摘要
利用现场的数据 ,采用 BP神经元网络预报热连轧层流水冷区集管组内的基本热流密度 ,将预报的结果用于上、下集管组的热流密度的数学模型计算 ,进而优化层冷集管组的水冷温降计算数学模型的精度。将结果与采用多元回归方法所得到的结果作比较 ,表明采用 BP神经元网络计算基本热流密度的精度要高于多元回归方法的计算精度 ,卷取温度的计算值与实测值的标准差比解析回归方法减少了近 2 0 % ,说明该方法具有良好的在线应用前景。
By use of BP neural networks the basic heat-flux density for laminar cooling system of hot strip mills was predicted on the basis of measured date,the result was then applied to calculate the top and bottom heat-flux density for cooling banks to improve the precision of calculation of temperature drop in each bank.In the mean time the basic heat-flux density was calculated by using traditional linear analytical regression program.The results showed that the calculation precision of neural networks is higher than that of analytical method,The standard deviation between the predicted and measured coiling temperature was reduced about 20 %.
出处
《钢铁》
CAS
CSCD
北大核心
2004年第3期29-33,42,共6页
Iron and Steel
基金
国家自然科学基金资助项目 (5 0 10 40 0 4
关键词
热轧带钢
层流冷却
卷取温度
热流密度
数学模型
多元回归
BP神经网络
laminar cooling,coiling temperature,basic heat-flux density,mathematical model,analytical regression,BP neural network
