Number of the records: 1
Model order reduction technique for large scale flow computations
- 1.0499769 - ÚT 2019 RIV CZ eng C - Conference Paper (international conference)
Isoz, Martin
Model order reduction technique for large scale flow computations.
Topical problem of Fluid Mechanics 2018. Praha: Ústav termomechaniky AV ČR, v. v. i., 2018 - (Šimurda, D.; Bodnár, T.), s. 153-160. ISBN 978-80-87012-65-9. ISSN 2336-5781.
[Conference TOPICAL PROBLEMS OF FLUID MECHANICS 2018. Praha (CZ), 21.02.2018-23.02.2018]
R&D Projects: GA MŠMT(CZ) EF15_003/0000493
Institutional support: RVO:61388998
Keywords : proper orthogonal decomposition * discrete empirical interpolation method * computational fluid dynamics * OpenFOAM
OECD category: Applied mathematics
http://www.it.cas.cz/fm2015/im/admin/showfile/data/my/Papers/2018/20-TPFM2018.pdf
Current progress in numerical methods and available computational power combined with industrial needs promote the development of more and more complex models. However, such models are, due to their complexity, expensive from the point of view of the data storage and the time necessary for their evaluation. The model order reduction (MOR) seeks to reduce the computational complexity of large scale models. We present an application of MOR to the problems originating in the finite volume (FV) discretization of incompressible Navier-Stokes equations. Our approach to MOR is based on the proper orthogonal decomposition (POD)
with Galerkin projection. Moreover, the problems arising from the nonlinearities present in the original model are adressed within the framework of the discrete empirical interpolation method (DEIM). We provide a link between the POD-DEIM based MOR and OpenFOAM, which is an open-source CFD toolbox capable of solving even industrial scale problems. The availability of a link between OpenFOAM and POD-DEIM based MOR enables a direct order reduction for large scale systems originating in the industrial practice.
Permanent Link: http://hdl.handle.net/11104/0292285
Number of the records: 1