[miquiita]

Hi,

I posted a while ago in regards to I am modeling a fan in a room. My simulation gives me the following code now:
2 isolated fluid regions were found in the following set of
coupled domains:

Fan
Fluid

If the isolated regions do not have the pressure level set either
by the boundary conditions or using a reference pressure equation,
you may encounter severe robustness problems.

This situation may have arisen because a domain interface was not
properly defined during problem setup. Please carefully check
the setup.

The solver will stop now and write a results file. The isolated
regions can be visualised in CFX Post by making plots of the
variable "Isolated Volumes".

If you are sure that the pressure level is set in each isolated
fluid region then you can force the solver to turn off this check
by setting the expert parameter "check isolated regions = f".

And I am not sure what is wrong, my interfaces should be good, I have one on the top of the smaller cylinder in the room, and one on the circumference of the smaller cylinder domain around the fan and also downstream from the rotor of the cylinder domain.

I have a frozen rotor and the walls of the room are stationary. See picture.

Please help me, I need to finish this by next week.

[ghorrocks]

The error is saying your domains are not connected. This means you interface must be wrong.

[miquiita]

Hi I know what it means and I am sure the interface is correct. I was thinkin maybe my Boolean on the rotor, room, cylinder domain is wrong...

[ghorrocks]

No, it means exactly what it says. If you can suggest what could result in this error which is not an error in your interface and/or the mesh defining the interface then I would be interested to hear it.

The boolean on the rotor, room and cylinder are solid modelling things. CFX is talking about the mesh as CFX only considers the mesh. Look at your mesh for the error, not the solid modelling.

[miquiita]

Hi,

I fixed the problem, it was the boolean because I accidentally defined it wrong.

But my solution is not good, it shows one color arrows of the velocity of the fan as in the pic. Could this be becasue is has not converged enough? But I would think it would be a slightly change of color of velocity.

Or is it solely my settings of results I need to change?

[ghorrocks]

You can manually change the colour scale to be anything you like.

If you are concerned about the accuracy of your simulation look at the FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

You cannot say anything about the accuracy of the simulation from a single image like that.

[miquiita]

Hi I had a look already at the link, I just think I made a mistake somewhere and cant find it, because I changed the color in pre, but still the same velocities in the whole room. I am not sure why.

[ghorrocks]

There are always going to be high velocities near the fan and low velocities everywhere else. So you would expect the room to be filled with low velocity air. Just change thje colour scale if you want to make the colour scale more visible in the room - but you will probably make the vectors around the fan go off scale. This just means you should not describe the whole flow field in a single image, as the velocity scales are very different around the fan compared to in the room.

[miquiita]

Hi I got the model to work and the problem was a boolean operation. But my question now is that I cannot get the coonvergency down to 1e-4. I have good valid results (compared with real life experiments). I was thinking if it could be because my mesh is not good quality enough, which I am fixing, although when I did that it took 4 hours to run the simulation for 100 iterations and still did not converge down to -4, residual.

My question is probably if I should let the simulation of the room and ceiling fan run for more than 100 iterations and if I can do something more than change the mesh? my setup is good and also the geometry.

Thanks so much for any help in the matter.

[evcelica]

It is rare that a simulation would converge in only 100 iterations. If the residuals were still decreasing let it run for longer.

[ghorrocks]

Convergence in under 100 iterations is common in low Reynolds number flows like in MEMS.

[miquiita]

Sorry what is MEMS?

I was thinking that I should see a good convergence in under 100 iterations.

[ghorrocks]

MEMS: http://en.wikipedia.org/wiki/Microel...anical_systems

In my experience:
Simple simulations with straight forward flows can converge in under 100 iterations.
Most flows with moderate complexity converge in 100-200 iterations.
Complex flows (eg shock waves, multiphase, radiation, chemistry, transitional flows) will take longer than 200 iterations to converge.

This is assuming a high quality mesh and correct convergence settings. If you cannot do a high quality mesh or don't have good convergence settings then things will take a LOT longer.

[miquiita]

Thanks, all of what I thought. I just have a bit of a problem understanding how good my mesh should be, and how to refine it.

I read all the guides and looked online. So for now I made a mesh which has inflations close to the rotor and on the cylinder around it. Then also sizing it to smaller pieces.

The room walls are all bigger meshed due to they are not specific in the model I am doing.

Any suggestions of how to know when the mesh is good and how to reach it?

[ghorrocks]

Always spend some time trying to optimise mesh quality. Time spent improving the mesh will save many headaches down the track.

There are lots of threads on mesh sensitivity analysis. Do a search on the forum for them. This allows you to determine the mesh size you require.

[miquiita]

Hi,

I will do that thanks,

I tried a new mesh and converged the solution with automatic timestep. This took for 100 iterations, 4 hours and still only converged to 1e-25. Is this normal? Or is it my automatic timestep that makes it take this long to run?

[ghorrocks]

I assume your convergence to 1e-25 is a typo. Otherwise you have amazingly tight convergence....

The number of iterations to convergence depends on many things, I listed a few in post #13. The time taken to convergence will depend on the number of iterations to convergence, but also computing power available, how much parallel you are running, other work the system is doing and many other things.

4 hours is quick by CFD standards. I have done simulations which took 6 weeks to complete. But overnight is common.

[miquiita]

Hi, thanks. I am just confused to the mesh. I am really trying to get a good understanding of what is good for a fan enclosed in a room. But I cant seem to understand. I have done resizing and inflation on the blades (picture 1 and 2).

I also did patch conforming and sizing with programmed controlled inflation (pic 3,4).

[ghorrocks]

The CFX documentation has some suggestions on mesh density on surfaces. I suspect you will find your mesh near the blade is too coarse, but your mesh away from the blades is finer than it needs to be.

[miquiita]

Hi thanks, what do you mean by the cxf documentation, the mesh guide?