基于神经网络的热带钢连轧弯辊力预报模型  

Prediction model of bending force based on neural network in hot strip mill

作　　者：张敬伟[1] 何安瑞[1] 杨荃[1] 

机构地区：[1]北京科技大学高效轧制国家工程研究中心,北京100083

出　　处：《冶金自动化》2007年第6期20-22,共3页Metallurgical Industry Automation

摘　　要：针对传统弯辊力预设定模型的缺陷和带钢热连轧轧制特点,利用日照钢铁有限公司1580 mm七机架热轧机生产数据,对精轧机组进行了基于神经网络的弯辊力优化预报。基于神经网络的弯辊力预报模型与传统模型相比,可进行高度非线性模拟,以大量实际数据作为神经网络训练输入,有模型结构简单、容易实现等优点。基于神经网络的弯辊力预报模型不但考虑各种输入参数相互之间的影响作用,而且考虑到各机架输出之间的关系,可用于提高头部板形控制精度,并为实际弯辊力设定提供了指导和试验基础。To counter imperfection of traditional bending force prediction model and according to feature of hot strip mill, a bending farce prediction model based on neural network is presented for finishing mill train, based on production data of 1 580 mm seven-stand hot rolling mill in Rizhao Iron and Steel Company. Large practical data was used as training input of neural network. Comparing with traditional model, the model is simple in structure and easy to implement. In the model, not only interaction among various input parameters was considered, but also relations among different stands were considered. Its application can improve control accuraccy of crop shape, and provide direction and test foundation for practical bending force setting.

关 键 词：热轧 板形 弯辊力 神经网络 

分 类 号：TG334.9[金属学及工艺—金属压力加工]