摘要
针对传统的回归方法的某些不足 ,采用了人工神经网络的方法预测力学性能。从宝钢 2 0 5 0热轧管理机中随机抽取数据 ,用人工神经网络中的BP网络建立原始化学成分和热轧生产的主要工艺参数与产品力学性能之间的关系。离线仿真表明 ,产品力学性能的预报值与实际值拟合良好 ,预报结果的相对误差很小 ,屈服强度相对误差 88%在± 4 %以内 ,抗拉强度的相对误差 86 %在± 2 %以内 ,伸长率的相对误差 78%在± 6%以内。
In order to improve the prediction accuracy,the artificial neural networks (ANN) are used to predict mechanical properties On the basis of the data collected randomly from Baosteel 2050 hot rolling supervisor, using BP algorithm of ANN, the relation between chemical composition and main hot rolling parameter and mechanical properties of product has been built The off line simulation indicates that predicted and measured results are in good agreement The relative error is very low for 88 % yield strength results, the error is within ±4 %,for 86 % tensile strength results within ±2 % and for 78 % elongation results within ±6 %
出处
《钢铁》
CAS
CSCD
北大核心
2002年第7期37-40.5,共4页
Iron and Steel
基金
国家"973"重大基础研究项目 (19980 615 0 9)
