Different results with Symmetry
 

Hallo Everyone,

I simulated a rectangular duct one time with and one time without symmetry boundary. I just wanted to obtain the velocity profile at the cross-section of the channel. But the results are different. (See the pictures below)

Do you have an idea why there is an such a difference?

Kind regards!

With Symmetry
http://imagizer.imageshack.us/v2/800...0/542/eaba.png
http://imagizer.imageshack.us/v2/800...0/138/r29x.png

Without Symmetry Boundary
http://imagizer.imageshack.us/v2/800...0/836/ow04.png

Before you start thinking about any physical reason for this, first you must make sure your result is an accurate representation of the simulation you set up. Is the simulation fully converged? Is the mesh fine enough (have you done a mesh sensitivity study)?

Thank you for your reply.

All simulations converged with a RMS Residual Target below 10^-5. I've done a mesh sensitivity study. I set up different meshes with unstructured tetrahedral objects with three mesh sizes and compared the velocity profiles of the simulations. After the second refinement the max velocity in the middle of the duct does not change anymore. Is that the correct way?

It sounds like you have covered the basic numerical accuracy stuff.

Can you explain what you are modelling? Especially what is happening upstream and downstream, with what boundary conditions you have applied.

This simulation is just about a straight rectangular duct to obtain the velocity profile. The profile will be used as initial condition for an inlet of a microchannel with different geometries, so you can save on the inlet zone in the following simulations.

The boundary condition at the inlet is a constant velocity over the cross section. The outlet is an OUTLET with average static pressure of 0. The wall is a non slip wall.

The profile is not changing upstream and downstream.

I don't know, why the use of symmetry ends in different results here.

http://imagizer.imageshack.us/v2/800...0/842/ipti.png

Have you checked the secondary flows? The symmetry plane might be stopping a secondary flow.

Plot vectors in the YZ plane on your cross sections and I bet the secondary flows are different.

Hmm, you mean the vectors in the pictures below? The flows are different but 10^-3 to 10^-4 less than the primary flow. Does this have a effect on the primary flow?

Maybe I don't understand the concept of the symmetry boundary, but if the flow is the same on each side of the plane, a symmetry boundary can be applied.


Symmetry
http://imageshack.com/a/img822/9061/fiqj.png

Without symmetry
http://imageshack.com/a/img845/1527/gome.png

You have not shown the one which looks different. The no symmetry and the symmetry plane on the XZ plane look the same in your first post. It is the one with the symmetry plane in the XY plane which looks different.

I actually placed the wrong picture in the first post. The profile is different for both symmetry boundaries.

Symmetry
http://imageshack.com/a/img200/8889/tx8.PNG
No symmetry
http://imageshack.com/a/img836/5332/ow04.png

That changes things - now I am thinking of a different error :)

Did you use a mass flow rate to specify the flow? When you use a symmetry you halve the inlet area so that would result in double the velocity due to the area being halved.

I specified the flow as a constant velocity over the cross section. In this case, I think I don't have to halve it.

OK, fair enough.

Can you post your output file?

Sure Attachment 28380

Your convergence on the symmetry plane one is quite a bit looser. Can you converge it tighter?

Also, use larger psuedo-time steps than auto time step. Run maybe 20 iterations with a time step multiplier of 10x or 100x. On long thin domains like this it can take a lot of time to get to a true converged solution when the convection time is so long.

I did the simulation again with a RMS residual target below 10^-6 and auto timescale with a timescale factor of 10. The .out files are attended Attachment 28450

It seems that the solution did not change a lot.

http://imageshack.com/a/img841/9709/kz07.png

This seems very strange.

Can you get the flow rates over these areas? If the flow is faster in one than the other then I cannot see how they are both conserving mass. Also integrate the velocity over the patch to check that as well.