基于发动机罩板体素数据的头部碰撞损伤快速预测

打开文本图片集

主题词：行人保护机器学习头部损伤指标罩板结构3D-CNN自编码器

中图分类号：U463；TP181 文献标志码：A DOI： 10.19620/j.cnki.1000-3703.2025008

Fast Head Crash Damage Prediction Based on Hood Plate Voxel Data

Chen Guangl²，Shang Yunlong1²，Zhang Junyuan4，Tang Hongbin³，Sun Xiuxiu1.2 （1.Engneering，ianjin3O3O;2.ianjinPower for s， 3.Global Center， FAW CorporationLimited，130o13;4.Colege ， ， ）

【Abstract】Inorder toquickly evaluate thepedestrian protectionperformance hood platestructures difrent vehicles atthe vehicle development stage，thispaper proposesa machine learning prediction method basedon3D voxel data. ，a3Dvoxeldatasetcenteredonthecolisionpointthehodplateheadisconstructedbasedonthefiniteelementmodels 7vehicles toaccuratelycharacterizethecomplex 3Dstructurethehoodplate.Second，unsupervisedfeature learning is performedonthe3Dvoxeldata byusing3D-CNNAuto-Encodertogenerate the1Ooo-dimensional coded dataset. Subsequently，thenoisefeaturesarefilteredbythePCA lineardownscaling techniquetoobtain the24-dimensionalPCA dataset.Finally，theperformancethetwodatasetsonthefourmachinelearning modelsiscompared.Theexperimentalresults showthatthe224-dimensionalPCAdataset hasabetter prediction effect，inwhich thecoeffcient determination R2 the SVR model for the prediction HIC a-peak reaches 98.86% 98.44% ，respectively，which significantlyimproves the predictionaccuracygeneralizationability.Inaddition，thecomputationtimemorethan1OOsetscollsionpoints fora singlevehicleisshortenedtolessthan1Os，whichsignificantlyimprovestheficiencythedevelopmentautomobilesafety performance.

words：Pedestrian protection，Machine learning，Head Injury Criterion （HIC），Cowl structure，3D-CNNAuto-Encoder

【引用格式】陈光，尚云龙，张君媛，等.基于发动机罩板体素数据的头部碰撞损伤快速预测[J].汽车技术，2025（10）：52-62.CHENG，SHANGYL，ZHANGJY，etal.Fast Head CrashDamagePrediction BasedonHoodPlateVoxelData[J]. ，2025（10）： 52-62.

1前言

在交通事故中，行人头部撞击发动机罩板是造成其致命伤害的主要原因之一]。（剩余12312字）