What would everyone think of an open source free development project ? I haven't seen any mention of this yet, though I might really have missed it if it has already been posted.

I really would love something like this.

if someone's interested, please contact me.

Viet

Dear Sir,

I am interested in this project. However, I do not understand your meaning fully. Would you please tell me about your thought in detail?

Truely your,

Lijun LIU

Hi,

I am interested by your project. I may contribute.

Farid++

To find more information on people who have discussed and worked on similar projects, go to the archive of old messages and search in page for Public CFD Code.

These links and the threads that follow may help you gain more information.

Public Domain CFD Project

More Public Domain CFD Project

Public CFD Finite Volume Project

Hi,

as was said before, there was a discussion on this topic.

I and also my colleague were hoping to contribute, but we had to stop on that, as my bosses were against such an project for different reasons, but I still think the idea is good. I suggested those who were interesed to look at the MOUSE project, you may have a look at it:

http://fire8.vug.uni-duisburg.de/MOUSE

I am sorry, that I was not able to do more about it.

Ciao

Heinz

(1). Even though commercial CFD codes have been available in the last ten , fifteen years, I have to say that the earlier versions of these codes were more or less for educational purposes. (2). On the other extreme, CFD has been heavily used in the defense industries in the high tech research fields. (3). For this reason, the project like this can only be participated by those who are working in the school environment, where free exchange of information is essential for their survival. (4). But sometimes, it is hard to get professors together because of the paper publication and related grant. I mean, it is very important for them to first publish the results under their names. (5). Under these constraints, I think, it is still possible to do something about it by defining the technology contents and the purpose of the final product first. (6). If the project's technology contents are limited to the knowledge available in the current textbooks level, then I think most people participated in the project would have no problem in finding the information to be used. (7). If the coding technology also uses standard Fortran language, then most participant would have minimum problems in reading it. (8). And if the problems selected ( CFD code is designed to solve problems) are typical benchmark test cases, then it will have a great impact on CFD user's attitude toward it. (9). I think, there is a possibility , if it is pursued along that direction. Something like a demo code which has every elements in it to solve some CFD benchmark problems. (10). So, for those who have free time and willing to work on such project with well defined and limited scope code development, I think, there is a possibility to make it work. (11). It is like working on a project to design CPU/286, while the real world is competing on CPU beyond P-III. (12). But if one can make the result of the project code useful for the newcomers to CFD field in a more professional way, then it will be a worth-while project.

Thanks for pointing me out those threads. I should have started with that search, of course ( This is very exciting. I never expected that much interest, since the CFD community is not that large after all. OK, now tell me, the thread dated back mid March : what's the status now ? have you all join MOUSE ? or KIVA ?

Viet

(1). In all CFD projects, the CFD source code is always available to the members of the project. (2). The problem is the unlimited access to the source code without any control of its consistency and reliability. (3). With many authors participated in writing the code, it is hard to guarantee a bug free code. I would say that it is much more important to exchange the ideas than to have a code listing. (4). In most existing cases of open source CFD projects, the code was originally written by one single author. The code was then made available for user modifications. Without proper documentation, the user modification of the code can easily run into trouble because of the mis-interpretation of the original code definitions. The code will quickly become useless. (5). So, for a short code, it is practical to make it open to anybody.( say a page long code) For a long code, it is better not to touch the source code at all.

I agree, there is a big risk of failure for this kind of project. However, with the right person in control and with good support from many users I think that a "CFD Linux" would be possible today - the interest these threads have generated here on the discussion forum clearly indicates that. The main question I think is if there is someone who is good enough and has enough spare time to lead this kind of project.

I think that the key to success is to have one good leader - you need a good basis to start from and this has to be written by one single individual, this person then has to take the main responsibility of deciding which user-additions that should go into the next release and thereby maintain overall controll of the code. Later, as more peoble get seriously involved, he can aslo delegate control to others within a core group - someone could be "solver responsible", someone "combustion responsible" and someone "turbulence model" responsible. But without the right person to start it and to keep it focused in the begging I think that there is a big risk of failure. I'll certainly help in any way I can in the little spare time I have if someone want to try to get this kind of project going.

(1). I don't think there is any risk involved in an open CFD project at all. (2). But I agree with you that " enough spare time" is very important. And in reality, it must be supported by his organization behind. In other word, there must be a sponsor organization. (3). The reason is very simple. In the CFD project, one needs to use a library, a computer, a network, and a printer maybe. (4). I think, it is possible to use the home environment, if he has a good computer systems. But then in this case, he is investing money and resources, and normally he is looking for some future rewards, like selling the code. (5). So, my conclusion is that the basic support of the time, the computer facilities are two fundamental issues. In a school environment, it would be possible to create a CFD project because the professor would be the project leader and the graduate students would be the contributors. And from there it can be expanded to the Internet world.(6). Actually, some commercial CFD codes were created this way. I think it is natural to ask the commerical CFD code developer to provide some limited version of the CFD codes as the base of the open CFD code. (7). I think, the fundamental concern is the issue of economy. Open CFD project still require financial support. So, the first step is really to look for financial resources for one to three years first. Once the financial support is available, the next step is to look for a project leader. This would be the practical solution. (8). In 80's, I have attended several PC user groups activities, including one which I created. Without support and financial resources, the best one can achieve is "discussion in the meeting about some new commercial software". (9). So, the cfd/forum is probably the most practical way to participate in CFD activities.

What you are describing is the classical way of developing CFD codes - a group of PhD students led by a professor. This is however not the way most public-domain large-scale internet-born software packages are developed. Look at Linux. I'm convinced that, as I think Wired wrote in the last issue, the fact that Linus lived in Finland and had to communicate with all his "helpers" via email etc. made this project such a success. If there had been a core group of developers that worked at the same place things might have developed differently - they would have looked at the code as theirs and other people would have been less prone to contribute time and effort to the project. In the end Linux might have gone commercial. Instead now Linus was alone up here in Scandinavia and he had to distibute responsibilities all across the world. This engages people and somehow makes people find time - you don't need financing for this kind of project. If it attracts interest and gets people involved they will work for free, whatever the boss says, belive me. And some will contriubte a lot more that a few posts in this discussion forum! No-one ever said that it was okay for me to start writing CFD Online, I just found the time and did it, it was fun.

(1). Yes, the example I gave has become a classical example. I would say it was in the pre-Internet era. (2). The last point first, last year, after a large European company acquired an American company for a couple of Billion dollars, the Internet use of the American company was monitored and engineer could no longer easily get on the Intenet when working in the company. To get on the Internet, the engineer has to justified that it is in the interest of the company to get on the Internet. Many connections to the Internet sites were disconnected. (3). Back to the basic issue of open CFD project. Assuming that the Linus project is a good representation of an open project on Internet, and we have to acknowledge that CFD is more international, and more important to mankind, then why is it difficult to get such project started on cfd-online? (4). I have not lived in Europe, and my knowledge about European people and countries are very limited. I would say that there is a fundamental difference between American and European in this area. If that is the case (people are willing to to work a lot for free when their boss say so), then I think, the future open CFD project will be started somewhere in Europe. (5). I think American value individual's contribution a lot. So, Innovation and reward are always linked together, especially for a person. (6). I think, between Linus and the future open CFD project, there is a basic difference between them. On the operating system side, the technology has been around for a long time for the main frame, the mini-computer, the supercomputer, the mini-supercomputer, the workstation, the PC. In other word, the technology is mature and there are tens of thousands of programmers in this field. So, we are not talking about any break-through at all. (actually, I don't like UNIX type systems at all because it is very hard to use. VI editor is a stone-age product, and it is still available and being used. really amazing.)(7). On the other hand, CFD technology is highly technical. It has been limited to some groups of people only. Even MIT is not known for its CFD work. From this point of view, the project development can be slow. If you have used a commercial code, you would have noticed that they normally cited the methods developed by well-known authors. In other word, the selection of the methods is limited. (8). There is also the dark side of the CFD technology, good 3-D CFD codes are normally under export control. That means the technology used in this open CFD project will be out of date from the begining. (9). So, if the financial issue is no problem, the next step would be to locate this ideal leader for the future open CFD project. Well, that is consistent with my thinking that a person is more important than a code. ( I would lay off a code first instead of a person.) (10). Based on what I have seen, many companies which have not invested in CFD engineer today will sure vanish in the next century.

We have the problem, that you cannot compare LINUX and CFD. There are a lot of people working for LINUX without having been at university. For an open cfd-code you need a lot of highly qualified people. I think, the only way to make such a code is to devide the tasks: one has to study the literature, one has to do the mathematical part and perhaps write a flux-diagramm and the third one has to write the code. And that has to be done for each chapter of the code. I don't think, that one person can do all the things.

(1). Yes, I agree with you. Open CFD is not Linux. (2). I think, we were talking about a leader of the project, like an editor of a book. The final code does have to be written in a consistent way. (3). I have no idea what will be in this open CFD code. (4). The normal procedures involved in CFD is something like : main_program, call geometry, call mesh_control, call mesh_generation, call display_mesh, call mesh_checkpoint, call boundary_condition, call initial_condition, call solver_control, call solver ( call solve_momentum_equations, call solve_turbulence_model, call solve_chemistry, call solve_energy_equation,...), call solver_checkpoint, call display_convergence, call display_flow_field ( call display_velocity_vector, call display_contour_plot, call display_particle_tracing,...), call display_checkpoint, call next_time. (5). I think, it's your project, so it's up to you to structure it. If there is something missing, you can always modify it.

I wouldn't take pre- and postprocessing at the first step into the project. But discuss some points for a better orientation: finite element or finite volume ? compressible (at the first step) or not ?. What could be a realistic goal for the first running case ? My problem is: this seems to be all really interessting and I really want to help, but I think, I do not know enough for this project and I have not enough time to "learn" all those things. But: never give up

(1).Everything starts with a problem definition, which must have a geometry and boundary conditions. (2). A useful problem would be the test cases commonly used in CFD validation. (3). The problem determines whether it is a compressible or an incompressible flow. ( if you use a solver which can handle both, then only one solver is needed. otherwise, it will be either incompressible or compressible.) (4). It doesn't matter whether the solver will be finite-difference, finite-volume or finite-element. You simply read in the geometry , boundary condition data and the mesh into the solver, the rest of the work is done in the solver. From the solver, there come the results. If you provide the finite-difference solver, you select the option-1. Otherwise, option-2, and option-3. (5). A solver is a solver, it doesn't have to be THE solver. (6). The project has to be problem-solving oriented. Otherwise, it will become a huge libraries, something like a commercial code. You can't say that you are going to develope a code to solve any CFD problem. You always have to define the problem first. (7). I am not going to write the code for you because I am busy trying to stay alive.