[Wandadars]

I'm frustrated at the current state of meshes for use in CFD applications. If we contrast the CFD mesh availability to that of easily available CAD models. There are probably millions of available CAD models. One repository that stands out to me is the Sketchup 3D Warehouse, where users can use simple search terms to populate a large list of user-created Sketchup models.

There is nothing like that today for CFD. If you want to run a simple case for flow over a sphere or over an airfoil, then you may need to generate the meshes yourself. Often for newcomers to CFD this stage may take weeks or months. It would also help for CFD applications to quickly run validation cases if we had canonical meshes to run accepted solutions like lid driven cavity flows and flows over backward steps.

I'm curious if anyone has any insight into the issues that are the impetus to cultivating a user-created database of CFD models. It could be that I'm just ignorant of some factor that is an absolute show-stopper for doing something like this. Is it because we can't decide on a good common format for meshes? CAD is pretty well centered around STL, and maybe two or three other formats. It seems like CFD has dozens of mesh formats that are all over the place.

Possible Database?
Mesh Info:
--Name of mesh
--Format of mesh( gambit, cfd++, etc.)
--Tags (viscous, airfoils, etc)
--Total Mesh Size in elements
--File size of mesh
--jpg of mesh
--Types of elements ( hex, tet, etc)
--Boundaries of mesh( unique marker numbers, names, etc.)
--Description ( notes about things like relevant flow conditions for mesh, etc)

[sbaffini]

Not only a, somehow, agreed upon, mesh format exists:

https://cgns.github.io/

but, also, it is plenty of CFD databases, or sort of:

https://www.cfd-online.com/Links/refs.html

Still, databases with grids are something that will never grow besides a certain limit, due to the required space, a thing which is not going to be better with the time, as grids tend to be bigger and bigger.

Moreover, there is more science in handling correctly and efficiently the IO of a CFD code (especially for large parallel cases) than most of the remaining things. Most of the biggest commercial players invested a lot of money in devising their particular IO with their particular input format.

Even if they also provide different input forms (and they do), they, or their clients, are not easily going to abandon their chosen format (there is also no scientific/commercial point in this, especially if agreed upon formats are not proven to be better than proprietary ones). This is even more true if you consider that some CFD vendors have implemented specific models/capabilities which are not used (or even known) by others and that require specific IO (a thing which is not going to change as, hopefully, the more we go on the more capabilities are developed and implemented).

However, the whole point here is also that there is no required knowledge to use a CAD, besides the manual. A couple of months is more than enough to be up and running with those instruments. That is, to have a quality output from a CAD, no previous knowledge is required besides the one concerning the CAD itself.

CFD is a totally different story, and the fact that unskilled people are now able to use such instruments (because of the robustness and simplicity which has been achieved by most of the tools)... well, doesn't mean it is the correct way to approach it. And how to produce a correct grid for a computation is just part of the required knowledge. Consider also that, mostly, each computation with different settings requires a different grid, making a database mostly superflous.

Finally, creating a good grid is always an iterative process and, obviously, a good grid for a given numerical method might not be good for another one.

Imagine introducing a single regulation for CFD computations to be admitted in a design: prove that the results are in the accuracy range of the chosen numerical method. Most of the people doing CFD today wouldn't even understand what this sentence means.

Unfortunately, CFD is perceived as a click and go possibility only because the field is lacking regulations, due to its very nature, and the difficulty in experimental testing. But, i ensure you that doing CFD in general (say, HVAC for parkings) and predicting correctly an airplane drag coefficient (maybe in landing configuration) or its acoustic signature are two different jobs... by orders of magnitude.

Fortunately for the formers, today there are several meshing tools which allow easy and fast grid generation. But, pretending that this is the whole CFD story? Far from true.

[Wandadars]

Thank you for your insight Sbaffini. I'm relieved that there is a database of validation cases. That was one of the major sources of my consternation. The community needs to have some agreed upon set of cases that are easily obtainable and in a format that is general. Standardization in an area like validation cases would be excellent. Thank you for the links to the resources that you provided.

[sbaffini]

Probably they are not exactly what you expect, but for some of them there are also the actual cases with their computed solutions (typically NASA).