7050铝合金蠕变本构模型及其泛化能力研究

Creep Constitutive Model and Generalization Ability of 7050 Aluminum Alloy

目的研究7050铝合金在温度为121~163℃、外加载荷为170~250 MPa条件下的蠕变拉伸行为。建立7050铝合金蠕变本构模型,对比本构模型精度并研究其泛化能力。方法采用CTM504-A1高温蠕变持久试验机进行蠕变拉伸实验,基于蠕变曲线及数据处理,分析蠕变温度和外加载荷对蠕变拉伸行为的影响,采用人工神经网络模型和幂律方程构建7050铝合金蠕变本构模型。结果7050铝合金在温度为121~163℃、外加载荷为170~250 MPa时,蠕变速率为4.44×10^(-8)~1.09×10^(-5) s^(−1),蠕变应变和稳态蠕变速率随着温度的升高而增大,在温度为163℃、外加载荷为250 MPa以及温度为177℃、外加载荷为170~250 MPa条件下,出现了蠕变第3阶段的加速蠕变,依据幂律方程构建的本构模型预测值的平均相对误差为2.45%,相关系数为99.02%;人工神经网络本构模型的预测值的平均相对误差为0.96%,相关系数为99.98%。结论在温度为121~149℃、外加载荷为170~250 MPa时,合金具有良好的低温抗蠕变性能。通过验证分析,与幂律方程模型相比,人工神经网络模型的预测精度更高;人工神经网络模型具有较强的泛化能力,在蠕变参数外仍具有很高的预测精度。

The work aims to study the creep tensile behavior of 7050 aluminum alloy at 121-163℃and 170-250 MPa(external load),establish a creep constitutive model of 7050 aluminum alloy,compare the accuracy of the constitutive model and study its generalization ability.CTM504-A1 high temperature creep endurance testing machine was used to conduct creep tensile experiments.Based on the creep curve and data processing,the effects of creep temperature and external load on the creep tensile behavior were analyzed.A constitutive model for creep of 7050 aluminum alloy was constructed with the artificial neural network and the power-law equation.The creep rate of 7050 aluminum alloy ranged from 4.44×10^(-8) s^(−1) to 1.09×10^(-5) s^(−1) at 121-163℃and 170-250 MPa.The creep strain and steady creep rate increased with the increase of temperature.The accelerated creep of the third stage of creep occurred at 163℃and 250 MPa as well as 177℃and 170-250 MPa.The average relative error and correlation coefficient of the predicted value of the constitutive model based on the power law equation were 2.45%and 99.02%respectively.The AARE of the predicted value of the artificial neural network constitutive model was 0.96%and the correlation coefficient was 99.98%.The alloy has good low temperature creep resistance at 121-149℃and 170-250 MPa.Through verification analysis,the prediction accuracy of artificial neural network model is higher than that of power law equation model.The artificial neural network model has strong generalization ability and high prediction accuracy except creep parameters.