多层前向神经网络在电动转向系统中的应用  

Application of Multilayer Foreword Neural Network on the Electric Power Steering System

作　　者：张海华[1] 许镇琳[1] 王豪[1] 尚喆[1] 

机构地区：[1]天津大学电气与自动化工程学院,天津300072

出　　处：《公路交通科技》2004年第8期121-123,共3页Journal of Highway and Transportation Research and Development

摘　　要：针对电动转向系统助力特性存在非线性的特点,应用多层前向神经网络,对电动转向系统全车速下的助力特性进行拟合。文中设计了多层前向神经网络结构,并采用基于综合目标函数的二阶学习算法对其进行离线学习,最后在电动转向综合试验台上实现在线控制。系统实现了全车速范围的非线性转向助力,克服了转向助力盲区。Considering the nonlinear characteristic of electric power steering (EPS) system's power control characteristic curve,Multilayer Foreword Neural Network is applied to fitting the power conlrol characteristic curve at any speed in this paper.The Multilayer Foreword Neural Network structure is designed.And the second-order learning algorithm,which is based on generalized cost criterion,is adopted to do off-line learning.Finally,the on-line control is realized on the EPS experiment rig.Then the full-speed nonlinear steering power control curves are generated and the blind zone is overcome.

关 键 词：电动转向 助力特性 多层前向神经网络 

分 类 号：U463.4[机械工程—车辆工程]