AMGNET: multi-scale graph neural networks for flow field prediction
Solving partial differential equations of complex physical systems is a computationally expensive task, especially in Computational Fluid Dynamics(CFD). This drives the application of deep learning methods in solving physical systems. There exist a few deep learning models that are very successful i...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2022-12-01
|
| Series: | Connection Science |
| Subjects: | |
| Online Access: | http://dx.doi.org/10.1080/09540091.2022.2131737 |
| _version_ | 1797684012140986368 |
|---|---|
| author | Zhishuang Yang Yidao Dong Xiaogang Deng Laiping Zhang |
| author_facet | Zhishuang Yang Yidao Dong Xiaogang Deng Laiping Zhang |
| author_sort | Zhishuang Yang |
| collection | DOAJ |
| description | Solving partial differential equations of complex physical systems is a computationally expensive task, especially in Computational Fluid Dynamics(CFD). This drives the application of deep learning methods in solving physical systems. There exist a few deep learning models that are very successful in predicting flow fields of complex physical models, yet most of these still exhibit large errors compared to simulation. Here we introduce AMGNET, a multi-scale graph neural network model based on Encoder-Process-Decoder structure for flow field prediction. Our model employs message passing of graph neural networks at different mesh graph scales. Our method has significantly lower prediction errors than the GCN baseline on several complex fluid prediction tasks, such as airfoil flow and cylinder flow. Our results show that multi-scale representation learning at the graph level is more effective in improving the prediction accuracy of flow field. |
| first_indexed | 2024-03-12T00:24:24Z |
| format | Article |
| id | doaj.art-8c8d90f277314cfb87062bb1da1aff48 |
| institution | Directory Open Access Journal |
| issn | 0954-0091 1360-0494 |
| language | English |
| last_indexed | 2024-03-12T00:24:24Z |
| publishDate | 2022-12-01 |
| publisher | Taylor & Francis Group |
| record_format | Article |
| series | Connection Science |
| spelling | doaj.art-8c8d90f277314cfb87062bb1da1aff482023-09-15T10:48:01ZengTaylor & Francis GroupConnection Science0954-00911360-04942022-12-013412500251910.1080/09540091.2022.21317372131737AMGNET: multi-scale graph neural networks for flow field predictionZhishuang Yang0Yidao Dong1Xiaogang Deng2Laiping Zhang3Sichuan UniversityNational University of Defense TechnologyAcademy of Military SciencesNational Innovation Institute of Defense TechnologySolving partial differential equations of complex physical systems is a computationally expensive task, especially in Computational Fluid Dynamics(CFD). This drives the application of deep learning methods in solving physical systems. There exist a few deep learning models that are very successful in predicting flow fields of complex physical models, yet most of these still exhibit large errors compared to simulation. Here we introduce AMGNET, a multi-scale graph neural network model based on Encoder-Process-Decoder structure for flow field prediction. Our model employs message passing of graph neural networks at different mesh graph scales. Our method has significantly lower prediction errors than the GCN baseline on several complex fluid prediction tasks, such as airfoil flow and cylinder flow. Our results show that multi-scale representation learning at the graph level is more effective in improving the prediction accuracy of flow field.http://dx.doi.org/10.1080/09540091.2022.2131737flow field predictionalgebraic multigridmulti-scalegnn |
| spellingShingle | Zhishuang Yang Yidao Dong Xiaogang Deng Laiping Zhang AMGNET: multi-scale graph neural networks for flow field prediction Connection Science flow field prediction algebraic multigrid multi-scale gnn |
| title | AMGNET: multi-scale graph neural networks for flow field prediction |
| title_full | AMGNET: multi-scale graph neural networks for flow field prediction |
| title_fullStr | AMGNET: multi-scale graph neural networks for flow field prediction |
| title_full_unstemmed | AMGNET: multi-scale graph neural networks for flow field prediction |
| title_short | AMGNET: multi-scale graph neural networks for flow field prediction |
| title_sort | amgnet multi scale graph neural networks for flow field prediction |
| topic | flow field prediction algebraic multigrid multi-scale gnn |
| url | http://dx.doi.org/10.1080/09540091.2022.2131737 |
| work_keys_str_mv | AT zhishuangyang amgnetmultiscalegraphneuralnetworksforflowfieldprediction AT yidaodong amgnetmultiscalegraphneuralnetworksforflowfieldprediction AT xiaogangdeng amgnetmultiscalegraphneuralnetworksforflowfieldprediction AT laipingzhang amgnetmultiscalegraphneuralnetworksforflowfieldprediction |