BP神经网络在女性假人颈部标定试验中的应用

Application of Neural Network BP in Neck Calibration Test of Female Dummy

采用BP神经网络代替响应曲面法的多项式函数,建立了基于BP神经网络的女性假人颈部减速特性模型。该模型以摆锤冲击速度、标定温度、蜂窝铝孔数、蜂窝铝切割形式为输入,10ms时刻颈部减速度为输出,实现了颈部减速特性的拟合预测功能。两种方法的预测结果表明,基于BP神经网络法的预测误差最大值为0.04m/s,而响应曲面法的预测误差最大值为0.05m/s。

This paper used neural network BP to replace the polynomial function of response curved surface method,and established neck deceleration character model of female dummy based on neural network BP.This model used pendulum impact speed,calibration temperature,hole numbers of honeycomb AL and cutting shape of honeycomb AL as input,the neck deceleration at the moment of 10 ms as output,and realized the fitting forecast function of neck deceleration character.The resuhs showed that the forecast error based on neural network BP was 0.04 m/s,the forecast error based on response curved surface method was 0.05 m/s.