一种改进的Elman网络及其在载荷识别中的应用  

IMPROVED ELMAN NETWORK AND ITS APPLICATION IN LOAD IDENTIFICATION

作　　者：田燕[1] 康海英[1] 唐力伟[1] 郑海起[1] 

机构地区：[1]军械工程学院兵器测试中心,石家庄050003

出　　处：《机械强度》2006年第z1期26-29,共4页Journal of Mechanical Strength

基　　金：国家自然科学基金资助项目(50375157);军械工程学院科学研究基金资助重点项目(YJJXM05017)~~

摘　　要：提出一种在关联层与输出层之间加入反馈连接的改进的Elman网络,给出相应的网络结构与学习算法。该网络关联层除从隐含层接收反馈信号外,还从输出层接收反馈信号,通过调节它们之间的权值系数,增加关联层的作用,提高网络的逼近能力和动态性能。在此基础上,建立具有时延反馈的Elman网络载荷识别模型,并用于齿轮箱的载荷识别研究。试验结果表明,该网络模型具有收敛速度快、识别精度高的特点,为复杂动态系统的载荷识别研究提供一种新的思路。An improved Elman network, in which feedback connections are added between the output layer and the associate layer, is developed and the corresponding network structure and learning algorithm are derived. The associate layer receives feedback signals not only from the hidden layer but also from the output layer. By adjusting the weights of the network, the power of the associate layer is improved and the network can have better performance. And based on this, the Elman network model for gearbox load identification with time delay feedback is established. The experiment results show that the proposed network model has the characteristics of fast convergence and high accuracy. The research done in the paper can provide a new approach for load identification of complex dynamic system and has definite practical use value.

关 键 词：ELMAN网络 载荷识别 齿轮箱 振动 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TH113[自动化与计算机技术—控制科学与工程]