事故再现结果不确定性分析方法及基于聚类的融合多模型事故再现结果的两种方法。在事故再现结果融合方法方面，提出了基于证据理论与OC算子法的系列多模型事故再现结果融合方法及一种无需专家意见的多模型不确定事故再现结果融合方法。除此之外，围绕弱势交通参与者损伤防护，开展了骑车人与行人损伤对比、骑乘人员损伤及运动学响应差异及骑车人耐受极限座高等探索性工作。相关成果为获更客观事故再现结果、更好保护道路弱势交通参与者提供了规律及方法支持，具有理论意义与实践价值。