[Nimish]

Hello,

I am running a duct flow analysis with compressible air using LES. The duct length (along X) and height (along Y) is considerably larger than the depth (along Z) of the duct. The duct has inlet, outlet, opening and wall boundary conditions (BCs) as shown below. Also, the planes along the depth direction have symmetry BCs. The inlet velocity BC is specified as the profile data using outlet velocity output from the steady state analysis, and also the the domain is initialized using steady state analysis.

opening
__________________________________
inlet __________________________________ outlet
wall

At low Mach numbers (~0.1), the analysis runs without any problem. When I rerun the problem with higher Mach number (~0.4), with same mesh and same boundary conditions, the analysis gives "floating point exception: overflow" error. I ran the low Mach number problem using "constant property air" and no heat transfer. For the higher Mach number, I am using air as compressible gas, and "total energy" as the heat transfer model.

When I process the results just below this error, the streamlines show the flow going from inlet to the opening. The Y velocity component is supposed to be very small compared to the X velocity component, because majority of the flow is expected to be along the duct length.

I think the mesh refinement might not be sufficient to handle LES type simulation. I am going to try refining the mesh, and improving mesh quality, but wanted to know if someone can offer any additional direction to look into. ANy help will be greatly appreciated.

Thanks.

[ghorrocks]

You can't use symmetry planes in LES. Symmetry does not exist in LES, so you have to model the whole thing.

Floating point exception: FAQ: https://www.cfd-online.com/Wiki/Ansy...do_about_it.3F

But your comment suggesting the flow is going out the opening sounds like the problem. Are you sure your boundaries are set up correctly? If you got the flows and pressures at the boundaries wrong then this sort of thing can happen.

[Nimish]

Thanks Glenn. Yes, I understand that symmetry is not the best option for LES, I have to use it to reduce the domain size. I am rerunning a case that ANSYS sent me, which too involves symmetry.

The BCs seem correct to me, I have submitted a run with a refined mesh. I will see how it progresses. Is there anything else that I need to consider?

[ghorrocks]

Using symmetry in LES has the following effects:
1) It limits the maximum turbulent structure size to half the size of the domain. If it wants to generate a structure the full size of the domain it cannot.
2) Structures on the symmetry plane are artificially constrained due to the symmetry condition (zero normal flow, zero normal gradient).

Be careful about your boundary condition setup. Refining the mesh will not help if you have a fundamental problem in your boundary condition setup.

[juliom]

Besides te issue with symmetry. The problem with compressible flows is the characteristic lines. In fact the B.C. You are using are not appropriate for compressible flows. Also check in reflective B.C.

[Nimish]

Originally Posted by ghorrocks

Using symmetry in LES has the following effects:
1) It limits the maximum turbulent structure size to half the size of the domain. If it wants to generate a structure the full size of the domain it cannot.
2) Structures on the symmetry plane are artificially constrained due to the symmetry condition (zero normal flow, zero normal gradient).

Be careful about your boundary condition setup. Refining the mesh will not help if you have a fundamental problem in your boundary condition setup.

Glenn, the size of the object interrupting the flow is fraction of the domain depth. Therefore, both of the points mentioned by you are expected to have little effect on the solution. moreover, i will be doing sensitivity study with respect to the domain width size to confirm this.

[Nimish]

Originally Posted by juliom

Besides te issue with symmetry. The problem with compressible flows is the characteristic lines. In fact the B.C. You are using are not appropriate for compressible flows. Also check in reflective B.C.

Thank you for your response, can you please be specific about which BC being inappropriate?

[Nimish]

UPDATE:

by changing the initial conditions slightly, the floating point error is no longer showing up. but, the simulation is still showing streamlines going out from the opening. the majority of the flow should go out from the outlet, and not through the opening. i am planning to increase the domain height to see if that can help. Does anyone have any other insight to prevent flow going out through the opening?

[juliom]

Grid sensitivity in les is Only possible using dynamic model or scale similarity. In regards to the BC, I usually use for compressible flow zero gradient at the outlets. I avoid using opening in compressible flows, thus I usually use far fields where you specify free stream conditions the same for inlets.

[ghorrocks]

Does anyone have any other insight to prevent flow going out through the opening?

It is because of the way you have set up your boundary conditions. If you want us to help you find exactly what issue is causing it you will need to post more details - images of your geometry and mesh and your CCL.

Therefore, both of the points mentioned by you are expected to have little effect on the solution. moreover, i will be doing sensitivity study with respect to the domain width size to confirm this.

It is good to hear you have considered this issue and will be checking it later on.