GINNs: Graph-Informed Neural Networks for Multiscale Physics
GINNs: Graph-Informed Neural Networks for Multiscale Physics
复制标题
DOI：
10.1016/j.jcp.2021.110192
复制
发表时间：
2020-06
期刊：
影响因子：
--
通讯作者：
E. Hall;S. Taverniers;M. Katsoulakis;D. Tartakovsky
中科院分区：
文献类型：
--
作者：
E. Hall;S. Taverniers;M. Katsoulakis;D. Tartakovsky
我们引入了图形信息神经网络（GINN）的概念，这是一种将深度学习与概率图形模型（PGMs）相结合的混合方法，可作为多尺度和多物理场系统的基于物理的表示的替代。GINN解决了消除基于物理的模型中固有的计算瓶颈和生成用于以高置信度估计感兴趣量（QoI）的概率分布的大型数据集的双重挑战。NN学习的复杂物理学的选择及其监督学习/预测都由PGM提供信息，其中包括制定可调控制变量（CV）的结构化先验知识，以考虑它们的相互关联性并确保物理上可靠的CV和QoI分布。GINN加速了对基于模拟的决策至关重要的QoI的预测，其中仅使用基于物理的模型生成足够的样本数据通常非常昂贵。使用基于能源存储的应用程序，我们描述了用于领域感知贝叶斯网络的GINN的构建，该网络嵌入了超级电容器动态的均质模型和用于朗缪尔吸附模型的数据驱动贝叶斯网络。这两个例子都证明了GINN能够产生相关非高斯的核密度估计，具有严格的置信区间的偏斜的生活质量。
We introduce the concept of a Graph-Informed Neural Network (GINN), a hybrid approach combining deep learning with probabilistic graphical models (PGMs) that acts as a surrogate for physics-based representations of multiscale and multiphysics systems. GINNs address the twin challenges of removing intrinsic computational bottlenecks in physics-based models and generating large data sets for estimating probability distributions of quantities of interest (QoIs) with a high degree of confidence. Both the selection of the complex physics learned by the NN and its supervised learning/prediction are informed by a PGM, which includes the formulation of structured priors for tunable control variables (CVs) to account for their mutual correlations and ensure physically sound CV and QoI distributions. GINNs accelerate the prediction of QoIs essential for simulation-based decision-making where generating sufficient sample data using physics-based models alone is often prohibitively expensive. Using applications grounded in energy storage, we describe the construction of GINNs for a domain-aware Bayesian network that embeds a homogenized model of supercapacitor dynamics and a data-driven Bayesian network for a Langmuir adsorption model. Both examples demonstrate the ability of GINNs to produce kernel density estimates of relevant non-Gaussian, skewed QoIs with tight confidence intervals.