论文部分内容阅读
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。
According to the actual situation of hot metal pre-desulfurization process, Visual Basic 6.0 was used for programming. A BP neural network pre-desulfurization magnesium powder with network structure of 4-12-1 and normalized range of model data were all [0,1] Consumption forecast model. The model determines the hot metal quality, the hot metal temperature, the initial sulfur content, and the end sulfur content as input parameters. The data of 210 furnaces were used to train the model. The field data of 46 furnaces showed that the accuracy of the model prediction was 65.2%, the absolute error of the furnace was within 0.04 kg / t, and the absolute error of 91.3% of the furnaces was 0.06 kg / t Within the mean absolute error of 0.033 kg / t. The predictions of this model are in good agreement with the actual production.