Scholarlink

Based on the car’s frontal collision test, this article uses MADYMO software to digitally model the occupant’s frontal collision. By adjusting the occupant’s sitting angle, the AEB (Autonomous Emergency Braking) pre-braking conditions, collision conditions and collision conditions under the action of AEB are simulated. The simulation data is put into the BP (Back Propagation) neural network for training, in which the simulation data of AEB pre-braking conditions and the simulation data of collision conditions are used as input, and the simulation data of collision conditions under the action of AEB are used as output. The input and output data include seat belt force, Simulation data such as head acceleration, neck bending moment, chest compression, etc. Then, the AEB pre-braking condition test data and collision condition test data are put into the trained prediction model to predict the data of the collision condition under the action of AEB. Finally, the EEARTH (Enhanced Error Assessment of Response Time Histories) method is used to score the simulation data separately from the predicted data and experimental data, and the predicted scores and actual scores of various data are compared and analyzed. The results show that the built BP neural network model has good prediction effect and the accuracy prediction method is feasible.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Collision Simulation Model Accuracy Prediction Based on BP Neural Network

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Hailan Xu,
  • Zhenfei Zhan,
  • Yajun Luo,
  • Changhao Li,
  • Hailiang Chen

摘要

Based on the car’s frontal collision test, this article uses MADYMO software to digitally model the occupant’s frontal collision. By adjusting the occupant’s sitting angle, the AEB (Autonomous Emergency Braking) pre-braking conditions, collision conditions and collision conditions under the action of AEB are simulated. The simulation data is put into the BP (Back Propagation) neural network for training, in which the simulation data of AEB pre-braking conditions and the simulation data of collision conditions are used as input, and the simulation data of collision conditions under the action of AEB are used as output. The input and output data include seat belt force, Simulation data such as head acceleration, neck bending moment, chest compression, etc. Then, the AEB pre-braking condition test data and collision condition test data are put into the trained prediction model to predict the data of the collision condition under the action of AEB. Finally, the EEARTH (Enhanced Error Assessment of Response Time Histories) method is used to score the simulation data separately from the predicted data and experimental data, and the predicted scores and actual scores of various data are compared and analyzed. The results show that the built BP neural network model has good prediction effect and the accuracy prediction method is feasible.