基于高维输入小波神经网络的热连轧机产品质量模型(英文)  

Using High Dimension Wavelet Neural Network Model Product Quality for Hot Rolling Mill

作　　者：李换琴[1] 万百五[2] 

机构地区：[1]西安交通大学理学院,西安710049 [2]西安交通大学系统工程研究所,西安710049

出　　处：《工程数学学报》2006年第4期614-618,共5页Chinese Journal of Engineering Mathematics

基　　金：The National Natural Science Foundation of China (60574060).

摘　　要：小波神经网络是一种以小波函数为激励函数的神经网络。现有的小波神经网络局限于低维,本文提出一种适合高维输入的小波神经网络建模方法,对几种小波函数与学习算法进行了比较实验,成功地解决了32维输入的大型多辊热连轧机钢板材质量建模问题。Wavelet networks now available are usually limited to problems of small dimension input. In this paper, a wavelet-based neural network （WNN） was introduced for handling high dimension input problems. Several different wavelet functions and different algorithms have been tested and compared. Simulation results demonstrate that the B-spline wavelet function and the Levenberg-Marquardt algorithm are effective. The WNN is taken as the production quality model of large-scale hot steel rolling mill and a 32-input modeling problem is successfully solved.

关 键 词：小波神经网络 质量模型 高维输入 热连轧机 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]