2.2 Non-Reacting Flows
2.2.1 DNS of two phase flow modelling (A. Massol,
F. Ducros)
The modelling activity in the field of two phase flow has started through
through the collaboration with EDF and IMFT (O. Simonin). The issue is to
measure the transfer coefficients (momentum and passive scalar) between the
the flow and an array of particles. The analysis is conducted through DNS
performed with AVBP (Fig. 2.4). The underlying motivation is to
improve the heuristic laws generally used for such transfers for both the
Lagragian/Eulerian and the Eulerian/Eulerian modellings. The first step has
been to estimate the validity of existing drag laws for fluidized beds, in
order to propose new and improved correlations.
Figure 2.4: Gas trajectories behind a particle. Trajectories are colored by the
U-component of velocity, and particles are colored by the pressure
coefficient. ad = 0.15, Re=300.
2.2.2 Assessment of U-RANS and related V-LES techniques:
(B. Caruelle, F. Ducros)
The main results here show a comparison between LES, U-RANS using the
Spalart-Allmaras model and the Detached Eddy Simulation (DES) approaches
on various test cases from attached boundary layer to full transonic 2,5
airfoil profile. The DES is shown to behave better than other investigated
modellings for buffeting (under restrictions that have been determined and
discussed
[Caruelle, 2000]).
As already said in the introduction, the PhD thesis of B. Caruelle received
the first BMW scientific award contest.
2.2.3 LES modelling fundamentals: (C. Jiménez, B. Cuenot, F. Ducros, F. Nicoud)
Theoretical work on LES modelling has continued in order to propose and/or
improve new formalisms with high potential.
Figure 2.5: LES of mixing in a gas turbine.
-
a former work related to the filtered standard models based on
resolved turbulents scales has come to an end with the assessment of a
clear formalism
([Schlüter, 2000 PhD]).
This kind of model is routinely used in AVBP (on the filtered Smagorinsky
form) for many contracts (see Fig. 2.5).
- a generalized relaxation procedure has been set up in order to
stabilize and control the variation of the constant coming along with the
standard dynamic procedure. This produces a new family of dynamic models,
strictly local in space and stable enough to be used without clipping or
averaging procedures
([Ducros, 2000a]).
These two works have been done in collaboration with ONERA Chatillon.
- The previous model for the subgrid scale (SGS) variance has
been checked against a-priori tests
([Jiménez, 2000]).
Moreover it has been extended to take into account the numerical
dissipation brought up by the scheme as a source term in the SGS
variance transport equation
([Ducros, 2000]).
- The DES technique proposed by Spalart (Boeing) some times ago has
been assessed as a mean of treating the wall problem in LES. This fundamental
collaborative work
[Nikitin, 2000]
underlined some of the resolution requirements of this technique and helped
to understand some results of Caruelle's thesis (see above).
- The fundamental work started in collaboration with the Center for
Turbulence Research (Stanford University) about the design of optimal
law-of-the-wall for LES
[Nicoud, 2000]
has been continued during the 2000 Summer program
[Baggett, 2000]
leading to a better understanding of the limitations of law-of-the-walls
for LES's.
2.2.4 Heat transfer calculation in RANS formulation
(F. Dabireau, F. Nicoud)
The task of deriving law-of-the-wall is much easier for steady RANS than for
LES. However, almost all the previous studies available in the literature
assume that density is constant in the process of deriving the relationship
between the velocity field and the wall shear stress. Two formulations
accounting for the thermophysical variations
[Nicoud, 2000;
Dabireau, 2000a]
have been tested against DNS data for subsonic
[Nicoud, 2000repa;
Nicoud, 2000a]
and supersonic flows. These law-of-the-wall formulations have been
|
|
