Anyone interested in sorting out Openfoam for the masses?
 

Hey Eco-geeks, I'm hoping to gauge the level of available person-power for openfoam (open source flow dynamics) and possibly some diy laser/camera stuff to build models with so that folks *might* be able to scan their existing cars in and do some virtual mods and determine the effects. As well as enabling folks to actually simulate things like vortex generators instead of speculating endlessly. If there's interest (as in people willing and able to help, and willing to work on a convergent solution so that it IS more comprehendable to a larger audience, not the "I think you should test all my ideas for me" variety of help) then speak up.

There is a huge pile of stuff to sort through, but the benefits to modders could be substantial if it can be made coherent.

Here is an example of laser scanning, so the concept works, but we need something open, and high resolution. There needs to be some more brainstorming on the acquisition side, as well as how to modify a model, but in the end it needs to work with openfoam (and be open source).

http://ecomodder.com/forum/showthrea...ing-10011.html

Also there is waaay to much stuff to sort through without adding divergent platform complications or various branches and etc. So contributors would be asked to focus on one platform for starters, i.e. ubuntu (I have to use virtualbox FYI). And we should agree on what version of what to sort out first, and yadda yadda to keep tail chasing and self inflicted problems to a minimum.

There likely aren't enough people to diverge AND focus on helping the larger audience, but it should still be somewhat democratic among the contributors (keeping the larger audience, the target end users, in mind) as to how we sort out the specific versions of the tooling.

This isn't going to be a big creative effort, this is going to be an exercise in cutting out all the chaff and a lot of testing and some repackaging and documenting and creating examples and lots of explaining things in the forum and perhaps the wiki and whatever else I am forgetting.

C++/unix/java would be most helpful, need some aerodynamic gurus to validate the results too and probably explain a few things, a few victims to try to make sense out of the "dumbed down" version, etc. The hope is by limiting the options/parameters to a few use cases specific to land vehicles that we can put a simpler front end on it without a paralyzing array of options.

Thoughts?

I've got no idea about all the tech stuff but if it's possible to convert and test an Alias model I'm all over this. Good luck :)

As far as I know, the test model has to be a solid model. I looked into using OpenFOAM for my CarBEN EV, but I used SketchUp, which is anything but solid. (Though the latest version can make solid components, I don't think that "organic" shapes can be made solid.)

I haven't used it a long time but Rhino was good at organic shapes and it could produce solids. Parametric solid modeling is what you would want to use to change and modify subtle aerodynamic shapes.

bump

I have played with it, However there are several things about computer simulations you need to understand. To simulate things like vortex generators and such you will need a wind tunnel to validate your assumptions about boundry layer thickness and air velocities to get any useful data.

CFD tool can provide insight but there are pitfalls. With out verification of your modeling you are likely to end up pursuing a wrong solution.

hmm, you are saying openfoam is not able to determine the boundries of the boundry layer or air velocities?

It can predict it provided you have modeled every thing right. What I'm saying is things like how smooth the surface is and gaps in body panels on the car are going to have some effect on the air flow that you have to quantify for the model.

Another thing that is very very important is getting a good mesh of calculation points. Too many points and your simulation will take too much memory and time to run. Too few and you won't get valid results. As I understand it there are formulas that can give you a starting point, but knowing this is something that comes with experience and looking at results.

There are ways to go about this to get a good approximation. One method is to model a test case that someone has gone through and tested thoroughly in a wind tunnel and start tweaking parameters in the simulation until you get very similar results. Then you can start testing your model. A common test case for car simulations is the Ahmed Bluff body.

start reading through this thread on this forum

Automotive test case -- CFD Online Discussion Forums

Then get a copy of openfoam running and look at the motorcycle test case and learn how to view the data with paraview.

Sadly my free time is not predictable so I don't have time to learn it to the degree required.

you can also chat with someone on the internet relay chat (irc) on

irc://freenode/openfoam-irc

be patient the people there are from all over the world and in different time zones

Edit: I was told this would be a good link for someone to look at http://www.icaen.uiowa.edu/~me_160/L...medcarrodi.pdf

Thanks for the comments/pointers conn. I envisioned this effort would include sorting out how to scan your car accurately and afford-ably, and come with recommendations and/or tools for making the process more foolproof (possibly a warning for a certain rate of surface change for a given mesh resolution for starters). A very large effort, needs many volunteers.

Ok, so last bump here, fair warning. Or I'll put it to rest for now.

When I played around with openfoam I used gmsh to make models with. Gmsh: a three-dimensional finite element mesh generator with built-in pre- and post-processing facilities

It will allow you to import STL files and mesh/re-mesh them as well as refine the mesh. It also allows you to link to other meshing programs. It also has some post processing capabilities.

for example

http://geuz.org/gmsh/gallery/f16_stream.jpg

http://geuz.org/gmsh/gallery/f18_stream.jpg

http://geuz.org/gmsh/gallery/screenshot2.gif

As far as modeling a car you can go and get a 3ds file off the net of your car that someone else has made. Then you can import it into blender and export it out as an STL file. Then you can import it into gmsh to manipulate.

Be for warned that doing this stuff can take a lot of CPU time and memory.