Premium
Equation‐free model reduction for complex dynamical systems
Author(s) -
Le Maître O. P.,
Mathelin L.
Publication year - 2009
Publication title -
international journal for numerical methods in fluids
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.938
H-Index - 112
eISSN - 1097-0363
pISSN - 0271-2091
DOI - 10.1002/fld.2219
Subject(s) - rendering (computer graphics) , projection (relational algebra) , galerkin method , mathematics , basis (linear algebra) , transformation (genetics) , algorithm , reduction (mathematics) , mathematical optimization , computer science , geometry , finite element method , artificial intelligence , biochemistry , chemistry , physics , gene , thermodynamics
This paper presents a reduced model strategy for simulation of complex physical systems. A classical reduced basis is first constructed relying on proper orthogonal decomposition of the system. Then, unlike the alternative approaches, such as Galerkin projection schemes for instance, an equation‐free reduced model is constructed. It consists in the determination of an explicit transformation, or mapping, for the evolution over a coarse time‐step of the projection coefficients of the system state on the reduced basis. The mapping is expressed as an explicit polynomial transformation of the projection coefficients and is computed once and for all in a pre‐processing stage using the detailed model equation of the system. The reduced system can then be advanced in time by successive applications of the mapping. The CPU cost of the method lies essentially in the mapping approximation which is performed offline, in a parallel fashion, and only once. Subsequent application of the mapping to perform a time‐integration is carried out at a low cost thanks to its explicit character. Application of the method is considered for the 2‐D flow around a circular cylinder. We investigate the effectiveness of the reduced model in rendering the dynamics for both asymptotic state and transient stages. It is shown that the method leads to a stable and accurate time‐integration for only a fraction of the cost of a detailed simulation, provided that the mapping is properly approximated and the reduced basis remains relevant for the dynamics investigated. Copyright © 2009 John Wiley & Sons, Ltd.