Direct Numerical Simulations Simulations Research Articles

G201 フリーメッシュ法による圧縮性を考慮したエッジトーンの数値解析

Numerical simulation of edge tone based on compressible Navier-Stokes equations is attempted using a node-based finite element method. Recent advanced computers have made possible calculation of both generation and propagation of aerodynamic sound based on CFD. i.e. direct numerical simulation (DNS). However, most part of DNS simulations handles aerodynamic sound generated from 2-dimensional objects with simple boundary shapes because of the limit of computer power and structured grids. Furthermore, conventional density-based numerical schemes for compressible flows encounter so-called stiffness problem under low Mach number. In the circumstances. we apply the C-CUP method and the CIVA method to the free mesh method (FMM), in order to achieve effective parallel computing for sound generated from 3-dimensinal objects. The C-CUP method enables calculation of compressible flows under low Mach number, and the CIVA method improves accuracy of interpolation within elements. The FMM is employed for an effective large-scale finite element analysis on parallel computers.

Effect of the subgrid scales on particle motion

In this study, the effects of small-scale velocity fluctuations on the motion of tracer particles is investigated by releasing particles in a turbulent channel flow at Reτ=175, and following their motion in time. Two types of numerical experiments were carried out: first, the Eulerian velocity field was evaluated by the direct numerical simulation (DNS) and the particles were advanced in time using the resolved and several filtered velocity fields to study the effect of the subgrid-scale velocity fluctuations on particle motion without the influence of modeling errors. In the second stage, the particle-motion study was performed using independent DNS and large-eddy simulations (LES), thus including the effect of interpolation and subgrid-scale stress modeling errors on the dispersion statistics. At this Reynolds number the small scales were found to have a limited effect on the statistics examined (one-particle dispersion, one-particle velocity autocorrelation, Lagrangian integral time scale, turbulent diffusivity, and two-particles rms dispersion). Only when a significant percentage of the energy was removed from the velocity field some differences were observed between filtered and unfiltered data (especially near the wall). It was found that when the dynamic eddy-viscosity model was employed, modeling errors did not affect the results as much as the filtering itself; the use of the Smagorinsky model, on the other hand, gave inaccurate results. Additional computations for finite-inertia particles have shown that these results represent a conservative estimate, in the sense that actual particles with inertia are less sensitive than the tracer particles examined in the first part of the investigation, and that LES provides improved results when particles with inertia are used.