magic multi-physics / multi-scale simulation

[roambotics_scott]

In a former life I did HPC and I've cautioned often and strongly here against doing CFD because even for people who know what they're doing it's almost guaranteed to give pretty but misleading or meaningless results for people who don't know what they're doing.

That said, there is clearly a great deal of demand for CFD (not certain if that's real of just «it'd be cool..», but still).

One highly non-trivial thing that could be done is automatic / dynamic meshing with smart multi-physics to automagically figure out what regime a particular part of the simulation should be in and adjust accordingly.

Doing this would be a huge undertaking and I remain skeptical that it's worthwhile, but it'd be awesome.

If you wanted to go nuts, you could also add chemistry (multi-scale combustion physics) and if you wanted to go really insane, something like FLASH for nuclear CFD simulations (though you'd probably run into ITAR issues if you did that..)

[kb9ydn]

Actually I think I would say the opposite.  The fact that people don't know how to use it means that it SHOULD be made available for them to try out and hopefully learn something from.

In terms of CFD specifically, I think even a very simplistic form of it could be useful, if only just for visualizing how a fluid would flow through a cavity based on a few variables.  You might not be able to get any significant quantitative results, but just watching the relative movement in different places would be helpful for flow optimization (for example).

C|

[roambotics_scott]

Totally get why you'd say that @kb9ydnand it sounds great but the problem is that they'd have no idea the physics was wrong. What's bad is that the results will probably certainly look plausible but be meaningless if not misleading

It's like giving people normal mode analysis, showing the exaggerated basis deformations, and expecting them to actually understand what those mean

[kb9ydn]

Well you could say the same thing about CAM.  You can simulate a tool path and it looks great on the screen, but then try to run it and breaks a tool, throws the part across the room, or even crashes the machine.  People will always misuse tools, but I don't think that's necessarily a good justification for not giving them said tools.  They have to learn somehow. 

That being said, I think good documentation goes a long way towards helping people use the available tools properly.  [hint] 

C|

[roambotics_scott]

The difference being that most people actually using CAM will actually run their G-code while almost nobody doing CFD will do any sort of validation

I've seen way too many grad students who went on to get PhDs in chemistry, physics, and engineering unknowingly screw up simulations badly even with good documentation to think putting anything but foolproof CFD in F360 is a good idea. Hell - I've seen too many people here misunderstanding NMA and other simulation tools already in the software to think it's useful.

Even with a very solid understanding of the underlying physics and Navier-Stokes' equations, it's hard to know where the Reynolds Number will be, what the right mesh type and resolution should be, and what approximation(s) are appropriate. That's not stuff that you can just wing and if the software isn't smart enough to let someone very easily specify the materials and boundary conditions without having to figure out the details themselves, it's just going to be GIGO - burning a ton of compute resources to make pretty but ultimately meaningless movies

[kb9ydn]

That is true, it's much easier to validate results from a CAM simulation than a CFD simulation.  And certainly it would have to be VERY simplistic (at least at first) with a lot of built in assumptions.  Have you seen Solidworks' FloXpress?  That's kind of what I'm thinking of for Fusion to start out.

C|

[vex]

I think CFD would be great. It doesn't need to be the magic wand that fits every problem. Even a simple caveat that deals only with incompressible flow and few basic turbulence models (LES and k-eps) would probably handle 95% of what people are considering: I have yet to see anyone attempting to simulate or model a tuned intake runner for pressure wave/Helmholtz utilization; nor trans-, super-, or hyper-sonic vehicles. I have done rocket motors in Inventor, but I shoe horned in the chemistry/combustion analysis using prop3(?) and standard burn rates and geometries by hand.

Mesh optimization could probably be handled similarly to their FEM optimization already built in. Granted, the key would be ensuring enough BL cells to ensure adherence and transition solution. Autodesk's acquisition of blueridge numerics (now offered under Autodesk CFD, I believe) should provide a very good starting point--not to mention Algor multi-physics already in their portfolio.

I know in many of my cases it's not so much "are the results reflective of real world application" but more design choice requirements: does the flow require tripping ? Does the manifold produce too much turbulence and starve a runner under condition x-, y-, z-? In that case, simply knowing your boundary conditions is enough to start you down the path of enlightened design choices.

Validation of the results should be the onus of the designer/engineer/PM/Hobbyist of the project, and not really the concern of the programmers. It's why all commercially available software packages include a lengthy disclaimer about results in the front of their documentation.

As way of something to not be overlooked: Having 2+ fluids with calculated boundary interaction would be nice (water, air, vehicle). This is something I've found OpenFoam does rather well and could probably be done here with little issue.