07.02.2013 Views

Scientific and Technical Aerospace Reports Volume 39 April 6, 2001

Scientific and Technical Aerospace Reports Volume 39 April 6, 2001

Scientific and Technical Aerospace Reports Volume 39 April 6, 2001

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

cally relevant flows. In particular, since the presence of complex boundaries is mimicked by body forces, the simulations are, in<br />

fact, carried out on ’simple grids’, thus taking advantage of the efficiency <strong>and</strong> accuracy of optimized solution procedures. IB/DNS<br />

simulations of complex-geometry turbulent-flows can be carried out with 1-2 million gridpoints on a PC with CPU times ranging<br />

between a few hours <strong>and</strong> a few days. In this paper we will show some examples of these simulations for an impeller-stirred cylindrical<br />

tank, <strong>and</strong> the results will be compared with experiments <strong>and</strong> RANS simulations.<br />

Author<br />

Turbulent Flow; Direct Numerical Simulation; Large Eddy Simulation; Reynolds Averaging; Boundary Conditions; Navier-<br />

Stokes Equation; Rotor Blades (Turbomachinery)<br />

<strong>2001</strong>0022654 Stanford Univ., Center for Turbulence Research, Stanford, CA USA<br />

Unsteady 3D RANS Simulations Using the v(exp 2)-f Model<br />

Iaccarino, Gianluca, Stanford Univ., USA; Durbin, Paul, Stanford Univ., USA; Annual Research Briefs - 2000: Center for Turbulence<br />

Research; December 2000, pp. 263-269; In English; See also <strong>2001</strong>0022631; No Copyright; Avail: CASI; A02, Hardcopy;<br />

A03, Microfiche<br />

Recent increased computational power has led to interest in the simulation of time-dependent flows for problems ranging<br />

from noise prediction to fluid/structure interactions. The computational cost <strong>and</strong> the resolution requirements are mainly related<br />

to the inviscid flow structures induced by geometry <strong>and</strong> wall layers; nevertheless, turbulence plays a crucial role in establishing<br />

such flow structures. Several approaches can be used to numerically simulate the behavior of unsteady flows: in the simplest, the<br />

Navier-Stokes equations are ensemble-averaged, converting turbulent fluctuations into Reynolds stresses (Reynolds-Averaged<br />

Navier-Stokes equations, RANS), while leaving the large scale, rotational motions to be resolved as unsteady phenomena. The<br />

large eddy simulation (LES) approach, on the other h<strong>and</strong>, employs a spatial averaging over a scale sufficient to remove scales not<br />

resolved by the particular grid being used. The subgrid scale turbulence is then modeled. A practical difference is in the degree<br />

of mesh resolution required: LES resolves the larger eddies of the turbulence itself, whereas the unsteady RANS approach models<br />

the turbulence <strong>and</strong> resolves only unsteady, mean flow structures - primarily larger than the turbulent eddies. Consequently, LES<br />

typically requires much higher grid resolution, at least locally, <strong>and</strong> is therefore more costly. On the other h<strong>and</strong>, LES only models<br />

the subgrid turbulence structures <strong>and</strong> is universally applicable to the extent that a universal grid-independent subgrid model can<br />

be established. In addition, LES resolves the complete range of scales of r<strong>and</strong>om motion, up to the cut-off frequency, while<br />

unsteady RANS aims to capture a single frequency (e.g. corresponding to coherent shedding) <strong>and</strong> to model the r<strong>and</strong>om motions<br />

using st<strong>and</strong>ard turbulence closures. Therefore, LES requires very long integration time to build a statistically-averaged solution;<br />

on the other h<strong>and</strong>, a few shedding periods are usually enough to obtain accurate phase-averaged solution with RANS, thus limiting<br />

its overall computational cost. The objective of this work is to apply the RANS approach with the v(exp 2) - f turbulence model<br />

to the solution of the flow around a surface mounted cube. A complete <strong>and</strong> reliable experimental database is available <strong>and</strong>, in addition,<br />

LES simulations have been carried out with considerable success. This problem was one of the test case presented at the<br />

’Workshop on Large Eddy Simulation of Flows Past Bluff Bodies’; several LES calculations were compared, showing a high<br />

degree of accuracy. RANS results were also included for comparison; they were obtained using simple turbulence models derived<br />

from k-epsilon <strong>and</strong> generally gave worse agreement with the measurements than did LES.<br />

Author<br />

Navier-Stokes Equation; Reynolds Averaging; Turbulent Flow; Turbulence Models; Applications Programs (Computers)<br />

<strong>2001</strong>0022655 Stanford Univ., Center for Turbulence Research, Stanford, CA USA<br />

Prediction of the Turbulent Flow in a Diffuser with Commercial CFD Codes<br />

Iaccarino, Gianluca, Stanford Univ., USA; Annual Research Briefs - 2000: Center for Turbulence Research; December 2000, pp.<br />

271-277; In English; See also <strong>2001</strong>0022631; No Copyright; Avail: CASI; A02, Hardcopy; A03, Microfiche<br />

There have been a few attempts in the literature to compare the performance of commercial computational fluid dynamics<br />

(CFD) codes, for instance, laminar arid turbulent test cases have been proposed to several CFD code vendors by the Coordinating<br />

Group for Computational Fluid Dynamics of the Fluids Engineering Division of ASME. A series of five benchmark problems<br />

were calculated, with all the mesh generation <strong>and</strong> simulations performed by the vendors themselves; only two of the problems<br />

required turbulent simulations. The first of these benchmarks is the flow around a square cylinder: the flow is unsteady <strong>and</strong> all<br />

of the codes predicted the measured Strouhal number reasonably well. However, poor accuracy was obtained in the details of the<br />

wake flow field. It was also noted that, depending on the code used (<strong>and</strong> assuming grid-converged results), the same k-epsilon<br />

model predicted very different results (from 2% to 16% accuracy in the Strouhal number, for example). The reasons for this difference<br />

include different grids, non-demonstrated grid convergence, different implementation of the models, <strong>and</strong> different boundary<br />

conditions. It must be pointed out that the predictions for this problem are strongly affected by the treatment of the stagnation point<br />

region: as reported in Durbin, the k-epsilon models predict a spurious high level of turbulent kinetic energy near the stagnation<br />

62

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!