【摘要】 根据铁水预处理的实际生产工艺,利用BP神经网络建立了铁水预处理脱硫时铁水中氮含量的预报模型。利用生产数据对网络进行训练后,可以用于预测铁水预处理脱硫后铁水中氮含量,预报相对误差可以控制在22%以下,w[N]的绝对误差值控制在8×10-4%以内.给各输入参量增加扰动后,模型计算表明,初始温度和初始[N],[C],[P]含量,四因素影响最大,且温度和[C]含量为负扰动,[N]为正扰动,说明增加温度和碳含量有利于脱氮.更多还原

【Abstract】 According to the practical production process of hot metal pretreatment,a model was established to predict the nitrogen content of hot metal in desulphurization pretreatment by BP artificial neural network.After training the network using production data,the model can predict the nitrogen content of hot metal in desulphurization pretreatment,the relative prediction error can be controlled under 22%,and the absolute error is below 8×10-4%.Changing the values of input parameters and calculating,the result indicates that the initial temperature nitrogen content,carbon content and phosphorus content have the maximum influence.The temperature and carbon content are of negative effect,and the nitrogen content has positive effect.High temperature and carbon content are favorable to denitrification.更多还原