Mg-Gd-Zn based alloys have better creep resistance than other Mg alloys and attract more attention at elevated temperatures. However, the multiple alloying elements and various heat treatment conditions, combined with complex microstructural evolution during creep tests, bring great challenges in understanding and predicting creep behaviors. In this study, we proposed to predict the creep properties and reveal the creep mechanisms of Mg-Gd-Zn based alloys by machine learning. On the one hand, the minimum creep rates were effectively predicted by using a support vector regression model. The complex and nonmonotonic effects of test temperature, test stress, alloying elements, and heat treatment conditions on the creep properties were revealed. On the other hand, the creep stress exponents and creep activation energies were calculated by machine learning to analyze the variation of creep mechanisms, based on which the constitutive equations of Mg-Gd-Zn based alloys were obtained. This study introduces an efficient method to comprehend creep behaviors through machine learning, offering valuable insights for the future design and selection of Mg alloys.

中文翻译：

---

通过机器学习了解 Mg-Gd-Zn 合金的蠕变行为和机制


Mg-Gd-Zn基合金比其他镁合金具有更好的抗蠕变性能，在高温下受到更多关注。然而，多种合金元素和各种热处理条件，再加上蠕变试验过程中复杂的微观结构演变，给理解和预测蠕变行为带来了巨大的挑战。在这项研究中，我们提出通过机器学习来预测蠕变特性并揭示 Mg-Gd-Zn 基合金的蠕变机制。一方面，利用支持向量回归模型有效地预测了最小蠕变速率。揭示了试验温度、试验应力、合金元素和热处理条件对蠕变性能的复杂且非单调的影响。另一方面，通过机器学习计算蠕变应力指数和蠕变活化能，分析蠕变机制的变化，并在此基础上得到Mg-Gd-Zn基合金的本构方程。这项研究引入了一种通过机器学习理解蠕变行为的有效方法，为镁合金的未来设计和选择提供了宝贵的见解。