View Issue Details

ID Project Category View Status Date Submitted Last Update 0003701 OpenFOAM Bug public 2021-07-29 12:35 2021-08-03 21:38 time-trader henry normal minor always closed suspended Linux Ubuntu 20.04 LTS 8 0003701: porousSimpleFoam solver results are highly sensitive to the mesh size at the porous media boundary The behavior is most likely due to the interpolation of pressure values to the faces of the porous zone. The error on pressure drop is proportional to the size of Cells at the porous zone boundary. In case of Forchheimer flow: ΔPerror ≈ 0.5* rho * fx * dx_cell*ux^2 Where: dx_cell = size of the cells at the porous zone boundary in the flow direction fx = Forchheimer coefficient in the flow direction ux = flow speed 1. User a rather coarse mesh at the boundary of the porous zone. 2. Use standard settings in fvSchemes ie. grad Schemes { default Gauss linear; } and interpolationSchemes {     default linear; } 3. Run a porousSimpleFoam solver Attached is a simple benchmark test, with several hex meshes generated by blockMesh, to demonstrate the behavior. With the thickness of the porous media of 0.5 [m] and fx = 20 [m^(-1)], the expected value of pressure drop:  ΔP = 0.5*1[kg/m^3]* (20 [m^(-1)] *0.5 [m]) *(10[m/s])^2 = 500 [Pa] The value of pressure drop from porousSimpleFoam: with 5 cells in flow direction, regular mesh size of 0.1[m] : 455 [Pa] with cells size in flow direction of 0.1[m] at boundaries and 30 cells of 0.01[m] inside the porous medium: 471 [Pa] with cells size in flow direction of 0.01[m] at the boundaries and 3 cells of 0.16 inside the porous medium 494 [Pa] with 5 cells in flow direction, regular mesh size of 0.1[m] and cellLimited interpolation scheme for p : 495 [Pa] Therefore, it was possible to obtain the results more in line with the analytical prediction with following strategies: 1. Decrease the size of cells at the boundary (only the boundary refinement is necessary). or, alternatively: 2. Limit the extrapolation on pressure gradient: gradSchemes {     default Gauss linear;     grad(p) cellLimited Gauss linear 1; } and introduce the cubic interpolation scheme: interpolationSchemes {     default cubic; } No tags attached.

Activities

 2021-07-29 12:35 reporter BenchmarkCase.zip (91,924 bytes) 2021-07-29 12:48 manager   ~0012117 It is not clear if you need anything from us on this. If you would like to fund development of the porous media implementation please contact us directly. 2021-08-03 21:38 manager   ~0012118 Waiting for response