摘要
基于天津天铁冶金集团30t转炉炼钢实际生产数据，首先建立了转炉炼钢终点静态控制的吹氧量及矿石用量统计模型，其预测100个炉次吹氧量和矿石用量平均相对误差分别为0．58％及10．4％。考虑到影响终点钢水温度和碳含量的因素比较复杂，设计了预测钢水终点温度和碳含量的人工神经网格模型，利用Levenberg-Marquardt算法和257个炉次的实际生产数据进行了模型训练，并对另外100个炉次的终点钢水温度及碳含量进行了预测，在终点钢水温度为1646-1698℃和终点碳质量分数为0．033％～0．128％的范围内，得到的终点碳温双命中率为55％。
On the basis of the practical production data of the 30 t converter steel-making process in Tianjin Tiantie Metallurgical Group Co. , ltd, a statistic model for prediction of the oxygen blow amount of the end-point static control BOF process and ore burden amount is established and the prediction accuracies of the oxygen blow amount an ore burden amount for over 100 heats are 0. 58 % and 10. 4 % respectively. In view of the complicated factors affecting the end-point temperature and carbon content of the liquid steel and artificial neural network based prediction model is designed and established for measurement of the end -point temperature and carbon content of the liquid steel and then simulation training is also carried out by way of Levenberg-Marquardt algorithm on the basis of the practical production data of over 257 heats. In addition, the end-point temperature and carbon content of the liquid steel of other 100 heats are also predicted. Within the controlled range of 1646℃ to 1698℃ of the end-point temperature and the end-point carbon content of 0.033 %and 0. 128 % the hit rate of the accurate prediction of both the end-point temperature and carbon content of the liquid steel attains to 55 %.
出处
《炼钢》
CAS
北大核心
2006年第1期45-48,共4页
Steelmaking
基金
天津大学"复杂过程检测与控制"创新平台资助项目(01BK-098-02-08)
关键词
转炉炼钢
终点控制
预测模型
converter refining process
end-point control
prediction model
