摘要
为准确预测复合喷吹铁水预脱硫终点硫含量,降低成本并减少二次喷吹,将多元回归模型、增量模型和神经元网络模型相结合,开发了复合喷吹铁水预脱硫综合预报模型。综合模型以增量模型为基础,利用神经元模型动态检索参照炉次,并用多元回归模型进行增量计算及结果筛查,比单一模型的偏差比例降低至少50%以上,提高了模型的准确性和稳定性。其在河钢集团唐钢不锈钢公司实际生产中获得了良好应用,钝化镁消耗率降低10 kg/炉,脱硫一次处理的终点命中率达8%以上,取得了显著的经济效益。
In order to predict the final sulfur content for hot metal pre-desulfurization,reduce the cost and reduce the secondary spraying,the composite pre-desulfurization forecast model is developed by combining multiple regression model,incremental model and neural network model.The comprehensive model is based on the incremental model,adopts neuron model to dynamically retrieve reference heats,utilizes the multivariate regression model to carry out incremental calculation and screen the results.The deviation ratio of the single model is reduced by at least 50%,which improves the accuracy and stability of the model.This model has been applied in HBIS Tangsteel stainless company.The consumption rate of passivated magnesium is reduced by 10 kg/furnace,and the final hit rate of desulfurization treatment is more than 8%,which has achieved remarkable economic benefits.
作者
高宇
李阳
巨伟峰
李健
张燕平
肖加海
贾新风
张明
Gao Yu;Li Yang;Ju Weifeng;Li Jian;Zhang Yanping;Xiao Jiahai;Jia Xinfeng;Zhang Ming(HBIS Group Research Institute,Shijiazhuang,Hebei,050023;HBIS Group Tangsteel Stainless Steel Company,Tangshan,Hebei,063100)
出处
《河北冶金》
2020年第3期7-10,共4页
Hebei Metallurgy
关键词
铁水预处理
复合喷吹
多元回归模型
增量模型
神经元模型
标准偏差
终点命中率
hot metal pretreatment
compound injection
multiple regression model
incremental model
neuron model
standard deviation
end point hit rate
