ENH: improve access to the inner content of turbulence models
WIP for log.linux64GccDPInt32Opt.zip (Gcc 7.4)
log.linux64ClangDPInt32Opt.zip (LLVM 9.0)
testLoopReport.PASS.zipClang compilation test.
Summary
The commits provide access to:
-
omegafield estimations for non-omega turbulence models, - model coefficients of turbulence models for
fvOptions.
Resolved bugs (If applicable)
n/a
Details of new models (If applicable)
n/a
Risks
- No change to the current behaviour
Merge request reports
Activity
changed milestone to %v2006
added enhancement label
added 8 commits
-
43e7098a...aa9be5b5 - 6 commits from branch
develop - d7622d1e - ENH: add virtual omega to turbulenceModels
- 2319762c - ENH: register internal fields of kOmegaSST
-
43e7098a...aa9be5b5 - 6 commits from branch
mentioned in commit b69360fe
-
Adjustments for OpenQBMM pending on github 150
@albertop and @kuti - we might wish to put in a concerted effort to merge in the OpenQBMM turbulence modifications if possible.
I think it would be possible.
However, there has been a long-standing issue in OF in this area which I wanted to tackle for a while but never had a chance. Kinetic theory models and non-Newtonian stress models (e.g. Maxwell) are treated as turbulence models, but they really are laminar stress models. The current implementation prevents some developments in both areas to be easily done. For example, if one is using the kinetic theory model, no turbulence model can be selected for that phase. Similar issues affect non-Newtonian models.
What do you think about moving these models out of the turbulence framework and make them usable with turbulence models?
The kinetic theory would need to access the epsilon field so that more recent models for dilute gas-particle flows can be incorporated (epsilon is needed in some source terms for Theta), but that can be achieved with lookup. The major challenge I see is changing how the calculation of laminar stresses, which affects nearly all turbulence models, since the divergence of the stress tensor is returned by divREff and the assumption is that laminar stresses assume Newtonian flow.
