problem with mixtureKEpsilon at epsilonWallFunction BCs intersected by processors boundaries
- Category: Bug
- Reproducibility: sometimes
- Severity: major
- Priority: high
- Profile: Ubuntu 16.04.3 LTS
- Product Version: OpenFOAM v1712
Summary: Problem with mixtureKEpsilon at epsilonWallfunction BCs intersected by processors boundaries.
Description:
When running in parallel, the mixtureKEpsilon turbulence model leads to unrealistic patterns in the fields of epsilon.air/epsilon.water (and consequently in the k.air/k.water fields) at epsilonWallfunction boundary condition that is intersected by processor boundaries. Cell values of k and epsilon change abruptly from one processor block to the next. Also the residuals of epsilonm stay at around 1, while all other residuals seem to converge.
I noticed that in the time directories of the individual processors, the boundary condition for epsilon.air/epsilon.water previously defined as epsilonWallFunction in the 0 directory, is automatically changed to a fixedValue BC.
The same case setup, but with the continuousGasKEpsilon turbulence model caused no problems.
Steps to reproduce:
- Change numberOfSubdomains in decomposeParDict to number of available cores. (I used 35 subdomains)
- Run decomposePar
- Run case in parallel until approx. 0.1s
- Inspect epsilon.air, epsilon .water or epsilonm field with paraview: view the long stretch somewhere between -30.5m<x<0m and rescale range to “visible data range”.
Additional Information: The initial conditions for U.water, U.air, p, p_rgh, etc. in the 0 directory are based on a stable laminar simulation of the case. Unfortunately, the case is too big to upload it directly here, but it can be found under: https://polybox.ethz.ch/index.php/s/CQKaYP9WvJKoIc4
## Reattaching the author to the issue ticket: @mbuergle ##