[Ciudo]

Hello everyone. This is my first post in this Forum. I hope I'll write everything properly.

I'm worinkg on a internal combustion engine, opposed piston, 2-stroke, using OpenFOAM 11's NCC approach, trying to simulate an entire cycle (without combustion for now).

The simulation starts before exhaust ports opening, with initial values set with setFields.

Actually I'm running with multicomponentFluid solver and piston motion is provided by displacementComponentLaplacian solver which uses pointDisplacementz field file.

The current turbulence model I'm using is kOmegaSST.

The case is constructed following these procedures:

1) Creation of intake, exhaust, cylinder meshes
2) mergeMeshes -overwrite cylinder intake
3) mergeMeshes -overwrite cylinder exhaust
4) stitchMesh
5) createNonConformalCouples -overwrite
6) system, constant and BNDs copied inside the case
7) topoSet
8) setFields
9) run

The problem I'm facing involves omega instability which causes oscillations of this scalar leading to divergence. This appens near liner and exhaust ports interfaces. In this phase the exhaust piston is "going down" to the BTC.

I've tried several things to solve this issue:
1) I've run with different turbulence models. However the divergence occurs more or less at the same CAD and same cells. To be clearer, standard k-epsilon and RNG versions give me the same problem giving epsilon oscilations even early and strongly compared to kOmegaSST.

2) I've tried several refinement levels, both to cylinder and exhaust meshes, but I always see bounding box omega values in the log file, which let me understand that instabilities don't disappear.


I really don't understand the reason of this behaviour. The NCC coupling is done using the createNonConformalCouplesDict as many tutorials do.

I try to share some Dict files and pictures. For privacy reasons I can't provide the entire geometry.

As you can see in the pictures, omega oscillations occur mainly on the exhaust side and on the cells near to the patch. Someone could advice me to avoid intersections but this should not be a problem with NCC approach.

If necessary I can explain further dubts.
I hope someone can tell me something about this strange behaviour.

dictionaries_zip.zip

picture1.jpg

picture3.jpg

[Ciudo]

Does anyone have suggestions?

[dlahaye]

Not sure.

Does it help here break down the problem into subproblems, allowing to ask mor specific questions.

Domenico.

[Ciudo]

Dear Domenico,

thank for your reply.

How do you suggest to "decompose" my problem?


Claudio

[Ciudo]

dummy reply

[dlahaye]

Not sure.

Reducing geometry and physics until having a working case might be a good way forward.

[Ciudo]

I simplified the geometry of my case.

Firstly I used original initial and boundary conditions. It's curious to see that the problem is still present, meaning that the reason is indipendent of my real geometry.

I found that reducing initial pressure of exhaust lines (reducing cylinder-exhaust delta pressure up to 1 bar) makes disappear bounding epsilon and omega. However, I see that dissipation (both omega and epsilon) doesn't converge at those CAD where divergence occur with original IC's and BND's.

For "real" ICs and BNDs I get supersonic flow (about 770 m/s), while for simplified ones I'm in subsonic regime (about 120 m/s).

Please note that simplified mesh is coarse but remember the first post where I said that even with fine refinement the problem is still present.

It seems to be a numerical problem. I've tried every king of discretisation schemes I know to stabilize it (even all upwind!) but it didn't help at all.

I also tried to use simplified physics properties but the divergence remain.

What else can I check?




k.jpg

epsilon.jpg

p.jpg

U(2).jpg


Claudio

[dlahaye]

What happens in case you switch motion off?

[Ciudo]

Thanks for the hint Domenico.

I re-started the simulation from 134 CAD without mesh motion and the simulation seem to go on without problems at least until I stopped it at 140 CAD.

It may be that cell distorsions near pistons causes turbulence wall values oscillations?

If it's the case I would need to setup a non-uniform motion diffusivity in order to deform only the cells far from interfaces, while preserving those near interfaces.

I know there're the possibility to use "file" option in dynamicMeshDict which reads a surfaceScalarField file in constant folder, but I don't know how to setup the internalField diffusivity in this file.

I've already tried to use setFields but it seems surfaceScalarField objects are non currently supported by this utility.

The idea is to release motion diffusivity on cells far from interfaces.

Thanks.

Claudio