Abstract
The present report focuses on the validation of the Computational Fluid Dynamics (CFD) code, OpenFOAM 1.7.1, for flow over complex terrain applications. With this purpose, a series of test cases is carried out comparing its accuracy and speed with EllipSys, a in-house CFD code developped by Ris?-DTU and already benchmarked in this type of applications.
This comparison is perfomed using the well known k-epsilon (k-?) eddie viscosity model.
When investigating the wall treatment used in OpenFOAM, significant differences are noticed in its formulation when compared with EllipSys. In particular, OpenFOAM uses a Nikuradse’s roughness length (ks ) law of the wall for rough surfaces which is inconsistent with the inlet boundary condition profiles. This inconsistency is present and well documented in other CFD codes. It will typically manifest itself with the appearance of streamwise gradients along the computational domain in two dimensional flat terrain cases. These effects have shown that accuracy in complex terrain simulations can be seriously compromised.
The first part of this project directly deals with this problem using a two-dimensional test case in flat terrain.
The results show that no gradients develop in the variables in the treamwise direction.
Moreover, the vertical profiles are maintained throughout the computational domain where only significant errors are observed in the lower part. Remedial measures are therefore proposed limiting
the position of the first cell height.
Later on, OpenFOAM is tested on a two-dimensional bump and on the Askervein Hill. Its results are then compared with EllipSys. In general, remarkably good agreement is observed from upstream to the hill top when doing mesh independent comparisons.
In the lee of the hills, some discrepancies are observed. In contrast, OpenFOAM is found to be much slower than EllipSys.
The meshing SnappyHexMesh utility is also tested with OpenFOAM in a three dimensional bump case and on the Askervein hill. When dealing with irregular terrain, the utility shows difficulties in the surface layer addition phase.
It also leads to a high number of cells in big domains. Nevertheless, simulations are performed successfully on the Askervein hill and the results show an acceptable agreement when compared with EllipSys and OpenFOAM on optimized meshes.
Hello Benjamin,
I am trying to recreate the simulations you perform in your thesis, and have some convergence issues – could you provide your email for further correspondence on this matter? or perhaps could you publish the cases you used for the simulation performed in your thesis?
Thanks,
Hanan