基于人工神经网络的宽幅铝合金中厚板厚度预测模型  

The model of thickness prediction for wildaluminum medium plate based on artificial neural network

作　　者：杨如民[1] 汤爱涛[1,2] 佘加[1] 潘复生[1,2] 李江宇 

机构地区：[1]重庆大学材料科学与工程学院,重庆400030 [2]重庆大学国家镁合金工程研究中心,重庆400030 [3]广西柳州银海铝业股份有限公司,广西柳州545006

出　　处：《功能材料》2015年第6期6102-6105,6110,共5页Journal of Functional Materials

基　　金：科技部国家科技支撑计划资助项目(2012BAF09B04);科技部国际科技合作计划专项资助项目(2011DFR50950-03);科技部国际合作资助项目(2011DFR50950);支撑计划资助项目(2012BAF09B04)

摘　　要：厚度预测模型的精度是影响厚度控制的重要因素。针对本项目国内水平领先、最宽幅的"1+4"热连轧生产线,根据生产现场获取的5083宽幅铝合金中厚板实测数据,在研究分析关键影响因素的基础上,运用人工神经网络技术建立了铝合金宽幅中厚板厚度预测的BP神经网络模型。其相对误差在0.5%之内,高于已有模型预测精度,能实现高精度预报。应用模型预测了5052宽幅铝合金中厚板的出口厚度,结果表明,模型能较好的预测轧件厚度的变化,有很好的泛化能力。The accuracy of the thickness prediction model is an important factor which was influenced the thickness control. Aiming at the '1 + 4'hot tandem rolling line which was domestic leading and most wide,According to the processed measured data of as-rolled 5083 wild aluminum medium plates,a thickness prediction model was developed by artificial neural network based on the analysis of key factors. The relative error of the model is within 0. 5%,which is better than that of previous models. And the high-precision prediction for rolling thickness was achieved. The developed model was successfully to predict the thickness of 5052 wild aluminum medium plate and exhibited good generation ability.

关 键 词：人工神经网络 热连轧 中厚板 厚度预测 

分 类 号：TG333.7[金属学及工艺—金属压力加工] TP18[自动化与计算机技术—控制理论与控制工程]