【摘要】 利用Split Hopkinsonbar装置上所得变形数据,研究并比较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.在此基础上比较研究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.本研究中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%.研究发现:Zerrilli Arm strong模型相对于神经网络模型有较低精度的原因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.研究认为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.图2,表1,参12.更多还原

【Abstract】 Data from the deformation on Split-Hopkinson bar are used for constructing an artificial neural network model and an Zerrilli-Armstrong model for shock-prestrained copper. Predictions by the two models are compared. It is found that the artificial neural network model is more accurate(δ=0.9%) for engineering use of copper deforming at high strain-rate than Zerrilli-Armstrong model(δ=8%) Since the former model has considered the factors varying dynamically in the deforming process which is regarded as constant in the latter one. The research shows that artificial neural network model can widen it’s engineering application to dynamical calculation of copper by consideration of more input nodes.2figs.,1tab.,12refs.更多还原