Learning mesh-based numerical analysis using graph neural networks

DOI
  • HORIE Masanobu
    Research Institute for Computational Science Co. Ltd. Graduate School of Systems and Information Engineering, University of Tsukuba
  • MORITA Naoki
    Research Institute for Computational Science Co. Ltd. Faculty of Engineering, Information and Systems, University of Tsukuba
  • IHARA Yu
    Research Institute for Computational Science Co. Ltd.
  • MITSUME Naoto
    Faculty of Engineering, Information and Systems, University of Tsukuba

Bibliographic Information

Other Title
  • グラフニューラルネットワークを用いたメッシュベース数値解析の汎用的な学習

Description

<p>Mesh-structured data is an important data structure to perform numerical analyses such as the finite element method and the finite volume method. It is known that graph neural networks (GNNs) can deal with mesh-structured data since meshes can be regarded as graphs. In this work, we demonstrate GNNs are useful in learning finite element analysis results. The proposed method efficiently leverages spatial information; that is, the input feature does not change under any rotation and translation. We show that our model generalizes to much larger meshes than these in the training dataset. Moreover, our model can perform inference for meshes, which have up to one million nodes.</p>

Journal

Keywords

Details 詳細情報について

  • CRID
    1391412326421287424
  • NII Article ID
    130007938956
  • DOI
    10.11421/jsces.2020.20201005
  • ISSN
    13478826
    13449443
  • Text Lang
    ja
  • Data Source
    • JaLC
    • CiNii Articles
  • Abstract License Flag
    Disallowed

Report a problem

Back to top