基于CA-CAMC网络的轧制力自学习预报模型  

Self-learning prediction model of rolling force based on CA-CAMC network

作　　者：赵文姣 闫洪伟 杨枕 温玉莲 孙祖乾 

机构地区：[1]首钢京唐钢铁联合有限责任公司冷轧部,河北唐山063200

出　　处：《冶金自动化》2016年第2期7-10,共4页Metallurgical Industry Automation

摘　　要：为了提高轧制力自学习模型的预报精度,将传统自学习模型的预报轧制力及影响轧制力的主要因素作为网络的输入,利用权值更新次数的倒数与单个样本本次激活的地址更新次数倒数和的比作为网络权值更新的信度,建立了基于信度分配的小脑模型CA-CAMC网络与轧制力自学习相结合的轧制力预报模型。通过大量在线数据实验分析,结果表明基于CA-CAMC网络模型的轧制力预报模型的精度高、稳定性好,能够更好地满足实际生产中越来越高的控制精度需求。In order to improve prediction precision,the rolling force of traditional self-learning model and the major factors affect rolling force are used as inputs of the CA-CAMC network,and the ratio of the reciprocal of address cell update times and the reciprocal sum of update times of all address cells that are activated by single sample is defined as the credit of network weight update,then the model of rolling force prediction is established by combining the traditional rolling force self-learning model and the CA-CAMC networks based on credit assignment. The experiment shows using the proposed model in the paper can attain a better accuracy and stability,it can better meet the increasingly high demand of control accuracy.

关 键 词：轧制力 自学习模型 CA-CMAC网络 信度分配 预报模型 冷轧 

分 类 号：TG334.9[金属学及工艺—金属压力加工]