用递阶控制思想改进基于结构的神经网络模型  

Improving the Architecture-Based Neural Network Model by Using Hierarchical Control Theory

作　　者：杨海威[1] 詹永麒[1] 施光林[1] 乔俊伟[1] 

机构地区：[1]上海交通大学机械与动力工程学院,上海200030

出　　处：《上海交通大学学报》2004年第8期1369-1372,共4页Journal of Shanghai Jiaotong University

摘　　要：分析了基于结构的神经网络建模方法,应用递阶控制的思想对基于结构的神经网络模型进行了改进.在原模型中增加了一个协调器,对每一时刻各子系统输出修正后再进行下一时刻计算.建立了52SFZ-140-270B型液压缓冲器基于结构的神经网络模型.测试结果表明,改进后的模型减小了子系统误差及子系统间的相互影响,提高了系统的整体模型精度.The modeling method using architecture-based neural network was analyzed. The architecture-based neural network model was improved by using hierarchical control theory. A coordinator is added into thee original model, each sub-system's output is modified before the next time's computing. The architecture-based neural network model of 52SFZ-140-207B type hydraulic bumper was established. The test result shows that the improved model can reduce the error of sub-systems and their interaction and the accuracy of the system model is raised.

关 键 词：神经网络 递阶控制 结构 液压缓冲器 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]