Npj Comput. Mater.: 玻璃结构预测—懂物理的机器学习
-
阅读： 2023/3/16 16:11:34
氧化物玻璃可以通过改变成分来调整物理性能以满足各种应用的需求，但可能的虚拟样本庞大，远超出实验室制备和表征的数量级。物理模型，可以提供对玻璃结构的洞察并准确推断简单体系，但有可能过度简化复杂体系中的问题。机器学习（ML）模型，可以准确预测用于模型训练的成分空间内的玻璃结构，但难以准确预测范围之外的结构。要解决这一挑战，就需要一种组合模型，通过将玻璃成分如何影响短程原子结构的物理知识，嵌入机器学习，以提高模型对氧化物玻璃结构的预测和外推能力。
来自丹麦奥尔堡大学化学与生物科学系的Morten M. Smedskjaer等研究者介绍了一种结合统计力学和机器学习的模型，运用多层感知器神经网络学习统计力学的结构结果，作为输入数据的附加层，来学习热力学对结构形成的贡献，通过提供结构组件允许组成-结构-性能建模来提高组成-性能模型的预测和外推能力，准确预测内插和外推二元氧化物玻璃的结构，并准确外推三元氧化物玻璃的结构。该研究运用纯统计力学模型、多层感知器神经网络 (MLP-NN) 模型和组合模型预测Na2O-SiO2玻璃的结构。与单独依赖统计物理或机器学习的模型相比，这种组合模型对非线性成分-结构关系的预测精度提高。最后，三个模型不经过训练，直接预测Na2O-P2O5-SiO2玻璃体系的结构。作者发现，组合模型能够外推在训练过程中完全隐藏在模型之外的玻璃体系的结构。该研究结合了统计力学的外推能力与机器学习的准确性的两个优点，将物理知识嵌入机器学习，提高组合模型对氧化物玻璃结构的预测和外推能力，也为解决其他物理问题提供了研究策略。
该文近期发表于npj Computational Materials 8：192(2022)，英文标题与摘要如下，点击左下角“阅读原文”可以自由获取论文PDF。
Predicting glass structure by physics-informed machine learning
Mikkel L. B?dker, Mathieu Bauchy, Tao Du, John C. Mauro & Morten M. Smedskjaer
Machine learning (ML) is emerging as a powerful tool to predict the properties of materials, including glasses. Informing ML models with knowledge of how glass composition affects short-range atomic structure has the potential to enhance the ability of composition-property models to extrapolate accurately outside of their training sets. Here, we introduce an approach wherein statistical mechanics informs a ML model that can predict the non-linear composition-structure relations in oxide glasses. This combined model offers an improved prediction compared to models relying solely on statistical physics or machine learning individually. Specifically, we show that the combined model accurately both interpolates and extrapolates the structure of Na2O–SiO2 glasses. Importantly, the model is able to extrapolate predictions outside its training set, which is evidenced by the fact that it is able to predict the structure of a glass series that was kept fully hidden from the model during its training.
转自：“知社学术圈”微信公众号
如有侵权，请联系本站删除！