[alinik]

Any suggestions Edmund?

[ghorrocks]

Can you explain what is happening around the 50th time step? I recall you have rod things which are closing up a gap. Is this just before the rods close the gap? An image of what the mesh looks like would help.

[alinik]

Actually I have posted a picture couple of posts back and if you take a look at it you can see that the rods are there only to produce some wakes. They are not closing up a gap.
By 50th time step the domain on left has barely moved. it has displaced only about 0.025m(2.5 cm)

[singer1812]

Did you verify that flow is actually moving through the interface in that little bit that did run? Just to verify that it is setup correctly.

But I am beginning to suspect mesh quality.

[alinik]

Here you can see some pictures. I do not know what is causing the problem?
For initial condition I have set the interface frame change type to "none" and you can see that the results are reasonable(the first two pictures). Then I set this solution as initial condition for the transient solution and as you can see in the next two pictures the velocity value goes through the roof.(for transient case the interface type is set to TRS)

any idea why this is happening?

[alinik]

the last two images are taken at the timestep that the solver crashes

[ghorrocks]

This happens when the simulation diverges. To fix this you need to improve the numerical stability. You can do this by some combination of:
* Better quality mesh
* Smaller time steps
* Double precision numerics
* Better initial condition.
* Use first order differencing (but this will compromise the accuracy)

[alinik]

I asked ansys support team to help me with that and they said that Transient-Rotor stator interface model(TRS) will not do the transnational sliding mesh modeling...It only can do the rotational sliding mesh

End of Story

[ghorrocks]

TRS works fine with a translating mesh when the translation is done as a moving mesh. If you are using a translating frame of reference it does not surprise me things don't work - that's why translating frame of reference is not a fully released feature.

[alinik]

Glenn,

Thanks, but I used moving mesh as you said. not moving frame of reference. I have spent over a month struggling with TRS and translational sliding mesh and now I can say with 100% confidence that CFX cannot do that.
especially after I opened a thread in CFX support portal and worked with CFX representative for more than a week. After that he told me explicitly that CFX cannot do that and suggested using a rotational sliding mesh with relatively large radius of rotation instead.

[alinik]

Glenn, If you have done that before and still have a .pre file or the ccl please give it to me somehow so that I can show it to CFX technical support.

[ghorrocks]

Interesting - because I have 100% confidence that is does work (well, at least did work) because I have used it several times. I will admit that the last time I used it was many years ago although I cannot remember exactly which version. The models were a 2-stoke port valve engine using GGIs which open and shut as the piston slides past them and a hydraulic pressure relief valve which cracked open with a translational motion. Both these models were with a previous employer and I no longer have the files available. I think I was doing the 2 stroke engine about 12 years ago and the hydraulic valve about 8 years ago - so a fair way back.

It is possible that this functionality has been broken in recent releases. It would be a shame if it does not work any more.

[alinik]

Glenn, You still can slide a mesh relative to the other one . but you can not use TRS model for the interface between the two grids. that is the issue. TRS is not compatible with translational movement.
For non-overlapping portion it can only uses wall type boundary condition and you can not change it to any other type of boundary.
In the picture you can see that the non-overlapping portion at the end of the simulation is treating as a wall

[ghorrocks]

I see, we might be talking to crossed purposes here.

The image you shows has the interface functioning perfectly - so it looks like a TRS interface is working fine. You issue seems to be the wall which is generated in the non-overlapping portion. In my case of port valve engines the non-overlapping section indeed was a wall so the default behaviour is correct for my application.

But would I be correct in saying that you are trying a model with translational periodicity, and you want this interface section to match up with the flow on the other side of the periodic condition, presumably at the top of the domain?

[alinik]

This is what TRS is good for. Aparently it is not working with translational movement of the domains

[ghorrocks]

OK, I think we have made some progress. TRS works fine for matching up the mesh interfaces as long as you are happy for the non-overlapping section to be a wall. If you want the non-overlapping section to map around to the other side of the domain (as it does in rotational cases) then this is not supported.

Now I understand what you are talking about the recommendation is simple - rather than modelling a single periodic unit and expecting the interface to handle the non-overlapping section, you model two periodic units. The top and bottom of the model is rubbish as an artificial wall is generated, but the middle section away from the non-overlapping section should be fine. If you find your results are affected by proximity to the non-overlapping region then model 3 or more periodic units until the effect is small enough to ignore.

This solution is not ideal as it doubles the size of the simulation, but at least you can run it without modelling the whole thing.

[haribhaskaran]

I am using a transient rotor stator interface between one rotating domain and one stationary domain. For this kind of an interface, if I am not wrong, I think the non overlapped regions should behave like a wall.

However in my case, the flow occurs even through the non overlapped regions. The simulation is a transient simulation and the overlap area changes with time. What could be going wrong ?

[Antanas]

Originally Posted by haribhaskaran

... I think the non overlapped regions should behave like a wall ...

I think you have to set this behaviour manually.

[haribhaskaran]

Thanks for replying Antanas

How do I set it manually ?

[Antanas]

Originally Posted by haribhaskaran

Thanks for replying Antanas

How do I set it manually ?

Inside properties of interface side that is not fully overlapped. There is nonoverlap conditions tab.