Accuracy of Models Inside Flow

[chrisbrinzo56]

Hello all, I'm new around these parts but I've been working with Autodesk programs for a while. I was excited to get my hands on Flow Design and test models out. However, I ran into an issue where my data was near exactly the same or opposite from what I was expecting when comparing two different models.

Below I've attached some pictures of the data I was receiving while testing an F1 car in the wind tunnel. One model with the DRS (Rear Wing) open and another with the DRS closed.

I would like to know if there is a limit to the accuracy of the models inside of Flow Design.

Open DRS:

Closed DRS:

Open DRS Close View:

Closed DRS Close View:

[Jon.Wilde]

Hi Chris,

Can I ask, did you change the Resolution under Simulation Settings at all? The higher this is, the more accurate the solution should be.

Regards,

Jon

[hartogj · Accepted solution]

Hi Chris,

Thanks for sharing this one. Interesting application.

I would echo Jon's comment on resolution - going up in resolution will help capture that part of the geometry better. You might start by going to 150% or 200%.

You will also certainly want to run this in 3D. The screenshots show a 2D run where all of the air has to go over the intake, which means the spoiler is almost entirely in the wake for both DRS on and off cases. In real life, a lot of the flow that reaches the spoiler is likely going around it. You need to run in 3D to capture this.

Also if you are looking at drag values let it run until the solution is fully stabilized. In the screenshot it looks like it is still "transient" meaning the solution is still evolving.

Lastly, it looks like we are working with a scaled model of the car. If this geometry went through several conversions from CAD it's possible the units info was lost (some formats don't even contain a reference to units). If you want to model it at full size, go to simulation settings and change the model dimensions to rescale the model. You can check the scale by looking at the wind tunnel dimensions in the top right corner of the screen.

Interested to see what you find.

[Royce_adsk]

I don't disagree with anything said before, but I wanted to highlight the 2D crossectional and where the mesh is probably located.  Once you increase the resolution you should see smoother velocity profile.

[chrisbrinzo56]

Thanks for the replies everyone!

I have managed to get substantially better data by changing the scale of the model and raising the resolution. Initially I had tried changing the resolution to the highest setting but it still wouldn't produce the results I was looking for. However, when raising the resoltuion to 350% and setting the model scale to inches from millimeters I started getting accurate data.

I'll post some pictures of the new data after I can compare the two models. It takes roughly 15 minutes for the flow data to solve.

[Royce_adsk]

Awesome!

Do you have a new picture you could share with the good results? I would enjoy seeing it.

Cheers,

[chrisbrinzo56]

I've been searching for the right method for getting the best representation in aerodynamic differences between the two models, however, it takes such a long time to simulate the air flow that it has been very difficult finding the best representation.

Do you have any recommendations for settings that I should use?

I have been using the 3D flow lines, but regardless of air velocity, the images have only been very minutely different. I seem to get much more detailed information from the 2D flow lines and I also get a much more detailed diffusion wake when the resolution is lower.

Here are some results I have so far:

Closed DRS:

Open DRS:

[Royce_adsk]

What if you were to seed the traces just upstream of the wing instead of near the inlet of the wind tunnel.  This way you isolate that componenet more and hope to get traces to go through and around the wing.

[sabaman]

Sorry to just drop this question in here as im having the same kind of problems but im also wondering, is there a reason why i cant go above 400% resolution, i tried project falcon about 2 years ago orso and i can remember that it did go higher if im correct, i could be wrong,...

[hartogj]

You may see the resolution being limited by the amount of memory available. If you go beyond what the machine can handle, Flow Design should limit the resolution to something more moderate. Do you have an idea of how much memory you have available and what kind of memory is being used when you go beyond 400%?

As an aside, for typical geometry and wind tunnel sizes we normally see that something in the neighborhood of 200% is sufficient. Some models may benefit from more but for most models the need to go beyond that is pretty rare.

Something else to be aware of is that the initial builds of Falcon 2 years ago used a somewhat different definition for mesh density.

[sabaman]

I dont know, i have a car model right now inthere at 400% res, takes 5,5gigs of ram, i have 32 available so even if 500% would quadrouple that (if thats the case) it still should be able to handle it, but ill try some of the options like changing the units and the size of the model.

Also, theres no way i can go past 400% it doesnt go higher, even if i fill in an amount?

Another thing, correct me again if im wrong but did project falcon have gpu support? is there anything like that planned?

And while im at it, another question, this is just in my lack of knowledge of these things, but why is the air so much detached from the actual surface?

Regards,

Sabino,...

[mayur_dhumal]

Hi Sabino,

Simulation Resolution is limited upto 400% that to is very high level, you can usually work with 200% odd figures & it will giv you more accurate results, so no need to go for 500%.

You can move the plane on the model so that air flow can be on model.

Regards,
Mayur

[sabaman]

Thanks Mayur,

The reason i wanted to go higher was simply cause of the voxel size being bigger than for instance the distance between 2 wing elements,

Now this is all ofcourse in my very limited knowledge so if you say theres absolutely no reason or the voxel size means something else maybe?

If anyone could explain the voxel size thing i would love to hear it,...

[Radon_Autodesk]

Hello,

I have a question related to resolution. I am trying to run a grille that contains openings as small as 10 cm, and I need to compare two grilles that have differences in the shape of those openings, I was using the highest resolution (400%) and 3D but it takes forever and the drag graph data does not make too much sense as it shows a negative average drag value and after 7 hours is still transient, my computer has 8 GB RAM and a 3.1 GHz processor; will I ever be able to obtain some data that shows the differences? should I scale the model? or use only a fraction of my model? thank you