基于物理参数和BP神经网络的9310钢本构模型研究

Research on constitutive model of 9310 steel based on physical parameters and BP neural network

采用Gleeble-3800热模拟试验机对9310钢进行了变形量为70%的等温恒应变速率压缩实验,在变形温度为800~1200℃、应变速率为0.01~50 s^(-1)的范围内研究了9310钢的热变形行为。通过不同热变形参数对自扩散系数D和杨氏模量E的影响,建立了基于物理参数的本构模型,同时基于实验数据构建了BP神经网络本构模型。结果表明:9310钢为负温度正应变速率敏感性材料,且流动应力随变形温度的升高和应变速率的降低而减小。基于不同条件构建的物理本构模型和BP神经网络模型的相关系数r均大于0.98,但BP神经网络模型的r值可达0.996,平均绝对相对误差为3.1%。经过流动应力曲线、相关系数和平均绝对相对误差的综合对比,得出BP神经网络模型对预测9310钢的流动行为具有较好的适用性。

The Gleeble-3800 thermal simulator was used to conduct isothermal constant strain rate compression experiments on 9310 steel with deformation amount of 70%.The hot deformation behavior of 9310 steel was studied within the range of deformation temperature of 800-1200℃and strain rate of 0.01-50 s^(-1).Through the influence of different thermal deformation parameters on self-diffusion coefficient D and Young′s modulus E,the constitutive model based on physical parameters was established,and the BP neural network constitutive model was constructed based on experimental data.The results show that 9310 steel is negative temperature and positive strain rate sensitive material,and the flow stress decreases with the increase of deformation temperature and the decrease of strain rate.The correlation coefficient r of the physical constitutive model and the BP neural network model constructed under different conditions is greater than 0.98,but r value of the BP neural network model can reach 0.996,and the average absolute relative error is 3.1%.After comprehensive comparison of flow stress curve,correlation coefficient and average absolute relative error,it is concluded that the BP neural network model has good applicability for predicting the flow behavior of 9310 steel.