应用本构模型和神经网络模型预测铝/镁基纳米复合材料的高温流变行为(英文)

Application of constitutive and neural network models for prediction of high temperature flow behavior of Al/Mg based nanocomposite

为了预测Al/Mg基纳米复合材料的高温流变行为,在不同的应变速率(0.01-1.0s-)和温度(523,623和1723K)的条件下进行热压缩试验,利用所得到的应力-应变数据,开发了本构模型,比如一般流动方程。阿累尼乌斯双曲模型、Johnson-Cook(JC)和改性的Zerilli-Armstrong(ZA)模型及人工神经网络(ANN)模型。通过使用统计参数,例如均方根误差(RMSE)、回归系数(R2)、平均相对误差(MRE)和分散指数(Is),比较了人工神经网络和不同的本构模型。结果表明,人工神经网络模型对AA5083-2%TiC复合材料的热变形流动应力的评估准确性更高。

To predicate the high temperature flow behavior of Al/Mg based nanocomposite, constitutive models such as general flow, Arrhenius hyperbolic, Johnson-Cook(JC) and modified Zerilli-Armstrong (ZA) models, and artificial neural network(ANN) models were developed using stress-strain data collected from hot compression tests carried at different strain rates (0.01-1.0 s?1) and temperatures (523, 623 and 723 K). The validity of the models developed was tested using statistical parameters such as root mean square error (RMSE), regression coefficient (R2), mean relative error (MRE) and scattered index (Is). A comparison between ANN and different constitutive models shows that the ANN model has a higher accuracy in estimating the flow stress during hot deformation of AA5083/2%TiC nanocomposite.