I wonder how I can start to make a code commercial, what are the procedure for this?

Any comment will be helpful.

Junseok Kim

Try to sell it. If people buy it, you can consider it a "commercial" code.

Could you more specific, for example, to make user friendly interface, what kind of program should I use, about registering, etc.

Junseok

I guess that depends on your programming skill, the capability of your code, your target customer, how much time and effort you are planning to invest, etc., etc. I am not sure what you have in mind? Do you plan to just sell a flow solver routine that someone can implement into an existing package? Or, do you intend to sell an entire application, including pre- and post-processor, where the source code is not accessible to the user?

If you would like to have a sample of a "commercial" code startup, then please take a tour through my web site and download my program MicroTunnel,

www.cfd4pc.com or www.microcfd.com

I started this as a hobby while still in school and gave it away for free the first year. But then I decided to charge people $20, then $30, and recently $90 per license, after making some upgrades. Educational institutions (my primary customer) still get a 3 for 1 discount.

Great site!

That will be a good step for me, Thank you~

Thank you! - And you are welcome.

Feel free to email me any time later, if you have more questions.

Junseok,

Surprising question given that Prof. Patankar is out there at Minnesota. I also remember some one mentioning that the people who started Fluent were/was at U. Minn for some time as grad. students or post-doc.

I guess the first thing you need in an unstructured, implicit flow solver. PISO, I believe is the preferred integration scheme these days. With increasing memory, you could perhaps go to fully coupled Newtons solver (although it could become a mess for multi-species chemistry problems). If you are good salesman (or you can hire someone), I am sure you can begin first by demonstrating the robustness of an unstructured, implicit flow solver for some benchmark problems (laminar to begin with) at various Mach numbers with all possible types of flow boundaries. The physical models (turbulence, chemistry etc.) come next. Multi-phase flow needs physical models but the need to deal with density interfaces should be built in while designing the solver.

I believe that the numerics in the commercial code are getting better by the day but still you can not write a general purpose code and compete with research codes which have been tailored to specific problems and whose formally order of accuracy can be made arbitrarily high. Therefore, I think it is important to get some result with a commercial code not necessarily the accurate. One can play with the grid (quality and refinement) to increase the accuracy later.

Please note that what is written here is based on my opinions. I do not use commercial codes and I do not claim to be an expert on how they work. I however encounter people who use these codes fairly often. This is what I would look for in a commercial code if I were to buy one.

Good luck.

People will buy the code if it matches their needs. Find out what their needs are first then supply appropriate technology. If you try doing it the other way round, mathcing a technology to an appropriate need , you'll commercially fail.

Fred.

The problem is that everybody's needs are different, and if you want to develop a general purpose flow solver like the one described by Kalyan - AND do all the work yourself - you are in for quite a formidable task.

Also, keep in mind that the flow solver is only one small building block in a CFD application. Far more important is the user interface for the preprocessor. This is where the real ingenuity comes in, because in a general purpose code you cannot anticipate the user's problem setup. The interface has to guide the user along an invisible path to ensure that he stays within the limitation of the software and that the problem setup is physically sound as well.

OK, so general or specific. That seems to be the choice. In a way general purpose is better defined (everything is more clearly defined than something).

Axel, your tunnel code is well suited for acadmic purposes, a good tailored example. Did you start with the applicarion or the technology (or was the technology inherent in the application)?

Fred.

Developing a general purpose CFD code is indeed a formidable task and is not a one person job. You have to make a living while developing a code. I guess the easier way is to start a company that does some consulting and SBIR type work and get projects to do research. Start modestly with a research code and over time, you can make it general purpose by improving the integration, robustness, grid handling and routines for physical modeling.

This process isn't easy either since the existing companies which do SBIR type work will let you work on their code but you can not take it with you when you leave no matter how much you have contributed to it. The other option is to start a new company and unless you have the backing or direct involvement of a well respected CFDir (a "guru" in the CFD-ONLINE terminology), it will not take off.

Fred's idea isn't bad either. I had visited a startup in Detriot which deals specifically with flow/heat transfer in automotive systems and I heard recently that they are doing very well. At the time, they were planning on moving into electronic cooling and electronic manufacturing processes. The people who work in this company are experts in heat transfer and can deal with clients at an engineering level. They are very design-oriented. Most CFD code developers on the other hand are analysis-oriented and do not even know what units the industry folk use.

About MicroTunnel:

I had the 'technology' long before I started the application. My Master's thesis was a CFD study of transient supersonic jet exhaust. The jet was 'cold' (non-reacting), because I was more interested in its unsteady development from the explosive start, up to the point in time where the jet is fully developed and the familiar chevron pattern sets in. The jet was 2-D, and so a uniform Cartesian mesh was well suited. Van Leer's flux vector split technique was one of the algorithms I had acquired during my M.S. course work. It was easy to program, computationally inexpensive, very robust, and thus it seemed well suited at the time. - (Years later, I replaced the FVS algorithm with a TVD scheme and got much crisper results, at a larger expense in CPU time and memory though; the current version still runs FVS because it is faster).

As a Ph.D. student I had to take another graduate level CFD course, in which everyone had to choose his own term project. Since we had other computer projects due every two weeks, I decided to dig out the old code (instead of starting from scratch on something new) and just build on it. It was at that time, when I first had the 'crazy' idea to put obstacles into a uniform flow field (after eliminating the jet), which were simply cut out of the Cartesian mesh, such that inclined surfaces appeared as small steps. Somewhat crude, but I thought it was an interesting concept. I thought maybe one day computer power would be sufficient such that the mesh could be refined to a level where the jagged surfaces would appear 'smooth'.

Well, computers have gotten pretty powerful, and a few months ago I decided to write the FVS flow solver in assembly language such that I can run an unsteady test (from impulsive start) on an 800x600 mesh within a few hours. But the secondary wavelets created by the jagged surfaces still have not disappeared, although each flow cell is represented by only one pixel on the display. For blunt body analysis this is not a problem, but for slender bodies, e.g. airfoils, these secondary surface effects are clearly visible.

Currently, I am working on the next upgrade, which will eliminate the jagged surface problem. I have just implemented a cut-cell interpolation algorithm such that the flow is indeed smooth over any (originally) curved surface. The next version of MicroTunnel should be available within a few weeks, once I have everything thoroughly tested.

So in summary, Fred, I would say that some of the technology was there from the beginning, while other things evolved over time. I never really tailored it to any specific customer, I simply followed my own impulse. People in academics just turned out to be the ones most interested, although over the past 2.8 years (I went online in January of 1999) I had a rather wide variety of people from all backgrounds contact me. Some people were not even remotely aerospace related, such as interior designers who were just interested in the 'artwork'.

I don't know much about the word commercial, but if your food makes people sick or your code has a virus in it, then you will have hard time to deal with it. And if the content of the product is not consistent with the descriptions or ads, then the consumers can also complain about it. Even in a free forum, the person who makes the suggestion is responsible for the content of his comments. In the market, there are basic rules to follow. Use your common sense.