石钢高炉铁水含硅量神经网络预报模型  

NEURAL NETWORK PREDICTION MODELS FOR CONTENT OF Si IN SHIGANG BLAST-FURNACE HOT IRON

作　　者：郭建斌[1] 郭建国[2] 

机构地区：[1]石家庄钢铁有限责任公司 [2]河北经贸大学计算机系

出　　处：《冶金丛刊》2006年第5期4-7,10,共5页Metallurgical Collections

摘　　要：本文按照现代控制理论,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水含硅量BP神经网络模型。通过引入动态步长和“惯性项系数”提高了网络收敛速度。采用“修正式”预报模式提高了铁水含硅量预报的命中率。结果表明:在允许误差为0.1%时,命中率达到了86.67%,可以为高炉操作提供指导。According to modem control theory, blast furnace was regarded as a multiple input single output system. Combined with the production experience, the BP neural networks were used to predict the content of Si in blast-furnace hot iron. The rapidity of convergence was improved with introducing dynam- ic step size and inertial coefficient, and the prediction precision was improved with introducing modified prediction model. The results showed that the prediction value was 86.67% under the permissible error 0.1%.

关 键 词：高炉 铁水含硅量 预报 神经网络 

分 类 号：TF53[冶金工程—钢铁冶金] TP183[自动化与计算机技术—控制理论与控制工程]