[manjunathanand]

Hello All,

I am running a radial fan simulation using Solidworks 2014 Flow Simulation tool. The mesh settings used is at level 5. I am defining the rotating speed of the rotating region.

The problem here is that I am unable to reproduce the same results each time when I solve using the same mesh generated. The values are quite random and the number of iterations are different each time.

The parameters that I am looking at is the volume flow rate and the air velocity.

Any suggestions would be helpful.

Thanks

[lcarasik]

Are you able to contact Solidworks directly? This seems like a question for them directly. A solver should never be giving out a random set of answers. You're not going to get the exact same answer every time, but they should be in the same range.

[manjunathanand]

Hi,

Thank you very much for responding to my request. I contacted Solidworks and they are yet to respond.

The root cause of this problem is the oscillating results and the solver finishes randomly without knowing where to stop. I am am not aware how to deal with oscillating results that will never converge. Any suggestions would be helpful.

Here is the snapshot of three iterations and the results vary b/w 12 - 20m/s

[lcarasik]

That is really concerning. You should have more control over solver in terms of how long it runs and how many iterations it goes through. Are you doing unsteady or steady simulations?

[tas38]

Originally Posted by lcarasik

That is really concerning. You should have more control over solver in terms of how long it runs and how many iterations it goes through. Are you doing unsteady or steady simulations?

I agree. You should have enough control to run transient vs. steady-state.

Also, it seems me that your results may in fact oscillate due to the nature of the problem. In other words, the steady-state solution is really only psuedo-steady, i.e. periodic. Look for example at a vibrating mass with harmonic forcing or Stokes 2nd problem (motion of a viscous fluid locate above and oscillating plate.)

Check and see if the frequency of the oscillations you presented coincide with the blade pass frequency of the fan. If so, the solution is akin the aforementioned examples.

[manjunathanand]

Hi Lane and Troy,

Thank you again for responding to my question. I tried having better control over solver conditions. Unfortunately the frequency of oscillation is not repeatable during each iteration and so I could not decide where to stop. Also for the goals, the flow simulations has only max flow rate conditions. Average flow rate could have been a good option to converge.

@ Troy, could you please lead me to examples you have mentioned. I will also look for them.

Meanwhile I will check the frequency of oscillation with the frequency of blades and let you know.

Thank you and I really appreciate your help.

[Murph]

manjunathanand,

I have seen your thread and re-ran a model that I have involving an internal analysis with an idealized fan. I have the same problem, that the simulation is showing a different result even though I have the same setup and am running on the same machine. My results are close, however this should not be happening at all. It means my results are useless in this case.

I have sent my model to SW support who have sent to SW vendor. They have confirmed that this shouldn't be happening. Initially they have told me that it is something to do with the way SW distributes to multiple processors - they are looking into the issue as a matter of urgency. So they have accepted a fault in the software and are trying to address it as we speak.

I saw your post and thought it off becuase...

The solidworks solvers should give you the EXACT same result if you are simulating on the same machine. SW runs on the RANS (or DNS) equations and they are deterministic ie... same input should give you the exact same output (your convergence plots should look the same - exactly the same).

If you are on a different machine or OS - the result may differ slightly due to rounding differences of different systems. CFD software that runs on multiple platforms will run on multiple platforms, but will give slightly different results. They should be close to the same answer.

If your answers are way off, it is likely that you have a psudo-steady or unsteady system in reality. But given the error in the software - it may be impossible to say! In this case you may find difficultly getting any meaningful results from SW Flow sim.

It would help if you would also pass ur simulation onto the SW vendor. This will help them diagnose and fix the issue.

Try ur simulation again on a single core (turn of all but one core in BIOS if you can). If the problem is due to issues with multi-processing, you then might be able to get some results. if you do this let me know!

[manjunathanand]

Hi Murph, I tried resolving this issue few weeks back.

I used "Rotating - Local regions (Sliding)" and this minimized the randomness of the results to a great extent. There is no change in the boundary conditions or the definition of the rotating regions. Furthermore, I had better control with the results using transient state analysis though I am still trying master this concept.

Can you share the model/study you have been after, what type of meshing has been used for the study?

I will run the single core study as suggested by you and will let you know the outcome.

You can also find this discussion in Solidworks forum
https://forum.solidworks.com/message/461956