VEHICLE CRASH ACCIDENT RECONSTRUCTION
BASED ON FEM AND NEURAL NETWORKS

ZHANG Xiaoyun¹  JIN Xianlong¹  QI Wenguo¹  HOU Xinyi²

(1.School of Mechanical Engineering, Shanghai Jiaotong University, Shanghai 200240;

2. Traffic and Police Office of Shanghai, Shanghai 200070 )

Abstract: To utilize the deformation of the vehicle and the collision objects in the accidents fully, a new vehicle crash accident reconstruction method by means of the finite element method and neural networks techniques is presented. In the present method, the digital measurement technology is adopted to acquire the deformation of key points on the main energy-absorbing parts as the indices to evaluate the accident. And the non-linear explicit finite element code is adopted to simulate the crash accidents in order to acquire the calculation values of these indices. On the basis of the numerical results of the crash accidents, a three-layer recurrent neural network is applied to generate an approximated function of the initial crash parameter and the deformation index. The finite element analyses are used to generate the examples for the training and test sets of the neural network. These results could be used to train the neural network by back-propagation learning rule. Application the present method to one vehicle-to-wall accident, at first, the finite element model of the auto-body, wall and the surrounding are finished, then the eleven key points on the frontal longitudinal beam and the mudguard are chosen, reconstruction result are solved by comparing the deformation of measuring in the real accident with the data of the simulation results. It is proved to be effective on analysis of this kind of accidents, and so can provide a scientific foundation for accident judgments.

Key words: Finite element Neural networks Crash Accident reconstruction Simulation

CLC No: U491.3

国家自然科学基金资助项目(60174023). Received 20060404, received in revised form 20061129