Abstract

The hot deformation behavior of coarse-grained (CG), ultrafine-grained (UFG), and oxide dispersion-strengthened (ODS) AA6063 is experimentally recognized though carrying out compression tests at different temperatures (300–450 °C) and strain rates (0.01–1 s⁻¹). Microstructural studies conducted by TEM and EBSD indicate that dynamic softening mechanisms including dynamic recovery and dynamic recrystallization become operative in all the investigated materials depending on the regime of deformation. Moreover, the high temperature flow behavior is considerably influenced by the initial grain structure and the presence of reinforcement particles. The constitutive and artificial neural network (ANN) models were used to study the high-temperature flow behavior of the investigated alloys. To establish an accurate ANN model, material characteristics along with the processing parameters are deliberated. An Arrhenius type constitutive model with a strain-compensation term is employed to predict the flow stress of AA6063 alloys. The relative error associated with the constitutive and ANN models in the prediction of the flow stress is obtained 9.56% and 2.02%, respectively. The analysis indicates that the developed ANN model is more accurate in the prediction of flow stress with at least 78% less error in comparison to the constitutive model.

Graphic Abstract

中文翻译：

---

晶粒尺寸和氧化物弥散纳米颗粒在AA6063热变形行为中的作用：实验和人工神经网络建模研究

摘要

尽管在不同温度（300–450°C）和应变速率下进行了压缩试验，但通过实验认识到粗晶粒（CG），超细晶粒（UFG）和氧化物弥散强化（ODS）AA6063的热变形行为0.01–1 s ^-1）。TEM和EBSD进行的微结构研究表明，取决于变形方式，在所有研究的材料中，包括动态恢复和动态再结晶在内的动态软化机制都可以发挥作用。此外，高温流动行为很大程度上受初始晶粒结构和增强颗粒的存在的影响。本构和人工神经网络（ANN）模型用于研究所研究合金的高温流动行为。为了建立一个准确的人工神经网络模型，需要考虑材料特性以及加工参数。采用带有应变补偿项的Arrhenius型本构模型来预测AA6063合金的流变应力。本构模型和人工神经网络模型在预测流动应力时的相对误差分别为9.56％和2.02％。分析表明，与本构模型相比，已开发的ANN模型在流应力预测中更准确，且误差至少减少了78％。

图形摘要