M4 Meeting in Manchester


Attendees: Sylvain Lardeau, Alistair Revell, Alex Skillen, Flavien Billard, Paul Dawson

SL has presented update on the elliptic-blending k-eps model with all y+ treatment.

exact implementation of the model: Compared to the original published version, SL has made the following changes:

  • The cross diffusion term of the phi equation has been dropped. FB is asking to check in the initial SL implementation whether it is 1/k(nut/sigma_k) grad k * grad phi or 1/k(nu+nut/sigma_k) grad k * grad phi which is used. The former is correct whereas the latter implies the cross diffusion term (explicit) to be involved in the near-wall region which is problematic.
  • SL has modified the Durbin realizability time scale limiter. He is using it only in the production term of k appearing in the source term of k and phi (= two occurences, respectively in P_k and f_h) and nowhere else. In particular the limiter should not be used in the model for turbulent transport (in particular those for the turbulence variables, which means that the limiter could be retained in the momentum turbulent transport (Re stress) model for energy conservation reasons). He recommends to use a=1 instead of a=0.6 in most cases (e.g F1 applications) whereas a=0.6 should be used in the impinging jet test case.
  • Ceps2 variable is now an option
  • SL calculates the E term as $\displaystyle \sum_{i=1}^3 \displaystyle \sum_{j=1}^3 \displaystyle \sum_{k=1}^3 (\partial_k \partial_j U_i)^2$ and FB calculates it as $\displaystyle \sum_{i=1}^3 \left( \displaystyle \sum_{j=1}^3 \displaystyle \sum_{k=1}^3 (\partial_k \partial_j U_i)\right)^2$. SL has raised issues related to the E term calculation (+ incresaed numerical cost of calculation of 27 double derivative). SL results on 2-D diffuser, periodic hill and hump exhibit spurious wiggles on the cf profiles, presumably due to the E term calculation.
  • In the AdWF implementation SL is not using the recommended epsilon (lookup table) wall function but a near-wall / log blended formula a la Hanjalic Popovac, for stability reasons. This originates from the fact that In the STARCCM+ implementation the first cell turbulence dissipation rate $\varepsilon$ comes from the wall function value and not from the computed value of epsilon at the previous time step, as recommended in the BLV2K -AdWF documentation. FB suggests that following the latter recommendation should allow the use of the lookup table epsilon WF instead of the blended formula.

Problems to be solved:

  • SL has presented buoyancy driven flows results (Betts, Kasagi, Rayleigh Benard): SL results are different from those presented in FB thesis and they are much more laminar (in particular in the near-wall BL along the vertical wall of the Betts cavity and the aiding side of the Kasagi vertical channel test flow). The reason for this (needs to be checked) maybe the way buoyancy production term is accounted for in the epsilon equation (SL uses $C_{\varepsilon 1}*(P+G)/T$ whereas FB uses $C_{\varepsilon1}*(P+max(G,0))/T$ ). Another reason may be the choice for turbulent Prandtl number (FB uses 1 and SL uses 0.9). FB and SL to iterate on this until same results obtained.
  • SL presented BLV2K results with rotation curvature correction (Durbin's or A Hellsten - 29th AIAA Fluid Dynamics Conference,(Albuquerque, …, 1998 - aero.hut.fi) applied on rotating channel flow and NACA 3-D airfoil. In the rotating channel flow the corrected BLV2K appears to be over-sensitive (near-wall turbulent damping) to acceleration. (improvement of LRE effects modelling needed?). To assess this FB will run the case with Hellsten correction.
  • SL and FB obtained the same behaviour on the impinging jet case (Nusselt overshoot near the impingement point). considering including the Yap term (as suggested in Craft, Graham, Launder "Impinging jet studies for turbulence model assessment-II. An examination of the performance of four turbulence models ).

Actions:

  • FB will revisit his bluff body calculations (cylinder, NACA) with SL modification of the stagnation point anomaly treatment.
  • FB send his Sholten and Murray mesh to SL for testing possible nut undershoot at the wake/free stream region (as observed by FB in this report).
  • SL and FB check their buoyancy driven flows calculations
  • FB tries Hellsten RC correction on rotating channel flow (+cyclone case)
  • FB continues development of the sleep mode / transition modelling for BLV2K . Application to T3A case.
  • FB continues all y+ development for BLV2K and EBRSM.


Current Tags:
create new tag
, view all tags
Topic attachments
I Attachment Action Size Date Who Comment
pdfpdf AIAA_98-2554-CP.PS.pdf manage 913.5 K 2012-12-19 - 00:13 FlavienBillard  
pdfpdf report.pdf manage 1217.2 K 2012-12-19 - 11:36 FlavienBillard  
Topic revision: r5 - 2014-10-27 - 11:18:42 - AlastairWest
Main Web
12 Dec 2017

Site

Manchester CfdTm
Code_Saturne

Ongoing Projects

ATAAC
KNOO

Previous Projects

DESider
FLOMANIA

Useful Links:

User Directory
Photo Wall
Upcoming Events
Add Event
 

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.