加速度峰值RBF神经网络预测  

Prediction of acceleration peak based on radial basis function network

作　　者：刘卫[1] 孙晓霞[1] 沈瑞琪[1] 叶迎华[1] 李创新[1] 

机构地区：[1]南京理工大学化工学院,江苏南京210094

出　　处：《南京理工大学学报》2013年第5期761-765,771,共6页Journal of Nanjing University of Science and Technology

基　　金：2011年江苏省研究生培养创新计划(CXZZ11_0271)

摘　　要：为了准确地定量预测加速度的峰值、脉宽和类型,针对自由式霍普金森压杆高过载实验,提出了一种基于径向基函数(RBF)神经网络的过载加速度的预测方法。分别在不同的撞击速度和LY12硬铝整形片尺寸下开展波形整形试验,获得过载加速度数据,并进行提取和归一化处理,用于RBF网络的学习。利用随机选取的5个样本数据对网络进行测试。结果表明,该方法可以有效地根据撞击速度和铝整形片的尺寸来预测过载加速度的峰值和脉宽,预测结果真实可靠。In order to predict the peak,type and duration of acceleration more exactly,a novel method is proposed based on the radial basis function(RBF) neural network model for the free Hopkinson pressure bar overload technique.Pulse shaper experiments are carried out under different striking velocities and sizes of the LY12 aluminium shaper,and acceleration data are obtained and normalized to(0,1) and used for the RBF network learning.Five sample data are selected randomly to predict the acceleration peaks with the trained network.It is concluded that the presented network is credible to predict the peak and duration of acceleration according to the striking velocity and the size of the shaper.

关 键 词：冲击动力学 霍普金森压杆 加速度 径向基函数 

分 类 号：TQ019[化学工程] TP183[自动化与计算机技术—控制理论与控制工程]