BP神经网络对斜轧穿孔轧制力的预测  

Prediction on rolling force of oblique rolling piercing based on BP neural network

作　　者：林伟路 丁小凤 双远华[1] Lin Weilu;Ding Xiaofeng;Shuang Yuanhua(College of Materials Science and Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学材料科学与工程学院

出　　处：《锻压技术》2018年第10期175-178,共4页Forging & Stamping Technology

基　　金：山西省留学基金资助项目(2017-084)

摘　　要：为了有效预测AZ31镁合金管坯在三辊斜轧穿孔变形区的轧制力,保证穿孔后镁合金管材的优良性能,借助MATLAB工具箱建立了三辊斜轧穿孔的BP神经网络模型。结合穿孔过程中影响轧制力的因素,从实际生产过程中抽取了325个数据作为试验样本;根据AZ31镁合金在不同穿孔参数下的变形特点和三辊穿孔的有限元结果分析,利用轧制力的经验公式实现了理论计算,并将预测结果与理论结果进行了对比。结果表明：实际值与计算值的误差为14%,网络预测的最大误差为5%,平均误差为2. 4%,最小误差为1.4%。因此,网络预报精度高,操作简洁,可以代替复杂的数学计算模型。In order to predict the rolling force of AZ31 magnesium alloy tube billet in the oblique rolling piercing deformation zone with three-roller effectively and ensure the excellent performance of magnesium alloy tube after piercing,the BP neural network model of oblique rolling piercing with three-roller was established by MATLAB toolbox. Combined with the factors affecting the rolling force in the piercing process,three hundred and twenty-five data extracted from the actual production process were applied into experiment simples.According to the deformation characteristics of AZ31 magnesium alloy with different piercing parameters and the finite element analysis result of three-roller piercing,the theoretical calculation was realized by the empirical formula of rolling force,and the predicted results were compared with the theoretical results. The results show that the error between the actual value and the calculated value is 14%,the maximum error of the network prediction is 5%,the average error is 2.4% and the minimum error is 1.4%. Thus,the network prediction has high precision and simple operation,and the complex mathematical calculation model can be replaced.

关 键 词：BP神经网络 AZ31镁合金 斜轧穿孔 轧制力 MATLAB工具箱 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]