摘要
在基于WinCE操作系统的嵌入式平台上,实现基于RBF神经网络的工业过程软仪表。应用无监督K均值聚类确定聚类中心,用可避免矩阵求逆运算的递推最小二乘法训练网络权值的两阶段学习算法训练网络。进一步构造双重RBF网络结构,用一个RBF神经网络训练样本,用另一个独立的RBF神经网络训练误差提高精度。以嵌入式工业计算机作为硬件平台,在基于WinCE的嵌入式系统上用EVC编程实现。在某炼油厂轻柴油凝固点的工业现场,用该软仪表实测数据进行仿真测试,取得了较好的效果。
Implement problem of the RBF neural network soft sensor on WinCE embedded system is discussed. By using K-means procedure, the cluster center is established and recursion algorithm is given to train the network weight value. A Double network structure is built to improve the network accuracy. One is used to train sample data and the other is used to train network error. To implement this soft sensor instrument, industrial embedded computer is used for hardware platform, and the software is programmed with Embedded visual C ＋ ＋. This method is used to estimate the freeze point of diesel oil in modeling data and good results are obtained.
出处
《控制工程》
CSCD
2006年第6期536-539,共4页
Control Engineering of China
