JAMSTEC > 地球情報基盤センター > 研究者のページ > セミナー情報



  第12回 CEISTセミナー

場所:シミュレータ研究棟 1 階 会議室
講師:國嶋 雄一(大阪大学)


An improvement of finite-difference lattice Boltzmann method for turbulent flow

Aerodynamic sound is sound caused by fluid motion. There are many types of aerodynamic sound, cavity tone, edge tone, and air column resonance, etc. Aerodynamic noise is one of important factors in designing cars, planes, windmills, and so on. Computational fluid dynamics (CFD) becomes a powerful approach to analyse them by effective modeling and development of computing power. Direct simulation considers compressibility of fluid, where flow and sound are solved simultaneously. It can handle various types of aerodynamic sound with flow-acoustic interactions.
The finite-difference lattice Boltzmann method (FDLBM) is a powerful scheme in direct simulation of aerodynamic sound. It is an extension of lattice Boltzmann method (LBM) with finite-difference discretization. It inherits some merits of the LBM, explicit algorithm and high scalability in massive parallel computing. The finite-difference discretization gives us to use body-fitted grids with high aspect ratio which can capture wall turbulence efficiently.
Turbulent flow around complex structures is often seen in many types of industrial devices.
We need to simulate flow and aerodynamic sound simultaneously under such a condition. It is important to assure reliability of a numerical scheme not only for aerodynamic sound but also for turbulence.
The first topic of this presentation is improvements of the FDLBM for the analysis of turbulence. The FDLBM had already achieved high resolution for sound wave, but resolution for turbulence has been inferior. We pay attention to the governing equation of FDLBM, discrete Bhatnagar-Gross-Krook equation. The equation is consisted with linear advection and collision of particles. Upwind schemes are used for the advection term. These are stable but include numerical dissipation, which diminishes sound wave and turbulence. We introduce higher order numerical dissipation which diminishes higher wave number fluctuations selectively.
We also reduce the amount of the numerical dissipation. We propose a local coefficient of the numerical dissipation which is decided from a symptom of the instabilities. The coefficient is set by a total validation diminishing scheme for higher derivative of each distribution function.
The second topic is implementation of the improved FDLBM for fully developed channel flow. The fully developed channel flow is a benchmark test of wall turbulence. Simulations are conducted in various resolutions in streamwise and spanwise directions. Those improved FDLBM are comparable with conventional CFD methods for incompressible flow. These methods are eminent for analyses of aerodynamic sound with turbulence.

大西 領