Dear all

I'am starting my PhD research which the subject is the simulation, whit a CFD code (Fluent), of the control sewer overflow structures (sructures into the sewer networks). So, I would know the state of the art of the CFD technology.

Regards

(1). We don't know what is exactly inside a commercial code. So, there is no way of telling the state of the art of a CFD code. (2). The CFD technology itself covers many different fields, such as, the geometry modeling, the mesh generation, the numerical formulations (finite-difference, finite-volume, finite-element,...), the solution procedures, the turbulence modeling, the graphic post processing. (3). There are many technical journals available in these fields. So, you should be able to find additional information in the library. (4). I would strongly suggest that you first define your problem, then try to derive the equations, and see whether you can understand it and solve your problem. (5). Using a commercial code takes only a few weeks to learn, and a few months to get a solution (assuming that it is a difficult one). A PhD is not required to run a commercial code.

You are dreaming too much to get a PhD. using commercial codes. Commercial codes have limited applications. Small schools for example like University of Nebraska or say University of Alabama may give you PhD. if you use commercial softwares. There are lot of PhD. degrees you can buy with money where you don't have to do any creative work. I agree with John, "We don't know what is exactly inside a commercial code. So, there is no way of telling the state of the art of a CFD code."

I don't think the gentlemen at the universities of Nebraska or Alabama would like you to characterise their institutions as 'small', actually these schools are large in size and in reputation

Usually graduate schools ranked in top 50 or top 60 are considered big schools (in USA) with good reputations. I just mentioned some schools name for showing some example, nothing more than that.

That's ridiculous! If you are a Civil engineering student studying sewer overflow structures, using anything other than a commerical code would be irresponsible. It's one thing if you are studying CFD for your thesis, but this fellow is *using* CFD.

No university would insist that he actually roll his own code, and any reputable CFD software vendor will completely describe the methods used in their code (which are almost always well-known from the research literature anyway). The value in commercial codes isn't coming from in-house secret developments as much as the total integration of many capabilities. So I don't see the 'black box' theory as a real issue. And to say that commercial codes are more specialized than research codes??? What are you smoking?

going on...I am aware that the ThermoFluids department at Stanford (is that reputable enough for you?) has PhD students using commercial codes for 2 purposes: (1) in support of their experimental work, and (2) using the "user-defined subroutines" extensions to develop a new model or something like that without having to write an entire code. To get to the state-of-the-art in CFD nowadays, you can't possibly write your own code unless you are going to stick with simple geometry.

In my opinion, if you ask enough CFD people, you'll very quickly get the picture that the days of rolling your own CFD code as a PhD project are very much long gone, and there never were such days for Civil engineering students.

1). We don't know what is exactly inside a commercial code. So, there is no way of telling the state of the art of a CFD code. (2). The CFD technology itself covers many different fields, such as, the geometry modeling, the mesh generation, the numerical formulations (finite-difference, finite-volume, finite-element,...), the solution procedures, the turbulence modeling, the graphic post processing. (3). There are many technical journals available in these fields. So, you should be able to find additional information in the library. (4). I would strongly suggest that you first define your problem, then try to derive the equations, and see whether you can understand it and solve your problem. (5). Using a commercial code takes only a few weeks to learn, and a few months to get a solution (assuming that it is a difficult one). A PhD is not required to run a commercial code.

I think so for you CFD is only set of tools for example as a set of pencils. But can you say more about your engineering problem?

And to say that commercial codes are more specialized than research codes??? What are you smoking?

Sorry, I don't smoke.

J'espère que vous pourriez lire ce message.

Je suis ingenieur civil et je connais tres bien mos sujet de thèse. A chaque fois que lis vos messages vous ne m'apportent rien. Il y a des gens comme Monsieur Chien qui m'ont aidé a mieux comprendre ce qui est l'etat de l'art de la tecnology CFD. Donc, merci quand même.

Vous devriez savoir que ce moyen de communication ce n'est pas pour insulter les autres et pour ceux qui me demandent qu'est ce que je fumme je vous repond: RIEN.

En tout cas pour le Monsieur qui me dit que son Université est trés reconnu (Stanford je crois), je lui dit que la technology CFD que je vais utiliser est pour valider un autre simulation dans un logiciel hydrodynamique de simulation de reseaux d'assainissement.

Je crois que vous drevriez savoir aussi que les structures de déversement (Deversoir d'orage en Francais et CSO en anglais) ont un rôle primordial dans la recherche en hydrologie urbaine. Donc, je crois que je me suis trompé de groupe de discusion car ici j'ai l'impression que existen seulement de savants mais en CFD seulement.

Merci a tous vous, mais vous ne m'apportent rien, sauf Monsieur Chien.

I think that you could know this message posted octobre 8, 1999.

My name is Juan Carlos. I am a civil engineer from National Autonomous University of Mexico (UNAM). In the Civil Department of the National Appliend Sciences Institute of Lyon (INSA-Lyon), France I was study a posgraduate course in which I developed like finaly work the research "Descrption and modelling of CSO models (frontal and lateral) into an HD (hydrodynamic sewer system software)". Now I am starting my PhD research at Civil Engineering Research Unit-Urban Hydrology at INSA-Lyon and where the principal subject is a construction, analysis and validation of hydraulics models of CSO into an HD software. This procedure will be calibrated and verified with a CFD softwere in order to establish the practical means for CSO's HD modelling.

Juan Carlos GARCIA-SALAS

Hi

I would definitely disagree with your statement concerning creative work and the use of a commercial CFD-code. I have spent a number of year investigating cooling tower performance experimentally. Currently I am using a commercial CFD-code to determine whether the major trends that I observed experimentally can be duplicated numerically and the results thus far have been very positive. The flexibility provided by numerical simulation has opened new avenues of thinking and a lot of creativity has gone into forging these avenues. I do not advocate the use of commercial CFD codes to do fundamnetal research prior to the necessary validating of numerical data against experimental results, but if you can show that the numerical results compare favourably with experimental data the commercial CFD-code becomes a powerful research tool.

I would agree that if your goal is to make a fundamnetal contribution to the CFD field then producing your own code is essential. From what I understand of Juan's message it would seem that he is not that interested in making a fundamnetal contribution to the field of CFD, but rather he hopes to use CFD as a tool in his research. I do not see how his PhD would be "bought" if he can show that his numerical results compare favourably with experimental data.

regards

Chris

Daer Mr. Chris

This is exactly, I am not interesed to make a contribution in the field of CFD research. I only hoped to know this technology because my PhD research requires, for validating and calibrating the CSO models into an HD sewer system software, the technology CFD and the experimental data from special structures into sewer networks, nobably in Lyon France.. and I say: My PhD research is in the field of URBAN DRAINAGE.

Cordially

Juan Carlos

Hi

My advice to you is to be very rigerous in validating your numerical results. It is very easy to end up in the situation where you don't know whether you are observing a physical phenomena or just some numerical anomaly. I have spent the better part of two years on the development and testing of my models before I started looking at the real topic of my PhD. What some of the other people have said is very true: you don't know what is inside the commercial codes so proceed very carefully and make sure you understand what you are doing. I spent a lot of time reading up on discretization schemes, turbulence models etc. I also tested there influence on my results. Commercial codes are deceptively easy to use and the results always look very pretty!!! But that says very little.

I wish you well for the task ahead!

Regards

Chris

(1). I like your comments. I think this is the right place to address the proper use of a commercial code. (2). Number one, you have spent a number of years investigating the problem experimentally. This is the key requirement. I am assuming that you understand your problem completely. (3). At this point, you are sure that your experimental investagation is reliable. (4). Number two, you are using the commercial code to check whether the numerically predicted major trend is consistent with the experimental results. So, basically, you are doing code validation. You don't trust the commercial code, you trust the experimental data. (5). From your comments, so far you are pleased with the commercial code results. And that is a good new for you. (6). At this point, you then make a positive comment on using a commercial code as a tool. (7). The fact is to reach this conclusion, you need to spend a number of years investigating your problem experimentally, and then validate the code against the test results. (8). I think, that is the right approach. But it will add a number of years to the PhD program doing experimental investigation first. (9). In turbomachinery, even the experimental results are not reliable in our case, not to mention the convergence of the commercial codes. (10). I like the idea that: a PhD student using a commercial code must demonstrate that he can obtain reliable experimental results first. This will demonstrate his real understanding of his problem. Once this is done, he can try to validate the commercial code against his test results. (11). Although this is a practical approach. it tends to make the person using the commercial code looks like a "second-class citizen". (12). I would hope that some days we can just go ahead and use the commercial code, without spending a number of years doing the testing first. This is probably the state of the art of CFD. (if we make the vendors of the code responsible for every bad results obtained by the users, then that goal should be realized earlier. The current picture is: my code is good, but don't tell me you are getting bad results, I have said clearly in the fine print that I am not responsible for your bad results, but I still say that my code is a good code. So, the only way a user can protect himself is to learn how to write his own code. There is no other alternatives. The PhD problem is an unknown, and if the commercial code is also an unknown, unknown + unknown is not = real answer. Unless, you have just redefined the + operator in a different way.) (13). To repeat my suggestion here: a commercial code is best suited for a person who knows how to write a code to solve his problem, or a person who is willing to spend a number of years doing experimental investigation first,followed by careful validation of the code against his results. Anyone else using the code is not protected against the bad results.

Your main work seems like experimental work in fluid dynamics which requires lot of expermentation. Comparing your results with numerical simulations to validate results seems auxiliary work. Probably you will survive but remember that "What some of the other people have said is very true: you don't know what is inside the commercial codes so proceed very carefully and make sure you understand what you are doing."

And I also wish you best of luck.

Can someone explain to me what they mean by the phrase 'CFD is a black box'. From my point of view, CFD can be a black box if one is a Fluid dynamist and knows very little about numerical analysis. It could also be a black box to some numerical analyst who knows very little fluid dynamics. This point is well put in the preface section of Ferziger's CFD text. He said that you have to be both a numerical analyst and a fluid dynamist to be a good CFD analyst. If one has this ideal background I do not see any reason why one cannot pick up a Fluent or CFX manual and get a very succint picture of what is happening in ones area of interes in these codes. Put differently, these codes are not black boxes, if I understand the term. They may have some constants that they use that they do not want someone else to know about but that does not make the code a black box. They may have some errors but so do the programs we write. Errors should not make them black boxes. As responsible users of CFD, we should tell other users that commercial CFD packages are easy to use but very difficult to use properly. Note the word properly.

We should be thankful to the academic community whose R&D efforts have helped CFD develop into a useful design tool. We should also be thankful to the 'commercial packages' for implement these good ideas. During the implementation of these ideas mistakes are made, we should help out by pointing out the mistakes to the developers, so they can be corrected. This way we help improve the code. Several years ago, I used to write code to solve specific problems. I almost always finished writing the code after the project was over. The reason being that it is difficult to debug and test a piece of code properly. Now, I let better programmers do the programming and I do my careful analysis. I hope I have not offended any one because that is not my goal.

Can someone explain to me what they mean by the phrase 'CFD is a black box'. From my point of view, CFD can be a black box if one is a Fluid dynamist and knows very little about numerical analysis. It could also be a black box to some numerical analyst who knows very little fluid dynamics. This point is well put in the preface section of Ferziger's CFD text. He said that you have to be both a numerical analyst and a fluid dynamist to be a good CFD analyst. If one has this ideal background I do not see any reason why one cannot pick up a Fluent or CFX manual and get a very succint picture of what is happening in ones area of interes in these codes. Put differently, these codes are not black boxes, if I understand the term. They may have some constants that they use that they do not want someone else to know about but that does not make the code a black box. They may have some errors but so do the programs we write. Errors should not make them black boxes. As responsible users of CFD, we should tell other users that commercial CFD packages are easy to use but very difficult to use properly. Note the word properly.

We should be thankful to the academic community whose R&D efforts have helped CFD develop into a useful design tool. We should also be thankful to the 'commercial packages' for implement these good ideas. During the implementation of these ideas mistakes are made, we should help out by pointing out the mistakes to the developers, so they can be corrected. This way we help improve the code. Several years ago, I used to write code to solve specific problems. I almost always finished writing the code after the project was over. The reason being that it is difficult to debug and test a piece of code properly. Now, I let better programmers do the programming and I do my careful analysis. I hope I have not offended any one because that is not my goal.

What language is this? It is certainly not english.

in my opinion, even if you can write your own code, you still have to spend enough time to do your experiment to validate your own CFD code. it's like you cooking coffee, the CFD is the sugar you add into your cup. anyway you must have a cup of coffee first, And, not all people like sweet coffee. maybe Juan's superviser like,