以温度为523～723 K、应变速率为0.01～10 s-1以及总应变为0.69的实验数据为基础,应用神经网络建立具有外推能力的7050铝合金本构模型,实现了在大应变速率下(10～50 s-1)流变应力的预测,并基于实验数据和预测数据,建立了7050铝合金大应变速率的逐步回归模型。误差结果分析表明,回归结果与实验数据及预测结果之间的平均相对误差不超过6.5%,能够准确地预测7050铝合金的流变应力,适用于分析热变形过程中的问题。 Based on the experimental data of temperature 523 ~ 723 K, strain rate 0.01 ~ 10 s-1 and total strain 0.69, the 7050 aluminum alloy constitutive model with extrapolation capability was established by using neural network, (10 ~ 50 s-1), and based on the experimental data and forecasting data, a stepwise regression model of the large strain rate of 7050 aluminum alloy was established. The error results show that the average relative error between the regression results and the experimental data and the predicted results does not exceed 6.5%, which can accurately predict the flow stress of the 7050 aluminum alloy and is suitable for analyzing the problems in the thermal deformation process.