铁水预脱硫过程镁粉耗量预报模型  

Prediction Model of Magnesium Powder Consumption During Hot Metal Pre-Desulfurization

作　　者：战东平[1] 张慧书[1] 姜周华[1] 

机构地区：[1]东北大学材料与冶金学院,辽宁沈阳110004

出　　处：《中国冶金》2010年第1期9-12,共4页China Metallurgy

基　　金：国家高技术研究发展计划(863计划)资助项目(2007AA04Z194);国家自然科学基金资助项目(50804009)

摘　　要：针对铁水预脱硫过程的实际情况,采用Visual Basic 6.0进行编程,建立了网络结构为4-12-1,模型数据归一化范围均为[0,1]区间的BP神经网络铁水预脱硫镁粉耗量预报模型。模型确定铁水质量、铁水温度、初始硫含量、终点硫含量为输入参数。采用210炉数据进行模型训练,经46炉数据现场验证表明,模型预报结果误差有65.2%的炉次绝对值误差在0.04 kg/t以内,有91.3%的炉次绝对值误差在0.06 kg/t以内,平均绝对值误差为0.033 kg/t。本模型的预测结果较好地符合了实际生产情况。Based on the productive practice of a steel plant, adopted the back propagation （BP） algorithm with the network configuration of 4-12-1 and the range of normalization from 0 to 1, used Visual Basic 6.0 software, the prediction model of magnesium powder consumption during hot metal pre-desulfurization processing was established. Meanwhile, four parameters, which are the weight and temperature of hot metal, the initial and final sulfur content in hot metal, were selected as input parameters. The data of 210 heats were used as the training samples and the other 46 heats were randomly selected as the test samples. The results show that the prediction absolute errors of magnesium powder consumption less than 0.04kg/t and 0.06 kg/t are 65.2 percent and 91.3 percent of the total test heats respectively. The average absolute error is 0. 033 kg/t. The model greatly coincides with the actual production operation.

关 键 词：神经网络 铁水预处理 脱硫 模型 镁耗量 

分 类 号：TF704.3[冶金工程—钢铁冶金]