Physics Knowledge Discovery via Neural Differential Equation Embedding
Physics Knowledge Discovery via Neural Differential Equation Embedding
复制标题
DOI：
10.1007/978-3-030-86517-7_8
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Yexiang Xue;M. Nasim;Maosen Zhang;C. Fan;Xinghang Zhang;A. El-Azab
中科院分区：
文献类型：
--
作者：
Yexiang Xue;M. Nasim;Maosen Zhang;C. Fan;Xinghang Zhang;A. El-Azab
从实验数据中发现物理学知识的过程，虽然引起了人们极大的兴趣，但在很大程度上仍然是一个人工试错的过程。本文提出了神经微分方程嵌入（NeuraDiff），这是一种端到端的方法，可以直接从实验数据中学习由一组偏微分方程表征的物理模型。其核心思想是两个神经网络的集成-一个识别网络从实验数据中提取物理模型变量的值，另一个神经微分方程网络模拟物理模型的时间演化。通过匹配两个神经网络的结果来完成学习。我们将NeuraDiff应用于跟踪和学习晶体材料在辐照和高温下纳米级缺陷的物理模型的现实应用。实验结果表明，NeuraDiffproduces高度准确的跟踪结果，同时捕捉纳米级缺陷的正确动态。
Despite much interest, physics knowledge discovery from experiment data remains largely a manual trial-and-error process. This paper proposes neural differential equation embedding (NeuraDiff), an end-to-end approach to learn a physics model characterized by a set of partial differential equations directly from experiment data. The key idea is the integration of two neural networks – one recognition net extracting the values of physics model variables from experimental data, and the other neural differential equation net simulating the temporal evolution of the physics model. Learning is completed by matching the outcomes of the two neural networks. We applyNeuraDiffto the real-world application of tracking and learning the physics model of nano-scale defects in crystalline materials under irradiation and high temperature. Experimental results demonstrate thatNeuraDiffproduces highly accurate tracking results while capturing the correct dynamics of nano-scale defects.