函数增强型神经网络应用于钢的冷弯性能研究

Study of Cool Bending Nature of Steel by using Functional - Enhanced Net to Pattern Recognition

函数增强型神经网络（FunctionalEnhancedNet，简称FEN）是一种无隐含层的新型网络，作者应用其三阶联合激励增强特性对钢的冷弯性能进行判别研究，识别准确率近100％．对预测集的每一个输入信号添加噪音干扰，发现依然能准确判别．网络误差下降随着添加噪音的增大而逐渐变慢，可见该网络的容错能力令人满意．

Functional - Enhanced net is a novel neund network(FEN), without hidden layer. We use the three order joint - activition of FEN to study the discrindnation of cool bending natures of steels.The discrimination is neary 100%. After adding noise to every input signal of the testing set, the discrimination retains accurate. With the noise incnaing, the falling of the mean sqamed ermr is slowr.This proves that the fault - toleamt ability of the FEN is satisfactory.