一种基于遗传神经网络的热连轧带钢弯辊力预报方法  

The Prediction Method of Bending Force Based on Genetic Algorithms and Neural Networks in Hot Strip Mills

作　　者：张敬伟[1] 何安瑞[1] 杨荃[1] 张倩[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：《冶金设备》2007年第6期22-25,共4页Metallurgical Equipment

摘　　要：针对传统板形模型缺陷和带钢热连轧轧制特点,提出基于遗传神经网络的弯辊力优化预报方法。利用1580mm热轧Q235B钢种实际生产数据对网络进行训练和测试,并对结果进行分析比较。结果表明:遗传神经网络预报能力不但优于传统弯辊力设定模型,而且优于单纯神经网络的预报能力,能较准确预报热连轧带钢弯辊力。基于遗传神经网络的弯辊力预报模型可提高头部板形控制精度,有实际应用的价值。Based on the shape disfigurement of tradition math model and feature of hot strip mills,a meth- od using genetic algorithms and artificial neural networks is used to optimize the prediction of bending force.The net- works were trained and tested by the Q235B steel's data in RiZhao 1580mm hot rolling mill.And the results are ana- lyzed and compared.The prediction of genetic algorithms and artificial neural networks not only exceed tradition math model,but also exceed the prediction of neural networks.It can improve the precision of bending force and shape. So it is worthful in real-life rolling.

关 键 词：弯辊力 遗传算法 神经网络 热轧 板形 

分 类 号：TG333[金属学及工艺—金属压力加工]