模糊规则细化的神经网络方法  

Neural network method on fuzzy rule densification

作　　者：冀春荣[1] 何钺[1] 李燕[2] 

机构地区：[1]西安交通大学,710049 [2]西安工业学院

出　　处：《机床与液压》1998年第4期8-10,共3页Machine Tool & Hydraulics

基　　金：兵总九五预研基金!YJ96092;陕西省自然科学基金!96C27

摘　　要：系统分析了模糊控制器I／O接口的量化误差，构造了控制器输入变量矩形域，通过对变量矩形域网络结点所对应的模糊规则进行｛0，1｝编码，利用单隐层前馈网络BP算法的学习功能，完成了控制器输入｛E，EC｝落在矩形域内任一点时输出变量｛U｝的调整计算，细化了模糊规则，减少了从变量连续论城到模糊论域映射时非一一对应所造成的控制误差，提高了控制精度。Quantitative errors for fuzzy controller I/O ends are systentatically analyzed and the input variable rectangular domain is constructed. By 0.1coding on fuzzy rules associated with the node of the variable rectangle domainand using the leaming function of single hidden layer feedforward network BP algorithm, the adjusting calculation of output variables U is accomplished when thecontroller input E, EC fall onto a arbitrary point in the rectangular domain.Thefuzzy control rule is densified and the control error resulted from Not one-toone correspondence is reduced when the variables shift from continuous scope.This improves the control accuracy.

关 键 词：模糊规则 量化误差 神经网络 模糊控制器 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程] TP18[自动化与计算机技术—控制科学与工程]