RANS modelling had its days and now LES is taking over. CFX is not a very good software for serious LES work. I am not so sure about Fluent. Considering the cost and limited features that CFX and Fluent offer, is it still worth to work on such softwares?

On the other hand OpenFOAM offers much more than CFX for both RANS and LES modelling and its absolutely free of charge. OpenFOAM results have proven to be much more accuarte than other commerical codes. When OpenFOAM is the future then why work on CFX and Fluent. Why are you guys still using CFX and Fluent and have you considered switching to OpenFOAM? Comment please

Dear Marco,

You got to be kidding!! right?

OpenFoam with all the good stuff it might have is not even closed to be a full featured product for industrial use. It cannot be compared with any of the 3 major CFD codes: FLUENT, CFX or STAR-CD. Not even to smaller codes such a CFD-ACE, CFD-Design, etc.

You want to tinker with the code, fine. That is your choice to enjoy. Other people have real targets to obtain a solution on a given problem and do not have the time to play with someone else's code.

Have a nice weekend.

Opaque

Hi opaque

I dont know what made you say that OpenFOAM is not ready for industrial use. As far as i know most of the universities and research organistaions are already switching to OpenFOAM due to its accuracy and reliability. OpenFOAM is an open source code and we can amend what we like in it even build our own solvers. I consider it advantage not disadvantage!

OpenFOAM is not some one else's code, its everyone's code. Any one can install it for free and it will give you better results than any other commercial code. I will post here comparative studies conducted between OpenFoam and other codes, then you can make your mind up.

You too have a nice week end opaque.

Hello Marco,

What can you say about the order of accuracy of the solver ? I'm not sure but I understand that the discretization in OpenFoam is only second order ... in Fluent 6.3 you could use third order.

I'm interested in a fair comparation between Fluent and OpenFoam.

Dear Opaque,

It is quite clear that you haven't even tried running OpenFOAM, let alone run some serious comparative tests. My company supports a number of commercial clients switching their CFD from CFX, Fluent and STAR-CD to OpenFOAM and they are most certainly not kidding.

I have worked both at CD-adapco and Fluent during my career and know my CFD pretty thoroughly, so the statements like "real targets to obtain a solution on a given problem" and "playing with someone else's code" are way off the mark. Why don't you join us at the Third OpenFOAM Workship in Milan in July and come see for yourself what the fuss is about.

Sincerely,

Hrvoje Jasak, OpenFOAM developer

The very first starting point is about LES for industrial applications. Sorry, this is not the case for 99.9% of the industrial requirements on restitution time.

The second point is about LES in itself. LES does not consist barely in putting a sub-grid scale model. If a significant part of the inertial scales of the turbulence spectra is not really solved based on a filtered approach for the smallest scales, please do not call it LES. You can call it whatever you want, but LES with dissipative and dispersive schemes, on URANS meshes with URANS times steps, just using a sub-grid scale model instead of baseline RANS one is definitely not a LES to me. Some may smile, but I see more and more industrial work of the kind.

Your statement seems to indicate that there are people who are shifting from CFX fluent etc to openFOAM for the reasons that openFOAM provides something that these softwares do not provide. Which is entirely wrong. As you said you have worked both as fluent and at starCD then you know very well that these commercial software provide a lot more features that openFOAM do not provide.

Of course this all costs. And people do pay for features.

If there are people who are switching from CFX and Fluent to openFOAM the reason certainly is not that openFOAM is superior. It may be anything but not this.

Hi, Hrvoje,

OpenFoam is indeed powerful and goes to outperform others.

But just a aquestion:

Since you be consultant of some commercial company like Fluent, you know more about their internal structure, based on my understanding from Fluent (its UDF capability) it should be written in C and is not object-oriented, i guess to keeping performance, So what about performance of OpenFoam in contrast with Fluent on a same problem?

Two suggections:

1) as you see most concern of CFD users are pre-processing and particularly grid generation, I suggest to add a octree cartesian cut cell grid generator in OpenFoam (is fully automatic for health CAD file) and adding a class to support such grid (is quit simple, each cartesain cell, either formal or cutted is a polyhedron). Whit this feature a great degree of automation is added and you could draws attention of cartesain (immersed boundary) too.

(Of course such grid is not very suitable for viscus flow, but cover a wide range of applications)

2) portability of OpenFoam is still issue, if you enquiry about some end-CFD users a wide variaty of them use windows as promary OS, while OF does not compile under native windows compiler, i guess under cygwin is possible (not sure), but this is not a clean way and also, under cygwin performance is decreased (anout 40 % in CFD codes, my experience)

What you think?

-John.

Babu i agree with you URANS is not LES. As the speed of computers get faster there is definite shift from traditional RANS to LES. My point was that CFX as it stands today is not very good for serious LES work.

John it does not cost you anything to make a shift from Windows to Linux. Again, linux open source OS is better than windows it terms of performance. Then why not make a shift to Linux and OpenFOAM and destroy these money making monsters. Comparative studies between OpenFOAM and other commercial codes will follow soon.

marco

> My point was that CFX as it stands today is not very good for serious LES work

Note that LES is state-of-the-art and it is too soon to add LES as a feature in commersial pakeges (commercial packeges use mature and reliable models, their users do not know state-of-the-art by default). Also power of current computers are beyound to perform a LES (considering real scales) simulation of industarial application. Else adding LES model to a commercial pakceges is not a difficult task. e.g. you could easily add explicit SMAGORINSKY filter by a UFD or such tool to any laminar solver.

But, I think It is a silly comparision if someone try to compare a commersial packages with an open-sources, because we look for open-source to do a research on cutting edge which is not possible for commercial ones. So OpenFoam has its specific position without attention to its limitation in contrast.

> John it does not cost you anything to make a shift from Windows to Linux.

so you do not understand what i imply !

> Again, linux open source OS is better than windows it terms of performance

are you sure? googling on this issue does not give a sharp judge !

PS: do you modify linux's kernel sofar, that talk about its open-source benefit? do you have ability of this job?

Dear Dr. Jasak,

I just could not believe what Marco said, but it seems those supporting/developing OpenFoam are reading beyond what is written.

I said: "OpenFoam with all the good stuff it might have is not even closed to be a full featured product for industrial use". Is it? let us review it:

a - Geometry enviroment: I do not think so.

b - CAD import: what CAD formats does it support? None.

c - Meshing in OpenFoam: it does not qualify to be compared to any decent commercial product in the CFD, FEA or even plain meshing market. You read the messages in the OF forum and cfd-online, and even OF supporters are asking for decent meshing of complex geometries.

d - Physics pre-processing: can OF be used in a modern design environment along those like CFX, FLUENT, CCM, CFD-Design, etc? I do not think so.

e - Post-processing: it does not even has its own, but relies on Paraview. Can it do quantitative post-processing for common modeling applications, simulation reports?

f - Physical models: Yes, there are a bunch of them in the source packages, but I have not seen any resembling:

- Radiation: surface to surface modeling, discrete ordinates, or Monte Carlo for starters. What about non-gray media modeling? None to be seen that compares to either FLUENT, CFX, or STAR-CD

- Combustion: Yes, it has some simple mixture fraction based models, but where are flamelets, ISAT, EDC, of ECFM?

- Turbulence: that is probably the one with several implementations; however, not even close to the pletora of models you will find on commercial codes. LES is not state of the art as Marco called. It is a fashionable model just because the computers are faster these days. The concept is more than 40 years old, and it is just receiving more attention lately.

- MHD: no comment. You know that the implementation there is barely a try-out of some good ideas. But, it will require considerable validation to be used by experts in the area.

- Stress modeling: well, I leave that for the FEA guys to evaluate.

Should I continue? Was I way off the mark? No, I was exactly on the mark. CFX, FLUENT, STAR-CD are not just a solver, they are professional tools that try to cover the whole process (geom, cad, mesh, pre, solver, post) to address more than just a limited set of problems.

I never even touched the issue of the core solver, numerics, speed, robustness, etc.

Now to your response,

I skimmed through your thesis once, read some of your papers (as well as Weller's). I know about your expertise in the area. I have no basis to question it. However, I had higher expectations of the source packages after reading the seminal paper :

Weller, H.G.; Tabor G.; Jasak, H. and Fureby, C.: A Tensorial Approach to CFD using Object Orientated Techniques, Computers in Physics, 1998 v 12 n 6, pp 620 - 631

- Have I run OF? No need. I have CFX and FLUENT at my disposal.

- Do I know what the fuss is about? Yes I do. Good papers, good simulations, a lot of color too. I can also get a Wii, or PS3, or XBOX-360 to play. The "target" I refer to is to finish the problem at hand: modeling a device and obtain enough physical insight to make improvements. Improving the modeling tool (regardless of how good it is) is an expensive distraction for some companies.

- You said "My company supports a number of commercial clients switching their CFD from CFX, Fluent and STAR-CD to OpenFOAM", and later added "I have worked both at CD-adapco and Fluent during my career ..." Hopefully, the time frame while you supported your clients to migrate away from FLUENT /CD-Adapco never overlapped with the time frame you worked at FLUENT or CD-Adapco. That is called a conflict of interest in professional ethics.

Hopefully, my view is clear this time.

Sincerely,

Opaque

Sorry, I do sign personally. However, those in the forum already know I work for a commercial CFD vendor.

Opaque,

I think that comparison is not feasible as others say.

But most of ur critsism is on pre/post process,

Post process: is not very important!, when a top graphic team (kitware) develop an open post-processor for ours, why we look for own post-processor. We just should costomize its.

Pre-process: I do not think your mentioned packages has a good pre-processor, e.g. Gambit of fluent has a very weack CAD media and also its grid generator has its own flaw. There are a veriaty of specific preprocessors that we could rely on them: GID, Patran.

Processor: althouth commercial vendors claim some features but they have their own limitations, particularly for multi-physics simulations, e.g. in my experience with Fluent 6.0 its VOF option (free surface flow) did not couple with phase change option (to simulate casting process). It seems that handling multi-physics simulation is an outstanding feature of OpenFoam.

What about optimal control and design optimization which is one goal of CFD analysis. Do your mentioned commercail packages have such option, i do not think, at least for Fluent, but for an opensource we could implement its.

The most balanced post so far, it does not mean That I agree with all your comments, but the one that is 100% true is that "CFX, FLUENT, STAR-CD are not just a solver, they are professional tools that try to cover the whole process (geom, cad, mesh, pre, solver, post)" That is it. As for the rest of your comments, just one thing to mention, OpenFoam source code is available and a qualified user can add whatever he believes is missing.

Post process: is not very important!, when a top graphic team (kitware) develop an open post-processor for ours, why we look for own post-processor. We just should costomize its.

post process and preprocess are very important and integral part of analysis. As far as paraview is concerned (this is my personal view) that it is just a joke on the name of post processing. On my windows machine that has 3gb plus RAM I could not read a results file with say 1 million cells in it. On the same machine I could easily read 6 million cell's results on fieldview. No problems. Paraview is just meaningless for serious work. I find it funny that you find gambit very bad but very happy with paraview.

Gambit of fluent has a very weack CAD media and also its grid generator has its own flaw. what serious flaws you are talking about in gambit. CAD handling was bad, and it has been improved. In my more than 5 years of use of Gambit I never had any problems due to it though. As far as meshing goes, gambit is one beautiful. It is difficult to use for people who do not understand how to work with it. Those who understand like me, its their first choice.

Now lets take solver,about which you seems to be so sure

Fluent provides: pressure based segregated solver pressure based coupled solver density based coupled solver (these are major types of solvers we know)

Further OP and Fluent both provide algebraic multigrid. But where is Full multigrid option in OP. And there are many many things.

Really comparing OP to Fluent is just insult to a full fledged solver. (I am not adding the stablity and speed of solution on Fluent).

I think OPenFoam can only focus itself on academic purpose because it is open. It is never possible for openfoam to compete itself with commercial packages. If you do not agree with me, please try to check the following items in both openmfoam and commercial packages:

1)easier to use

2)easier to analyze simulation outputs

3)portable

4)have an obvious improvement in accuracies

5)much faster?

6) have much priority in handling industrial scale calculations?

7) have more physical models?

8)have some special models which others do not have

Unfortunately,in all above items I can not see openfoam has a positive position. I found that most applications of openfoam had been mainly conducted by some academics and the developers themselves. No big company want to buy a people to learn and use a complicated code. This is a dangerous thing as the the use of openfoam requires too much experience and time, and only works well in Linux or the like. Looking at our world, how many companies which are using Linux based OS in their workshop are there? More importantly when we operate a company, we hope our things are standard and can be quickly controlled by new guys. Or we have to spend extra money to keep the positions for experts. So the use of openfoam may cause more money in a long-term view.

On the other hand, at this stage if some software says it can solve single phase fluid flow in complex geometries or the like with good accuracies, it sounds really ridiculous. It is because most commercial packages can do it. These are the basic functions of CFD. BTW: When Openfoam users claim that Openfoam has advantages over the existing commercial codes in some aspects, will you please to tell me the reason and which advancing modes or numerical techniques make this work. Now let's look at one of the biggest applications of CFD: the so-called process and chemical industries. Here the multiphsics and multiphase flows are our targets. OpenmFoam seems like a package mainly applicable to these places. However, Openfoam is really poor at this point. It have no new models in handling multiphase flow and combustions etc. the physical models available can be ignored compared to Fluents etc.

At last I want to say, Openmfom has made a great contribution to push CFD package to walk toward correct direction. We need to elaborate our simulations by introducing some new modes. The "black" commercial packages make this quite difficult, even a simple modification. Even worse, some packages cheat us in order to let results looks good. Openmfoam has no such problem. If the developers try to make it easier for new guys, it will be very good to the whole community of CFD. Now the problem if the developers work in this way, who are willing to support them?

Cheers,

Chen

I do not like elongate this disscusion, becasue i basically belive it is not feasible to compare opensource with commersial package,

> As far as paraview is concerned (this is my personal view) that it is just a joke on the name of post processing.

silly sound, so you do not have idea about what you say

i think your comparision be not equivalent or you do not know how to use paraview, how to write data and manipulate them (of course PV is not user friendly) i have used it on my PC with 2.5 G for more than 5 M unstrutured grid without any memory problem, after suitable filtering its rendering speed is comparable with Tecplot 10 (is fairly fast)

Of course you do not want to elongate discussion because you have nothing to support to what you say that openFOAM and Fluent should be compared as solver and I tell you that openFOAM only provides fraction of what Fluent provides.

You did not come and tell me that OpenFOAM has Full Multigrid Algorithm for difficult problems. You did not tell me that OP can give me coupled solvers. You did not provide any proof that OP converges faster than Fluent and is more accurate.

Fluent and other commercial solvers provide much more features, this is a fact. Deal with it.

Of course you do not want to elongate the discussion. I can understand.

Now as far as Paraview is concerned, last time I checked it had lots of issues with Polyhedral handling and it is well documented.

I do not know how you would assume that I would keep banging my head with something like Paraview when I can do my work with Fluent's post processor. I gave up on it long back and that time it did not read 1 million cells. But to check the things again I will again download and give it a try. And yes, I would use ensight file written by Fluent. Lets see.

Marco, let me give you a data point here. I am a single-person consultancy. Every time I use commercial software, I have to pay for it out of my own pocket, it has a direct impact on how profitable my business is. I have also, sporadically, over the last few years, experimented with OpenFOAM.

So here is what I do. For commercial work, I use commercial codes. Every hour I spend experimenting with OpenFOAM is time I could be using to generate revenue. OpenFOAM is a fine piece of work, but when you have to answer somebody's design question quickly, you don't have time to mess about. There have been times when I could perhaps have used OpenFOAM's solvers, but I still needed commercial tools for meshing. Likewise, Paraview is a wonderful piece of software, and I use it nearly every day, but there are things (such as quantitative post-processing) that standard Paraview cannot do. Sure, with some user programming, one could perhaps extract the necessary numbers with Paraview, but my client does not pay me to sit and figure out how Paraview works.

I am very keen to see more use being made of OpenFOAM and other open-source codes, but to even suggest that users should ditch their current codes overnight in favour of OpenFOAM is just amazingly naive. When you start dealing with real-world engineering design problems, one of the things you really need is good documentation about the code and its features. The detail documentation for OpenFOAM just isn't there. This is not a criticism of OpenFOAM, after all, writing documentation is a huge expense, and nobody is paying the code developers to do that.

I have nothing but admiration for the OpenFOAM guys, but outside of academia, the license cost savings rapidly disappear when you need to do real work in a hurry.

I thought opemfoam was free up to a point (limited use) and then it was charged. Isn't that the case?

No thats definitely wrong. OpenFOAM is open source software.