Efficient treatment of complex geometries for large eddy simulations of turbulent flows |
| |
Authors: | DGE Grigoriadis JG Bartzis |
| |
Affiliation: | a NCSR DEMOKRITOS, Ag.Paraskevi Attikis 15310, Athens, Greece b LFMT, Department of Mechanical Engineering, Aristotle University of Thessaloniki, 54124, Greece |
| |
Abstract: | Incompressible turbulent flow over a backward facing step at Reh=5100 is investigated by large eddy simulations (LES). The ratio of the oncoming boundary layer thickness δ to the step height h was set to 1.2. Additionally channel flows at various Reτ numbers are presented for the validation of the numerical code. The results are compared with existing DNS and experimental databases. The present study focuses on different procedures for LES of engineering problems in complex geometries using structured rectangular grids. Two different methods that are able to treat complex geometrical configurations are implemented, examined and compared; namely the domain decomposition approach based on Schur’s complement and the immersed boundary method. In the present study both methods make use of a fast direct Poisson’s pressure solver based on a heavily modified version of the public domain package FISHPAK. The latter was optimised and fully parallelised for shared memory architectures, for solutions on rectangular grids stretched in one or two directions. The resulting code reaches performances of 1.0 μs/node/iter, allowing low cost computations on grids of the order of million points. The main objective of the present study was to investigate the potential of different methods for LES in complex geometrical configurations like bluff body flows and wakes. One of the main findings is that careful selection of numerical methods and implementation techniques can lead to accurate and very efficient codes, where the geometric complexity does not lead to algorithmic or numerical complexity. |
| |
Keywords: | LES Schur complement Domain decomposition Immersed boundary FISHPAK Parallel computing Turbulent channel flow BFS flow |
本文献已被 ScienceDirect 等数据库收录! |
|