摘要
由于受轧机自身特性变化等复杂因素影响,调控机构的影响系数不适合用恒定的常量来描述。将小波神经网络应用到影响系数的自学习过程,对预设定的影响系数值进行在线修正。介绍了冷连轧板形调控机构影响系数自学习的神经网络结构设计,结合目标能量函数的最小化,对影响系数自学习算法进行分析。结合生产现场的实际板形数据,采用Visual C++/MATLAB对控制算法的作用效果进行仿真。仿真结果表明,自学习算法对板形控制起到了预想的效果,具备现场在线运行的可行性。
The influencing factors did not be exactly described by constant values because of complex factors such as characteristics of cold rolling mill and so on. The wavelet neural network was applied in self-learning process of factors to modify the values online. The self-learning wavelet neural network structure of efficiency faetor for flatness control tandem cold rolling was introduced. With the minimization of target function, the self-learning algorithm of efficiency was analyzed. With the combination of actual flatness data in site, simulation for the control algorithm was made effective with Visual C ＋ ＋/MATLAB. Results revealed that the expected effect was attained with the self-learning algorithm, and it＇s feasible for site online operation.
出处
《上海金属》
CAS
2009年第5期38-41,共4页
Shanghai Metals
关键词
神经网络
自学习
效率因子
板形控制
Neural Network, Self-Learning, Efficiency Factor, Flatness Control
