摘要
针对冷轧过程中难以建立精确的板形控制数学模型的问题,设计了RBF模糊神经网络模型用于冷轧板带过程中的板形控制。基于一种无监督聚类方法确定了模糊规则数和网络初始参数以建立网络初始模型;采用BP算法训练网络,使网络能够快速收敛;同时,为使所建立的控制系统能够应用于在线控制,增加了模糊规则在线自学习功能,避免网络过度修正权值。仿真实验表明,所建立的RBF模糊神经网络模型控制精度高,满足板形在线控制要求。
An RBF fuzzy neural network controller was proposed for sheet shape control because it was hard to set up an accurate mathematics model in cold rolling process. The number of fuzzy rule, the initial parameters and initial model of the RBF fuzzy network were determined by one cluster method. BP algorithm was used to train the network for it could enable the network to have a rapid convergence. At the same time, the function of self-correcting fuzzy rule was increased to enable the control system built on-line, avoiding the network over corrected weight. The simulation result shows that the control effect is excellent, reaching the goal of the intelligence control for shape on-line.
出处
《重型机械》
2008年第2期5-9,共5页
Heavy Machinery
基金
国家自然科学基金资助项目(50675186)
关键词
板形控制
模糊神经网络
聚类
BP算法
flatness control
fuzzy neural network
cluster
BP algorithm