机器学习在行人保护概念分析中的应用研究

Application Research of Machine Learning in Conceptual Analysis of Pedestrian Protection

为应对在新开发车型的造型及总布置阶段,由于没有整车详细设计数据而对行人保护腿部伤害性能定量分析难度大的问题,引入机器学习理念。通过总结整理已开发车型造型特征及行人保护aPLI腿型伤害结果数据,形成关联数据库,并作为训练数据,搭建基于反距离加权算法的机器学习模型。经验证,机器学习预测分析结果与基于概念设计数据的概念计算机辅助工程(CAE)仿真分析结果准确性相当,表明方案可行;但仍与基于详细设计数据的详细计算机辅助工程(CAE)仿真分析结果存在一定偏差,还需要持续将具有不同造型特征的车型输入输出关联数据添加到训练数据库中,提升机器学习预测分析结果的准确性。

In order to address the difficulty of conducting quantitative analysis of pedestrian protection leg injury performance prediction for styling reviews due to the lack of detailed design data for the vehicle during the styling and general layout stages of new vehicle development,the concept of machine learning was introduced.An associated database was formed by summarizing and collating the styling features of developed vehicle and pedestrian protection advanced pedestrian legform impactor(aPLI)legform injury result data as training data,establishing a machine learning model based on the inverse distance weighted algorithm.It has been verified that the accuracy of the machine learning prediction analysis results is comparable to the conceptual computer aided engineering(CAE)simulation analysis results based on conceptual design data,and the scheme is feasible.However,there is still a certain deviation from the detailed computer aided engineering(CAE)simulation analysis results based on detailed design data.It is also necessary to continuously add vehicle model input and output correlation data with different modeling characteristics to the training database,to improve the accuracy of machine learning prediction analysis results.