应用神经网络表征变形高温合金的本构关系

Using artificial neural network to characterize constitutive relationship for wrought superalloys

本构关系是联系塑性加工过程中材料的动态响应与热力参数之间的媒介，因此，本构关系是用有限元对金属塑性加工过程进行数值模拟的前提条件。由于变形高温合金本构关系的高度非线性，在试验数据基础上，按Ａｒｒｈｅｎｉｕｓ型方程用数理统计方法建立的本构关系，结构十分复杂，不便工程应用。本文探讨用人工神经网络这种新一代信息处理工具表征变形高温合金本构关系的方法和效果。研究结果表明，应用人工神经网络表征的本构关系比用数理统计方法建立的能较宏观地反映材料在塑性加工过程中的动态行为，是一种有广泛应用前景的表征工程材料本构关系的方法。

Constitutive relationship is the medium between dynamic response of materials during plastic working processes and thermomechanical parameters, so it is regarded as the prerequisite for simulation of forming processes by using finite element method.Because of its high nonlinearity, the constitutive relationship for wrought superalloys formulated by mathematical statistics according to Arrhenius type equations is not convenient to incorporate into finite element software due to complexity of its structure.In this paper an attempt is made to establish the constitutive relationship for wrought superalloys using artificial neural networks (ANN) which is a new tool to process information.It is found that the predictions of constitutive relationship characterized by ANN are in more close agreement with experimental data than that of the constitutive equation established by mathematical statistics.It shows the promising prospect of using ANN to characterize the constitutive relationship for engineering materials.