石钢高炉铁水含硅神经网络预报模型  

Neural Network Forecast Model of Silicon Content in BF Liquid Iron

作　　者：郭建斌[1] 

机构地区：[1]石家庄钢铁有限责任公司炼铁厂,河北石家庄050031

出　　处：《中国冶金》2007年第2期38-42,共5页China Metallurgy

摘　　要：按照现代控制理论,把高炉视为多输入单输出系统,利用人工神经网络方法,结合高炉生产实际,建立了石钢高炉铁液中硅神经网络模型;通过引入动态步长和“惯性项系数”提高了网络收敛速;并不断更新学习样本集,提高了铁液中硅预报的命中率。结果表明,在允许误差为0.1%时,命中率达到86.67%,可为高炉操作提供指导。The neural network model of silicon content in liquid iron of blast furnace in Shigang is established based on the modern control theory and the artificial neural network methods, in which the blast furnace is considered as a system of multi-input and mono-output. The convergence rate of network is increased by introducing ＂coefficient of inertia item＂ and dynamic step length. The renewal of the specimen collection enables the＂ network to improve the hitting ratio of silicon forecast. The hitting ratio reaches 86.67% at 0.1% of permissive error.

关 键 词：高炉 铁液硅 预报 神经网络 

分 类 号：TF533.21[冶金工程—钢铁冶金]