BP神经网络的弯辊力预设定优化  

Optimizing of Bending Force Preset Based on BP Neural Network

作　　者：常安[1] 邸洪双[1] 佟强[1] 白金兰[1] 阳代军[2] 

机构地区：[1]东北大学轧制技术及连轧自动化国家重点实验室,辽宁沈阳110004 [2]首钢技术研究院,北京100041

出　　处：《轧钢》2006年第5期55-57,共3页Steel Rolling

基　　金：教育部博士点基金资助项目(20050145021)

摘　　要：针对国内某厂六辊单机架可逆冷轧机预设定弯辊力与实际轧制过程中弯辊力偏差较大的问题,采用了BP神经网络对其进行优化,结果其精度比回归的弯辊力模型的精度有显著提高,可使弯辊力预设定精度控制在15%以内,并为实现弯辊力的闭环控制奠定了基础。Because of the bending force difference between the preset and practical rolling proccess, the method of BP neural network is used to optimize the bending force based on six-high single reversing cold mill in some factory of China. Its accuracy is higher than the one of regression model, which is among 15 % -Nowadays this conclusion is used in the factory and the close loop control of bending force can be realized based on it.

关 键 词：BP神经网络 弯辊力设定 可逆冷轧机 回归模型 

分 类 号：TG335.55[金属学及工艺—金属压力加工]