Abstract
Advanced nuclear reactors often involve complex thermal-hydraulic behavior during transients, making multi-dimensional modeling capability a desirable feature for system analysis code. For multi-dimensional modeling in system code, coarse mesh setup is adopted to ensure computational efficiency and consistency with the overall system-level simulation. In this sense, the closure model is a necessary component for multi-dimensional modeling in system code for most scenarios. In this paper, we investigate a deep learning-based data-driven approach for multi-dimensional closure model development. The proposed approach utilizes fine mesh computational fluid dynamics data to develop a coarse mesh closure relation that is compatible with system code. In the proposed approach, deep neural network serves as the multi-dimensional field-to-field surrogate model. We select a loss-of-flow transient that involves large volume and complex thermal stratification and mixing as a case study. Two different types of deep neural networks, i.e. densely connected convolutional neural network and long-short-term-memory network based on proper orthogonal decomposition, are used in the case study. Promising results are obtained in the case study which demonstrates the potential of applying the proposed data-driven approach into advanced system analysis code.
Original language | English |
---|---|
Title of host publication | Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 |
Publisher | American Nuclear Society |
Pages | 867-876 |
Number of pages | 10 |
ISBN (Electronic) | 9781713886310 |
DOIs | |
State | Published - 2021 |
Externally published | Yes |
Event | 2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 - Virtual, Online Duration: Oct 3 2021 → Oct 7 2021 |
Publication series
Name | Proceedings of the International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 |
---|
Conference
Conference | 2021 International Conference on Mathematics and Computational Methods Applied to Nuclear Science and Engineering, M and C 2021 |
---|---|
City | Virtual, Online |
Period | 10/3/21 → 10/7/21 |
Funding
This material is based upon work supported by Laboratory Directed Research and Development (LDRD) funding from Argonne National Laboratory, provided by the Director, Office of Science, of the U.S. Department of Energy under Contract No. DEAC02-06CH11357. The authors gratefully acknowledge the computing resources provided on Blues, a high-performance computing cluster operated by the Laboratory Computing Resource Center at Argonne National Laboratory. The authors gratefully acknowledge the resources of the Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC02-06CH11357. The authors gratefully acknowledge the resources of the Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under ontC ract DE-AC0 -1HC 13. This material is based upon work supported by Laboratory Directed Research and Development (LDRD) funding from Argonne aN tional Laboratory, provided by the Director, fO fice of Science, of the .S. U Department of Energy under ontC ract o. DN EA02C-1HC0 13.
Keywords
- coarse mesh turbulence prediction
- convolutional neural networks
- long-short-term memory networks
- proper orthogonal decomposition