摘要
水泥窑优化控制建模问题一直是自动化界急于解决的难题。文章利用神经网络的非线性映射、特征抽取和容错能力来解决水泥窑优化控制建模问题,建立了基于BP网络的窑优化操作二层回路控制方案,并针对在BP网络训练中可能出现的一些问题如收敛、局部极小、隐层节点数确定和训练样本集对训练影响等问题进行了分析讨论,并采用了相应的训练策略,以保证训练后网络具有较好的记忆和归纳能力。仿真结果证明了采用这种策略进行训练的正确性。
Because it is difficult to control the cement kiln and to set up mathematical model on the basis of tradi-tional or modern control theories,the authors put forward a sort of two layer loop circuit of optimal control scheme in cement kiln based on BP networks for its nonlinear mapping,character take-out and error tolerance.Some problem,such as premature saturation or local minimum,the number of the hidden layer in the process of training and so on,are not resolved,this paper takes out some practical strategies and methods to reduce the training time and to improve the a-bility of memory and epagoge.Simulation results declare that the strategies and methods are effective in the process of training,and the calculated results which are compared with factual values are satisfactory.
出处
《计算机工程与应用》
CSCD
北大核心
2002年第14期235-238,共4页
Computer Engineering and Applications
基金
国家自然科学基金青年基金(编号:60103021)资助项目
