Results for case (not set)

Code: Code_Saturne

Version: 2.0-beta2

Authors: K. Osman

Method and Numerical Options

ID Turbulence model Options Code version Other
KE1 k-epsilon   1.3  
KE2 k-epsilon   2.0  
KE3 k-epsilon IDEUCH=0 1.3  
KE4 k-epsilon IDEUCH=0 2.0  
KE5 k-epsilon   2.0 CCRT (Platine) in parallel
KE6 k-epsilon flux reconstruction 2.0  
KEPL1 k-epsilon with linear production   1.3  
KEPL2 k-epsilon with linear production   2.0  
KEPL3 k-epsilon with linear production IDEUCH=0 1.3  
KEPL4 k-epsilon with linear production IDEUCH=0 2.0  
KEPL5 k-epsilon with linear production flux reconstruction 2.0  
SSG1 Rij-SSG   1.3  
SSG2 Rij-SSG   2.0  
SSG3 Rij-SSG IDEUCH=0 1.3  
SSG4 Rij-SSG IDEUCH=0 2.0  
SSG5 Rij-SSG ICLSYR=1 1.3  
SSG6 Rij-SSG ICLSYR=1 2.0  
SSG7 Rij-SSG ICLSYR=1, IDEUCH=0 1.3  
SSG8 Rij-SSG ICLSYR=1, IDEUCH=0 2.0  

Nota bene: If not specified the calculations are run on a Calibre workstation (Calibre 5 for the calculations run with version 2.0). Both RIJ-SSG and k-epsilon models use by default IDEUCH=1

Models

%$k-\varepsilon$% with and without linear production, %$R_{ij}-\varepsilon$% model

Mesh

The mesh used is the refined mesh used for the 1.3 validation.

Description of the results files

Boundary conditions

Reference Publications

Discussion

Comparison of the models

Figures (1), (2) and (3) compares the different models %$k-\varepsilon$% with and without linear production and %$R_{ij}-\varepsilon$% model with or without the partial implicitation of symmetry conditions. Figure (1) shows the Nusselt number predictions, with results using the versions 1.3 and 2.0. Figures (2) and (3) present the velocity and turbulent kinetic energy profile for the vesion 2.0. The new version of the code gives the same Nusselt number prediction as the previous version. The patch p03 regarding the algebraic multigrid for the pressure equation was used on some cases, with no influence on the results.

Activating the partial implicitation of symmetry condition enables to get results closer to the experimental measurements. This can be seen on the Nusselt plot figure (1) as well as the velocity and turbulence profiles, figures (2) and (3).

The linear production %$k-\varepsilon$% enables to reduce the problem of turbulence over-prediction observed with the original model towards the stagnation point.

Courbe1a.png
Figure1a: Nusselt profile, case KE1, KE2, KEPL1, KEPL2, SSG1, SSG2, SSG5, SSG6
Courbe1b.png
Figure1b: Nusselt profile, case KE1, KE2, KEPL1, KEPL2, SSG1, SSG2, SSG5, SSG6, close-up near axis,

Courbe4a.png
Figure2a: Velocity profile, case KE2, KEPL2, SSG2, SSG6, R/D=0.5
Courbe4b.png
Figure2b: Velocity profile, case KE2, KEPL2, SSG2, SSG6, R/D=1.0
Courbe4c.png
Figure2c: Velocity profile, case KE2, KEPL2, SSG2, SSG6, R/D=2.5
Courbe4d.png
Figure2d: Velocity profile, case KE2, KEPL2, SSG2, SSG6, R/D=3.0

Courbe5a.png
Figure3a: Turbulence profile, case KE2, KEPL2, SSG2, SSG6, R/D=0.5
Courbe5b.png
Figure3b: Turbulence profile, case KE2, KEPL2, SSG2, SSG6, R/D=1.0
Courbe5c.png
Figure3c: Turbulence profile, case KE2, KEPL2, SSG2, SSG6, R/D=2.5
Courbe5d.png
Figure3d: Turbulence profile, case KE2, KEPL2, SSG2, SSG6, R/D=3.0

Inter machine comparison

Same results are obtained using the %$k-\varepsilon$% model whether it is used on the Calibre workstation or the Platine CCRT machine in parallel.

Courbe2a.png
Figure4a: Nusselt profile, case KE1, KE2, KE5
Courbe2b.png
Figure4b: Nusselt profile, case KE1, KE2, KE5, close-up near axis

Influence of the flux reconstruction

The use of the flux reconstruction in the equations of %$k$% and %$\varepsilon$% for the models %$k-\varepsilon$% with and without linear production has little influence. This was to be expected since the mesh is quasi orthogonal (about 100 000 boundary faces with a non-orthogonality angle below 7, and less than 1000 above).

Courbe3a.png
Figure5a: Nusselt profile, case KE2, KE6, KEPL2, KEPL5
Courbe3b.png
Figure5b: Nusselt profile, case KE2, KE6, KEPL2, KEPL5, close-up near axis

Influence of using IDEUCH=0

Using a one velocity wall function ( IDEUCH=0) yields a behaviour different from the the experimentally observed one, towards the centre of the plate. Activating IDEUCH=1 (two-velocity wall function) leads to a better matching with the measures. With a one velocity wall function the number of cells falling into the viscous sublayer is much more important, which may explain the observed difference of results.

Courbe9a.png
Figure6a: Nusselt profile, case KE3, KE4, KEPL3, KEPL4, SSG3, SSG4, SSG7, SSG8
Courbe9b.png
Figure6b: Nusselt profile, case KE3, KE4, KEPL3, KEPL4, SSG3, SSG4, SSG7, SSG8, close-up near axis,

Courbe7a.png
Figure7a: Velocity profile, case KE4, KEPL4, SSG4, SSG8, R/D=0.5
Courbe7b.png
Figure7b: Velocity profile, case KE4, KEPL4, SSG4, SSG8, R/D=1.0
Courbe7c.png
Figure7c: Velocity profile, case KE4, KEPL4, SSG4, SSG8, R/D=2.5
Courbe7d.png
Figure7d: Velocity profile, case KE4, KEPL4, SSG4, SSG8, R/D=3.0

Courbe8a.png
Figure8a: Turbulence profile, case KE4, KEPL4, SSG4, SSG8, R/D=0.5
Courbe8b.png
Figure8b: Turbulence profile, case KE4, KEPL4, SSG4, SSG8, R/D=1.0
Courbe8c.png
Figure8c: Turbulence profile, case KE4, KEPL4, SSG4, SSG8, R/D=2.5
Courbe8d.png
Figure8d: Turbulence profile, case KE4, KEPL4, SSG4, SSG8, R/D=3.0



Current Tags:
Saturne_Validation_2.01Add my vote for this tag create new tag
, view all tags
Topic revision: r28 - 2011-09-02 - 09:44:08 - DavidMonfort
 

Computational Fluid Dynamics and Turbulence Mechanics
@ the University of Manchester
Copyright & by the contributing authors. Unless noted otherwise, all material on this web site is the property of the contributing authors.