Learning Forum

[papp]

Hello,

I would like to make sure of the following:

If a certain Fidelity value is set (e.g. to 100% on the slider), does Discovery Live produce the same mesh and results every time the simulation is restarted on the same GPU (Fluids module)?

My experience suggests otherwise, but I would like to rely on your opinion. I have also come across with more extended mesh on thin objects at lower Fidelity than at 100%.

I know that the software sets the Speed/Fidelity range to each worstation's GPU capacity. Is it possible that Discovery Live takes the current GPU usage into account as well?

I'm aware of the fact that DL is supposed to help the rough design, and we should obtain qualitative data only, but I believe reproducibility is crucial nevertheless.

Thanks for the explanation in advance.

Bálint

[Brian Bueno]

Bálint Papp

I will consult with the development team and get back to you regarding this issue.

[m.hamza_amu]

Hi  BBueno_ANSYS ,

 

Bumping up the topic again. Did you manage to get feedback from the development team on this topic?

 

Thanks

[papp]

Hi gm86 ,

I'm confident that every time the simulation is restarted, the exact same output data are created. This implies the genereration of the same mesh.

You can try this yourself: set up a chart or probe, and run the Fluids simulation for excatly the same time (this can be chosen arbitrarily) a couple of times. The numerical output on the calculator should be the same. As this applies to all time instants, this means that the same time series are computed at your selected location each time, which backs up the theory of having the same mesh created by the software.

My experience was poorly grounded bact then, and made prematurely. A more thorough investigation into this phenomenon (like the above experiment) proved the reproducibility of the results.

Setting the resolution at maximum Fidelity is only dependent on the GPU parameters of your computer, the usage is not taken into account, I believe.

In some cases, lower Fidelity meshes can fit a geometry better. This is probably a result of the automatic discretization schemes that Discovery Live uses, but in general, higher Fidelity values provide a better fit, especially for non-thin objects.

Bálint

[Brian Bueno]

Thanks for your input Bálint Papp ! gm86 we've also posted this benchmarking and verification document: https://discoveryforum.ansys.com/t/m2rw27/discovery-verification-and-benchmark-cases

[m.hamza_amu]

Thank you for your detailed input  Bálint Papp , I am new to the forum and just want to take a moment to thank you for sharing the results of all of your previous benchmarking studies in this and other posts. Highly appreciated!

 

Thanks  BBueno_ANSYS for sharing the official benchmarking results. This would all be very helpful.