Ti-5Al-2Sn-2Zr-4Mo-4Cr合金本构关系的新模型

本构关系是联系材料塑性变形行为与工艺参数的桥梁，也是用有限元法模拟金属成形过程的前提条件．开发了一个基于神经网络的Ｔｉ－５Ａｌ－２Ｓｎ－２Ｚｒ－４Ｍｏ－４Ｃｒ合金的本构关系模型．首先利用Ｔｈｅｒｍｅｃｍａｓｔｏｒ－Ｚ型热模拟机等温压缩Ｔｉ－５Ａｌ－２Ｓｎ－２Ｚｒ－４Ｍｏ－４Ｃｒ合金，研究在不同变形温度、变形程度和应变速率等工艺参数条件下流动应力的变化情况．然后用实验所得的热变形工艺参数与性能间的数据来训练具有ＢＰ算法的人工神经网络．训练结束后的神经网络即成为一个知识基的本构关系模型．预测结果和实验结果对比表明神经网络模型具有较高的精度和推广能力．

New Constitutive Relationship Model of Ti-5Al-2Sn-2Zr-4Mo-4Cr Alloy

Artificial neural networks were applied to acquire the consititutive relationship model of Ti 5Al 2Sn 2Zr 4Mo 4Cr(wt%) alloy. First, an isothermal compressive experiment using Thermecmastor Z hot simulator was studied to acquire flow stress at different deformation temperatures, equivalent strains and equivalent strain rates. Then, a feed forward neural network was trained by using the experimental data. After the training process, the neural networks became a knowledge based constitutive relationship system. The comparison of the predicted and experimental results shows that the ANN model has good learning precision and good generalization. The neural network methods show much better agreement than the existing methods with the experiment data and have advantage of being able to deal with noisy data or data with strong non linear relationships.