Combining fusion 360 generative design with CFD

npundir
Community Visitor

Combining fusion 360 generative design with CFD

npundir
Community Visitor
Community Visitor

Hey,

 

Currently generative design in fusion 360 focuses on strength and material utilization, but can generative design be used to iterate design based on CFD simulation results

 

Thanks

0 Likes
Reply
940 Views
4 Replies
Replies (4)

OceanHydroAU
Collaborator
Collaborator

What are you designing?

 

I did my generative-design work I need outside Fusion360 (took 8 years to write the code and many months to run), and then I wrote an add-in that makes use of the results - if you're seeking airfoil-related CFD, what I've done might help?

(e.g. here: https://forums.autodesk.com/t5/fusion-360-api-and-scripts/add-in-announcement-hydrofoil-and-airfoil-... )

 

If you're a programmer, the open-source package OpenFOAM looks pretty awesome, and Fusion360 add-ins are not too hard to write - combining those 2 things with a "generative middle bit" would result in what you're looking for.

 

I've done some work with OpenFOAM - it's very important to get all the fiddly bits right (no doubt the same with all CFD's) and a great many of those "fiddly bits" are tricky math concepts, and a fair few of the time-consuming bits are "programmer laziness" (grids and their resolutions etc - none of that should be left to the user, it's not very hard for any automated process to figure that out itself) - so while it might produce nice colourful looking reports, more often than not the stuff in them is bogus.

 

I've looked at a lot of other people's CFD work - the vast majority of stuff you can find is garbage (riddled with obvious mistakes) and the few things that are not totally wrong are typically supported by experimental data (e.g. wind tunnels etc) which in my vast experience just proves that they threw away their "wrong" stuff until their answers matched the experiments - so once again - even those examples are bogus too.  We never got to see what they got wrong at the start, and if they didn't do the experiment, they wouldn't have known they were wrong.

 

The last one I saw yesterday included charts that clearly showed that their "optimized propeller" UNDERperformed in comparison to an off-the-shelf shop-bought one (kudos to them for being honest), except in their write-up and synopsis they said their CFD one worked best (not supported by the chart).  They also defined the area of a circle to be (PI*D^4)/4 - I can't wrap my head around that - but there is one unit case where that flukes to work out - except they were not using that one case, so as best I can tell, all the results from all their CFD work was based on totally wrong math...

0 Likes

hindmas
Explorer
Explorer

Hey @npundir - you are correct, the current physics of generative design are based on structural objectives and constraints.  Internally, we often take design outcomes from GD and run them in a standalone CFD application to understand their performance.  We totally agree that the inclusion of flow and thermal physics into generative design would be a huge benefit.  We'll be able to share more insights into what we're thinking soon; thanks for the question.

1 Like

OceanHydroAU
Collaborator
Collaborator

It's such a privilege to have such easy contact to the Autodesk managers in charge of these things!

 

An amateur observation I have based on some small experience: today is different to years past - we now have widespread 3D manufacturing, so the very idea of CFD is outdated.  Math used to be necessary because it was infeasible to do real studies, which is no longer the case.  The huge problem with Computational fluid dynamics is that first word: "Computational" - to get that bit right, requires such superhuman understanding of every single complex facet of the problem that what we typically find in practice is that none of it really works properly.

 

Bottom line: do not believe the hype - pretty graphics and high-brow reports do not mean the answer is right, or that if it was right, that they knew it was so before they went and did a test to check it (i.e. they cheated, and changed their answer to match the question afterwards).

 

If you're evaluating stuff, do not look at their claims - give them a few tests where they cannot cheat (do not let them have the answer, and do not give them enough time to "validate their results" outside of their software), and see if it really works.

 

If you're still just thinking about this - outside the box might be a nice place to start.  Maybe instead of "CFD" an "AFD" answer would be far better: "Actual Fluid Dynamics" - an automated system to build and test a real design with embedded sensors, which is guaranteed to return real results, because it *was* real.  (or perhaps a hybrid, where customer pay more for real, or even an A.I. version, where the A.I. learns how to properly emulate "real" by first creating genuine "real" and learning from how that really behaves).

0 Likes

hindmas
Explorer
Explorer

@OceanHydroAU  this is water on our wheel!  Especially in the case of flow and thermal insights, we've been looking to provide more focus on users being able to select the insight or outcome desired vs the tool or process required to get there.  An example of this is in the newly debuted 'electronics cooling' solution for Fusion 360.  We are working to change to a common language interaction that defines what you want to know vs an articulation of all of the physics and study components that a user would need to be knowledgable of.

 

As a followup to the originating thread, check out the 2020 roadmap update that was published today (hint:  generative fluids).  https://www.autodesk.com/products/fusion-360/blog/fusion-360-roadmap-update-2020/

1 Like