作　　者： ; ; ;

机构地区： 湘潭大学

出　　处： 《湖南大学学报（自然科学版）》 2004年第6期23-27,共5页

摘　　要： 利用Split Hopkinsonbar装置上所得到的变形数据,研究了冲击预变形铜的神经网络本构关系模型以及Zerrilli Armstrong本构关系模型,比较了两种模型对冲击预变形铜在不同热力学状态下流变应力的预测精度.研究表明:神经网络模型的总的拟合度为0.9%,而Zerrilli Armstrong模型的拟合度为8%;Zerrilli Armstrong模型相对于神经网络模型精度较低,是由于物理模型把材料内部某些动态变量作为常数处理,而神经网络模型建模训练时已经包含了这些动态变化的因素;通过增加神经网络输入节点数可以扩大神经网络模型的应用范围. Data from the deformation on Split-Hopkinson bar were used to construct an artificial neural network model and a Zerrilli-Armstrong model for shock-prestrained copper. The predicting precison of the two models were compared. It was found that the artificial neural network model was more accurate(8%) for the engineering use of copper deforming at high strain-rate than Zerrilli-Armstrong model(0.9%),as the former considered the factors varying dynamically in the deforming process, which was regarded as constant in the latter method. The research shows that the artificial neural network model could widen its application in engineering for the dynamical calculation of copper through the consideration of more input nodes.

关 键 词： 冲击预应变 本构关系 人工神经网络模型 模型

领　　域： [金属学及工艺] [金属学及工艺]