一类新的RBF神经网络在非线性系统建模中的应用刘妹琴.pdf

DOI ：10.13195/j.cd.2001.03.21.liumq.005 第16 卷第3 期 2001年5 月 Vol. 16 No . 3 CON TR OL A N D DE CIS I ON 　 May 2001 : 1001-0920(2001) 03-277-05 RBF 刘妹琴, 沈　轶, 廖晓昕 ( , 430074) : 提出一种基于输入集分类函数的新的距离度量方法, 它与前传 归的正交最小二乘法相结合, 不仅可以学习分类超平面的参数, 而且可以选择重要的输入节点。这种结构的RBFNN 特别适用于非线 性动力学系统的辨识( 建模) 和控制。将改进的RBFNN 用于化工中的聚合反应过程建模, 结果表明该方 法是有效而适用的。 : 径向基函数神经网络( RBFNN) ; 分类超平面; 正交最小二乘法( OLS) ; 聚合反应过程 : TP 1 3: A Application of a Class of New RBF Neural Networks to Modelling Nonlinear Systems L I U M ei-qin, SH EN Yi, L IA O X iao-x in ( Department of Cont rol Science Engineering, Huazhong University of Science Technology, Wuhan 430074, China) : Abstract A new distance met ric based on a classification function of the set of input vect ors is ad- . ( ) , vanced Combining the new measure w it h the forw ard regression orthogonal least square OLS not only the paramet ers of the classification hyperplane, but also the important input nodes can be obtaind. T he im proved RBFNNs are suitable to identifying ( modelling) or controlling nonlinear dynamic sys- . tems Experimental results show that the im proved RBFNN has successfully modelled the polymeric re- action process in the chemical industry. Key words: radial basis function neural network( RBFNN) ; classification hyperplane; orthogonal least square( OLS) ; polymeric reaction process 1引　　言