摘要
针对热轧带钢层流冷却过程的复杂性,以国内某热轧厂层流冷却系统为例,分析了层流冷却系统的组成以及相应的空冷和水冷数学模型.采用神经网络与数学模型相结合的方法,对带钢实测卷取温度与目标值的偏差进行了预报,证明利用神经网络能较好预测卷取温度的偏差值,进而对数学模型中的参数进行调整,实现高精度的卷取温度控制.结果表明,卷取温度比传统数学模型控制的标准差降低了21.94%.
In view of the complexity of laminar cooling process of hot-rolled strip, the configuration and mathematical model of air and water cooling of the laminar cooling system in a domestic hot rolling mill were discussed. The model is combined with BP neural network to predict how much the measured value of strip＇s coiling temperature deviates from its target value. The predicted results prove that combining BP neural network with the model is beneficial to the deviation prediction and, further, to the readjustment of the relevant parameters in the model. Then it is available to control the coiling temperature with high accuracy. Compared with the standard deviation as controlled conventionally by a mathematical model only, the new approach shows that the deviation is decreased by 21.94%.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2005年第9期871-873,共3页
Journal of Northeastern University(Natural Science)
基金
国家自然科学基金资助项目(50227401)
关键词
热轧带钢
层流冷却
控制精度
神经网络
hot rolled strip
laminar cooling
control accuracy
neural network
