Hi everyone.

It is clear that there are a huge number of open source projects for all kind of different applications; most of them are incomplete are useless. However, once in a while, you can find great open source projects such as linux, mysql, openoffice, and eclipse. Benefits from these projects are extensive.

In a similar way, we can find a wide range of CFD open projects on the internet. Can we say that this proves that CFD open source is a good idea?

I believe that the CFD community could benefit from open source projects if these projects have well defined objectives and the total support from the CFD community. In that respect, I am interested in exploring the idea of creating a CFD open source project that will be beneficial for most of us. I would like to know how many of you are in favor or interested in this idea. What do you think?

Thanks a lot.

Dear JF,

Have look at OpenFoam? It is an open source CFD code.

Opaque

PS. I do not endorse/agree with their implementation details; however, it is active, it has a lot tools, and a lot of people to ask for help..

> In a similar way, we can find a wide range of CFD open projects on the
: internet. Can we say that this proves that CFD open source is a good idea?

How many of these open source projects are active?

> I am interested in exploring the idea of creating a CFD open source project
: that will be beneficial for most of us.

If open source is a good idea why aren't you helping an existing open source project instead of starting yet another one?

> What do you think?

I think a successful open source project needs to be supported and maintained and this requires people to spend a lot of time performing boring and uninteresting tasks. Enthusiasm to do this voluntarily soon wears off.

see this: http://www.lmm.jussieu.fr/~zaleski/zaleski.html

http://www.lmm.jussieu.fr/~zaleski/OpenCFD.html

The cfd community is already profiting fom open source. They are even making a profit out of it.

Where do you think the dualization for polyhedral meshes initially comes from?

Where do you think some of the Delaunay and advancing front triangulation routines comes from?

The Newton-Raphson solvers, some the post-processing mapping/interpolating tools, cantera chemical reaction solver, GPU flow solver etc...

They all come from open sources downloaded from the web, and inserted with a hammer into a commercial CFD code.

Then it might take a year or so to debug a small problem because the guy just downloaded the file and made the connection, it was not actually programmed by the guy.

Currently you may assemble and build a commercial cfd code a bit like a puzzle by downloading and connecting open-source elements.

Try it, you will see.

Sure!!!

The last I recall, most of the big FEA codes like ANSYS, ABAQUS, MARC, SAP, etc were written long before the Internet and open source movement came about..

For CFD, Phoenics/Fluent and others were also started before the popular open source movement..

Similar for other pieces of engineering software in the market place..

Wake up!!!

Opaque..

I think you're confusing the proliferation of ideas with Open Source Coding. Information (as long as it isn't Commerically Classified) is freely available in scientific journals, and many of the numerical models and algorithms that are widely used in CFD have come from them.

Coding these up into a coherent code is quite a different matter altogether. Groups of people have a hard enough time writing good CFD software when they hold regular face-to-face meetings, let alone trying to organise a CFD code based on e-mail traffic and forum posts. I personally would be astounded if any code that was built from downloaded elements gave anything approaching sensible simulation results.

new parts of the codes you are mentioning were taken from open source and plugged in afterwards and sold to you.

"I personally would be astounded if any code that was built from downloaded elements gave anything approaching sensible simulation results."

Have you tried some of the radiation or some chemical reaction models in cfd codes?

Nobody's astounded that they don't work

You an opaque are so naive on what is stuffed in there for marketing purpose only sometimes...

Oh no, I agree entirely, a lot of things are marketing-relevant only! What I meant was someone like us on this forum making their own code from various pieces of open source code and hoping it to give decent results....

Dear Blue,

I do not recall any open source code dealing with radiation (I assume you mean thermal radiation) that is relevant to commercial CFD. Most of the source code available is meant for 1D calculations and highly tuned for atmospheric, astrophysics, and plane slabs..

Most thermal radiation models in current commercial CFD include: 1st order spherical harmonics (P1), discrete transfer, discrete ordinates, finite volume radiation, monte carlo, or flux methods. Never seen any of these models in open source, have you?

If you know any link/pointer to thermal radiation open source, please share them with the forum.

You do not need to mention company names, but knowing which open source projects you are referring to will be of help for the community in general..Otherwise, your statement become baseless..

Opaque..

Hi all,

I love it when this forum gets a good topic like this. ;-)

There are some good points above but the devil is in the details.

Some parts of a "full featured commercial cfd code" can be gleaned from something that is "open source" some parts cannot.

Opaque.....of course a radiation guy...says no open source for "full featured" radiation. I would agree with that on the other hand I hear that Shaw's stuff was put in CFX-4 pretty much out of his thesis...you probably know better than I?? Would all radiation models just plug in .... obviously not as radiation is the most unusual model in a CFD code that i have ever seen......integral differential equations solved on a different (usually) grid!! So no none of that is going to just plug in!! I totally agree!

Blue.... you have a good point that certain points were plugged in and sold. A good example is the spray models in fluent which basically came out of KIVA. These do not need a lot of mods for a "full featured" CFD code if you already have the basic Lagrangian tracking working! When I say working i mean working in parallel, with most other features, etc. But in general No you can't pull together some pieces from the flee market and get "world class full featured cfd code"

Shit I should be in marketing that sounded good! Just kidding I think the marketing guys are the biggest part of the problem$

On the other hand....as a user I see the price of commercial cfd is too high and rising and the openness of the code, support, user input, etc decreasing.

Just my input.

Regards,

Bak_Flow

Dear Bak_Flow

From what I know, the radiation module in CFX-4.x was written from scratch in the very early 1990's. It includes the discrete transfer model (Shah's algorithm) and the Monte Carlo model as well.

One thing is using Shah's ideas, and another to reuse his code (that is what open source is, right?). I read Shah's thesis back in the late 80's, and from I recall his was for simple geometries (Cartesian mostly), and far from a "full featured" module..

Other CFD codes had closer ties (Imperial College) to his work.. He worked on the same building (perhaps in the same floor) where STAR-CD and PHOENICS were conceived.

Opaque..