[nskelly]

Hi all,

I'm having a bit of a problem with a LES backward facing step.

I am replicating the Armaly (1983) paper of their backward facing step, with particular attention to the recirculation zone lengths. I have managed to create the geometry and mesh using blockMesh with extra fine mesh at the step wall and upper and lower walls, to account for the recirculation zones. However, the simulation is taking a long time to run, for a "relatively" low velocity. I am aiming to keep the Courant number < 1, as I heard that is best for LES, but this means having a time step of 1e-6.

One of the reasons this may be is because of the mesh. Running a checkMesh has shown that I have a large aspect ratio, resulting in the checkMesh to fail. I suppose what I want to know is that if it is the aspect ratio is causing the simulation to solve very slowly.

I have attached an image of my mesh, blockMeshDict, checkMesh and fvSchemes.


______________________________________________
checkMesh
______________________________________________

Create time

Create polyMesh for time = 0

Time = 0

Mesh stats
points: 1805103
faces: 4203400
internal faces: 2996600
cells: 1200000
faces per cell: 6
boundary patches: 7
point zones: 0
face zones: 0
cell zones: 0

Overall number of cells of each type:
hexahedra: 1200000
prisms: 0
wedges: 0
pyramids: 0
tet wedges: 0
tetrahedra: 0
polyhedra: 0

Checking topology...
Boundary definition OK.
Cell to face addressing OK.
Point usage OK.
Upper triangular ordering OK.
Face vertices OK.
Number of regions: 1 (OK).

Checking patch topology for multiply connected surfaces...
Patch Faces Points Surface topology
outlet 1600 2403 ok (non-closed singly connected)
inlet 800 1203 ok (non-closed singly connected)
lowerWall 1800 2706 ok (non-closed singly connected)
upperWall 1800 2703 ok (non-closed singly connected)
stepWall 800 1203 ok (non-closed singly connected)
front 600000 601701 ok (non-closed singly connected)
back 600000 601701 ok (non-closed singly connected)

Checking geometry...
Overall domain bounding box (-0.2 -0.004889 -5e-05) (0.5 0.0052 5e-05)
Mesh has 3 geometric (non-empty/wedge) directions (1 1 1)
Mesh has 3 solution (non-empty) directions (1 1 1)
Boundary openness (-4.47912e-20 -2.15041e-17 -2.43961e-13) OK.
***High aspect ratio cells found, Max aspect ratio: 10307.9, number of cells 94606
<<Writing 94606 cells with high aspect ratio to set highAspectRatioCells
Minimum face area = 2.05117e-12. Maximum face area = 3.43272e-07. Face area magnitudes OK.
Min volume = 1.02559e-16. Max volume = 1.71636e-11. Total volume = 6.0845e-07. Cell volumes OK.
Mesh non-orthogonality Max: 0 average: 0
Non-orthogonality check OK.
Face pyramids OK.
Max skewness = 2.4958e-12 OK.
Coupled point location match (average 2.78878e-17) OK.

Failed 1 mesh checks.

End

[piu58]

If you want to increase accuracy by refining the mesh, this is only possible to a degree with only refinig one coordinate (as you did). If the aspect ratio is too bad, the system of equation need more time to solve. I recommend to increse the resolution in the y coordinate too. If the mesh gets too large then and may be more suitable to reduce the resolution in the recirculation area.

[nskelly]

Hi Uwe,

Thanks for you advice, I will increase the resolution in the y direction as you said.

I am actually trying to replicate a mesh that is from another paper and they had a very fine mesh near the step however their domain was a lot shorter, and was restricted to laminar solver. So you may be right in reducing the resolution of the recirculation zone.

[nskelly]

So I managed to refine my mesh and got my aspect ratio down to a reasonable 48 rather than 10000 but I am having problems with the simulation run time.

I am just looking at a relatively low Reynolds number Re=389 using laminar solver, not LES atm. However, I still need a small timestep 1e-6, which isn't ideal for the amount of simulations I need to conduct. Currently, 0.2s takes 1 day to solve, this is when I am parallel processing using 15 processors.

Does anyone know why this would be? I do not think the mesh is the problem as my checkMesh passes.

__________________________________________________ __
checkMesh
__________________________________________________ __
Create time

Create polyMesh for time = 0

Time = 0

Mesh stats
points: 1127202
internal points: 0
faces: 2243600
internal faces: 1116400
cells: 560000
faces per cell: 6
boundary patches: 7
point zones: 0
face zones: 0
cell zones: 0

Overall number of cells of each type:
hexahedra: 560000
prisms: 0
wedges: 0
pyramids: 0
tet wedges: 0
tetrahedra: 0
polyhedra: 0

Checking topology...
Boundary definition OK.
Cell to face addressing OK.
Point usage OK.
Upper triangular ordering OK.
Face vertices OK.

Number of regions: 1 (OK).

Checking patch topology for multiply connected surfaces...
Patch Faces Points Surface topology
outlet 200 402 ok (non-closed singly connected)
inlet 100 202 ok (non-closed singly connected)
lowerWall 3400 6804 ok (non-closed singly connected)
upperWall 3400 6802 ok (non-closed singly connected)
stepWall 100 202 ok (non-closed singly connected)
front 560000 563601 ok (non-closed singly connected)
back 560000 563601 ok (non-closed singly connected)

Checking geometry...
Overall domain bounding box (-0.2 -0.004889 -5e-05) (0.5 0.0052 5e-05)
Mesh has 3 geometric (non-empty/wedge) directions (1 1 1)
Mesh has 3 solution (non-empty) directions (1 1 1)
Boundary openness (4.28338e-20 -2.12895e-17 -3.085e-14) OK.
Max cell openness = 3.30554e-16 OK.
Max aspect ratio = 48.1711 OK.
Minimum face area = 1.05228e-10. Maximum face area = 7.74067e-08. Face area magnitudes OK.
Min volume = 1.05228e-14. Max volume = 7.74067e-12. Total volume = 6.0845e-07. Cell volumes OK.
Mesh non-orthogonality Max: 0 average: 0
Non-orthogonality check OK.
Face pyramids OK.
Max skewness = 9.09589e-12 OK.
Coupled point location match (average 2.77471e-17) OK.

Mesh OK.

End