[BrendaEM]

I wish OpenFoam would realize that in the real world, meshes aren't all made of neat diagonally sliced cubes.

I've even tried the Salome export script, and that's all I get is
*Number of severely non-orthogonal (> 70 degrees) face...

Well, it should still work, even if the volumes are useless slivers, right?

One something has rounded corners, which seems to generate a lot of them, OpenFoam all comes apart.

Yes, I am using Salome, but people are still having problems with snappyhexmesh. And if openFoam has it's own pet mesh conversion utility, will it turn its back on the rest of them? Are there any other open pre-processors that are being developed?

~

I wish there was an open source alternative to OpenFoam. Reading these forums, I cannot believe that most people have not had major problems with OpenFoam.

Why must we get out experience and documentation from other users?
In the large sense, do you know how woefully inefficient that is, to have the blind leading the blind?

Just copy xx case, is not as good as taming the syntax or documenting the beast.

(Feel better now : )

[kmooney]

snappyHexMesh is a pretty capable mesher. I would take some extra time and figure it out so you can get away from using tet meshes if at all possible. Any finite volume code will be less stable on those elements. There are a few free GUI's out there that can help with meshing setup if you have STL based geometries too.

A good way to start might be to setup a few cases with a GUI, take a look at the snappyHexMeshDict it produces, and reverse engineer from there.

[alexeym]

Hi,

Originally Posted by BrendaEM

I've even tried the Salome export script, and that's all I get is
*Number of severely non-orthogonal (> 70 degrees) face...

70 is just a constant in checkMesh utility. In reality it's OK to have meshes with higher non-orthogonality.

I wish there was an open source alternative to OpenFoam. Reading these forums, I cannot believe that most people have not had major problems with OpenFoam.

Rather interesting thought But who managed to close sources of OpenFOAM?

[elvis]

Hi,

I am not sure how do you convert your mesh from salome to OF.
Do you use ideasUnvToFoam,
or the script https://github.com/nicolasedh/salomeToOpenFOAM mentioned http://www.salome-platform.org/forum...3165#994489164 ?

ideasUnvToFoam does not convert everything
unfortunately http://pythonflu.wikidot.com/hybridflu is not actively developed

[BrendaEM]

kmooney, I appreciate your reply, but you validated my apprehension about favoring an OpenFoam internal mesher.

Meshing from a mesh is sub-optimal, because the user has to do twice the work to change it. I build my objects from nurbs surfaces, that's why i had been using stp/step. I can import my object, from stp, which is currently the defacto standard for surfaces(like it or not), and I want to mesh from that.

I am not aware of pre-processors that can use snappyHexMesh. At least with Salome there is some workflow. I can choose and label surfaces to give them boundary names. As a thought experiment, what percentage of the time how often do you use OpenFoam's data without Paraview to investigate it? And if we use a powerful tool for PostProcessing, then why not preprocessing?

[I am trying to a CFD solve on a replica of a real world object, in one case a NACA duct. With filleted corners it has 32 surfaces. Meshed so I have around 10 volumes high in the duct, so I can see the vortices, it has roughly 2,000,000 volumes. Not something I want to be doing in gedit.]

alexeym, speaking from experience OpenFoam will surely crash like a cheap airplane if you try to run a case with too many non-orthaganal volumes. There are several threads on this forum which support my claim. Additionally, decomposePar will fail under the same conditions. If I am understanding the problem correctly the problem is: OpenFoam just cannot deal with tetrahedron meshes that include volumes that have large aspect ratios.

elvis, I have done it both ways, the (your?) script does better, but things still work poorly. I finding it interesting that no one fixed ideasUnvToFoam? Should I even attempt to put in a bug report, when it's such common knowledge that it doesn't work well?

[alexeym]

Hi,

Originally Posted by BrendaEM

alexeym, speaking from experience OpenFoam will surely crash like a cheap airplane if you try to run a case with too many non-orthaganal volumes. There are several threads on this forum which support my claim. Additionally, decomposePar will fail under the same conditions. If I am understanding the problem correctly the problem is: OpenFoam just cannot deal with tetrahedron meshes that include volumes that have large aspect ratios.

Well, if we look at the forum there will be LOTS of posts where people can't run simulations on perfectly orthogonal meshes. I successfully ran simulations on meshes with 87 degree non-orthogonality, during OF workshop in Zagreb this year Prof. Jasak mentioned successful simulations on meshes with 115 degrees non-orthogonality. Surely if initial post was just to release steam - OK, bad OpenFOAM

[kmooney]

Originally Posted by BrendaEM

kmooney, I appreciate your reply, but you validated my apprehension about favoring an OpenFoam internal mesher.

Meshing from a mesh is sub-optimal, because the user has to do twice the work to change it. I build my objects from nurbs surfaces, that's why i had been using stp/step. I can import my object, from stp, which is currently the defacto standard for surfaces(like it or not), and I want to mesh from that.

I am not aware of pre-processors that can use snappyHexMesh. At least with Salome there is some workflow. I can choose and label surfaces to give them boundary names. As a thought experiment, what percentage of the time how often do you use OpenFoam's data without Paraview to investigate it? And if we use a powerful tool for PostProcessing, then why not preprocessing?

[I am trying to a CFD solve on a replica of a real world object, in one case a NACA duct. With filleted corners it has 32 surfaces. Meshed so I have around 10 volumes high in the duct, so I can see the vortices, it has roughly 2,000,000 volumes. Not something I want to be doing in gedit.]

alexeym, speaking from experience OpenFoam will surely crash like a cheap airplane if you try to run a case with too many non-orthaganal volumes. There are several threads on this forum which support my claim. Additionally, decomposePar will fail under the same conditions. If I am understanding the problem correctly the problem is: OpenFoam just cannot deal with tetrahedron meshes that include volumes that have large aspect ratios.

elvis, I have done it both ways, the (your?) script does better, but things still work poorly. I finding it interesting that no one fixed ideasUnvToFoam? Should I even attempt to put in a bug report, when it's such common knowledge that it doesn't work well?

Many people here are actively building their careers based on OpenFOAM CFD. I'll be honest that you're not going to get much support on this forum if this is how you plan on interacting with the community.

Put the effort into understanding the tools available and put the required effort into pre-processing and meshing your geometry. You're not going to convince anyone here that OpenFOAM is 'too hard' for basic cfd.

[BrendaEM]

Respectfully, the people who have built their carreers from OpenFoam will be resistant to any change, bad or good.

Those who make thier living from commercial CFD solutions will not want OpenFoam to be any easier to use, any more reliable, or any better documented.

No, I am sorry but I am not going to agree with your peer-pressured statement and say that OpenFoam is fine cannot be improved.

[BrendaEM]

I've learned that perhaps I might not be alone in my observations, suggestions for OpenFoam.

kmooney, Reenforcing an elitist attitude at the mere suggestion that a OpenFoam could be better and better documented I find quite interesting. Fortunately, up to now, I've been protected from academic peer-pressure.

OpenFoam is open source. Logically, I wouldn't be looking a gift-horse in the mouth, if it were carrying me well, now would I?

[philippose]

Hi there,
A Good evening to you.

I thought I would also put in my two cents (euro) worth in here too :-)

I have been using OpenFOAM for over 7 years now with completely real industrial cases in the field of Hydraulics (complex valves, piping, pump geometries, etc...etc...etc...), and I have results which are less than 5% off from real measurements performed on Test-stands.

And... those results are in some cases, with pure tetrahedral meshes, and...with non-orthogonality up to around 82 degrees.

I have used OpenFOAM with meshes created using Netgen (starting with STEP and STL geometries), GMsh (STEP and STL), snappyHexMesh (STL Geometry), cfMesh (STL Geometry) and blockMesh (for simpler test cases).

Mesh sizes range from around 400000 to around 3,5 - 4 Million Cells.

As mentioned already in this thread, non-orthogonality above 70 degrees is not really a problem for OpenFOAM as long as you set up the rest of the case properly (for example, using the right dicretisation schemes, the right number of non-orthogonality correctors, etc...etc...).

Also, depending on the kind of simulations you are running, the physics of the system might have different mesh requirements. For example, if you are dealing with extreme shock waves, or high pressure gradients etc, non-orthogonality might be a problem. However, this is not something to do "only" with OpenFOAM... even other solvers would require specific mesh qualities for specific types of simulations.

I guess you will have to delve a little deeper to investigate where and what is exactly causing the simulations to crash in your cases.

Wishing you a great day ahead!

Regards,
Philippose

[wyldckat]

Greetings to all!

I happened to browse this thread sometime ago, but didn't have the time back then to give my 2 cents. Since this apparently popped up back again, and since I have some time today, here goes my many cents

Originally Posted by BrendaEM

I wish there was an open source alternative to OpenFoam. Reading these forums, I cannot believe that most people have not had major problems with OpenFoam.

Well... if you had accompanied more open source projects, you would have gotten a stronger sense that "open source" can be... ephemeral at times. Open source requires a lot of things for it to work properly, such as:

• a good user base
• a good community
• a strong and determined leader/community
• a good development platform
• sponsorship, may it be money-wise or time-wise

Now, there is a lot of popular open source software that has thrived over time, but there is also a lot more open source software that barely has any documentation, and sometimes barely works.

OpenFOAM is somewhere in the middle term of all of this. It has a good community and there are platforms available to assist in its evolution, may it be wiki-wise, forum, courses and so on.

Originally Posted by BrendaEM

Why must we get out experience and documentation from other users?
In the large sense, do you know how woefully inefficient that is, to have the blind leading the blind?

Well, many of those who "see the light" many times don't take or simply don't have the time to figure out how to help others reach said "light". They just move on with their lives and don't look back. If anyone asks them, some will likely say "it's in the forum, Google it"

Then don't forget that many of those who do have the necessary experience to help out, do not have enough time to do so... because the burden of knowledge and experience is usually responsibility, which can easily drain ones free time.

Originally Posted by BrendaEM

OpenFoam is open source.

Perhaps you might want to read this page: https://www.gnu.org/philosophy/selling.html

Originally Posted by BrendaEM

Logically, I wouldn't be looking a gift-horse in the mouth, if it were carrying me well, now would I?

I thought of a neat analogy sometime ago, but this seems a nice enough basis for another analogy:

The analogy premise:

1. Let's imagine that there is a new community service "adopt a horse", endorsed by some animal protection group, that has managed to gather enough funds to give horses "for free", along with the necessary starter's kit "How to live a happy life with your new horse".
2. The starter's kit comes with a around 30 horseshoes sets, one group for each major terrain, donated by the local "horseshoe blacksmith". It also comes with a book on the basics of what to feed a horse and how to change horseshoes.
3. But said kit doesn't specify what each set of horseshoes is for, namely for which terrain.
4. In addition, it also doesn't state all types of food the horse can eat, nor several other necessary details on horse's physiology.

Now, I could go on and list all of the problems that will arise from this kind of situation, but the analogy that can be drawn from this premise is as follows:

• A person that has an "adopted" horse claims that he/she spends 250 USD every month, in food, veterinary, horseshoes and so on. In addition, it takes him/her 20 hours per week of time to maintain the horse, for the first 3 months.
• A person that has purchased a horse from a company that has a dedicated support office, might spend on average 2500 USD per month, and only spends 3 hours per week to maintain the horse.
• Your horse is currently having a lot of problems dealing with a new kind of cheap ration... something along the lines of something that a goat would easily eat and not complain about, but the horse just keeps getting sick and weak. Then there is the problem with the horseshoes, because you've been taking the horse to go around in terrain unsuitable for said horseshoes... which can be inferred as being the horse's fault.
• On the other hand, the person that purchased the horse for a high price, has none of these problems. Possibly, the horseshoes are even self-adaptive to each terrain specification and so on.

Anyway, the conclusion to draw from all of this analogy is this: OpenFOAM is a lot like doing DIY CFD (Do It Yourself Computational Fluid Dynamics); but it's not like Ikea furniture, where a set of pre-defined furniture components has already been prepared to be just a matter of following the instructions for setting it up.


-----------------

Now, coming back from the analogy: It can be defended that OpenFOAM is as finicky as real life, akin to Earth's gravity certainly not being user friendly, but it's 99.9% reliable.
OpenFOAM can seem unforgiving, but keeps always on one's toes because we're using it to simulate real life. One small error or overlooked detail, is enough to be simulating things with the wrong unit scale or fluid properties.


Meshing-wise, there have been some recent discoveries, which are somewhat documented on this thread: http://www.cfd-online.com/Forums/ope...mesh-pipe.html
Also, indirectly related to this: http://www.cfd-online.com/Forums/ope...e-changes.html

As for mesh diagnosis, here are some ideas:

• http://www.cfd-online.com/Forums/blo...onditions.html
• http://openfoamwiki.net/index.php/FA...is_in_ParaView

In order to make using OpenFOAM interactively easier, here's a list of known GUIs: https://openfoamwiki.net/index.php/GUI - if you feel that the open-source GUIs one that list feel too... finicky, then keep in mind that they are open source


As for tips for CFD newbies in general:

• http://www.cd-adapco.com/blog/dr-mes...rt-finish-race
• http://www.symscape.com/blog/cfd-tip...ctive-new-year

Best regards,
Bruno

[BrendaEM]

wyldckat, you are questioning my idea of what open source software is about, and the economics of open source?

So, the OpenFOAM process goes as follows: download software, struggle, ask questions on forums, get answers, get your degree, and walk away.

Don't work on the wiki.
Don't make another tutorial cases.
Don't create a structured knowlege-base.

Do you realize that I have had someone write to me personally, stating that they indeed have had problems with OpenFOAM, that at least some of the problems I've posted are valid?

I don't think they are going to post on the forum, what they wrote to me, because there is peer-pressure here.

I have no interest in fitting in this community. I want to play the devil's advocate, and say to the community, that the emperor has no clothes, that things needn't be quite so challenging to learn how to run a case.

The tutorials jump from a hello-world extruded 2d case, to the motorbike, which has linked case set-up files.

OpenFOAM's documentation is space. I would like to see comments on every line in all tutorial case files.

I have seen valid results from Open Foam. I've crashed it too, setting timesteps up, throwing untrapped, cryptic errors.

The non-orthognal issue is an issue.

[wyldckat]

Originally Posted by BrendaEM

wyldckat, you are questioning my idea of what open source software is about, and the economics of open source?

I actually didn't ask any question... I only partly outlined how things are working at this moment (didn't manage to outline every detail), somewhat in the same way you have only complained and pointed fingers at the problem

So I guess my first question for you is: Are you willing to help getting OpenFOAM where you and many others want this to go to?

Originally Posted by BrendaEM

So, the OpenFOAM process goes as follows: download software, struggle, ask questions on forums, get answers, get your degree, and walk away.

Don't work on the wiki.
Don't make another tutorial cases.
Don't create a structured knowlege-base.

From what I've seen so far, possibly somewhere between 50 to 90% of the people that have used OpenFOAM, have done it that way. Very few have looked at how things are, and have rolled up their sleeves and are doing their best (almost desperately) to fix this problem.

Originally Posted by BrendaEM

Do you realize that I have had someone write to me personally, stating that they indeed have had problems with OpenFOAM, that at least some of the problems I've posted are valid?

That's only one of many problems. Pressure isn't even the main issue. Simply there isn't enough people/time to help on the forum, let alone document every single detail that OpenFOAM has got
And if you ask the OpenFOAM Foundation to fix any and all of these problems, the answer you'll get is something along these lines: "we have not yet done that, because no one has sponsored it yet"

Originally Posted by BrendaEM

I don't think they are going to post on the forum, what they wrote to me, because there is peer-pressure here.

On the other hand, I've had people who simply don't want to post on the forum because they don't want to wait for an answer. Instead, they prefer that someone else who does know something about the solution, should take 2 or 3 times the effort (aka time) to both help that person who wants the solution immediately and to also post the question+answer in public, because the person who has the question is too important to be bothered by such trivial things...
And on top of that, some of those don't even care if they are pretty much stealing time from others who asked publicly, with each private question they ask Because their logic is apparently "if it's free anyway, then it's not stealing".

One possible diagnosis is peer-pressure, but another (from my point of view) is the insufficient resources, namely people who are willing to take the time to help others and also the lack of people willing to support this cause, namely to fix this exact problems you're complaining about.

Perhaps the real problem is the lack of coordination in the community... but guess what, that takes time as well

Originally Posted by BrendaEM

I have no interest in fitting in this community. I want to play the devil's advocate, and say to the community, that the emperor has no clothes, that things needn't be quite so challenging to learn how to run a case.

Well... from the way I look at it, instead of me posting here twice, I could have answered to two others who did have problems that were properly described and for which a solution was attainable within 20 to 120 minutes, for someone who does know about how to solve their problem. The answer would have been posted on the forum and available to the public. Some would consider that this would be a good step in the right direction.

The reason why I'm answering to you is actually because:

1. I'm too tired today to solve most of the problems still open on the forum.
2. I see in you what many have felt before, who do have the strength, but end up never rolling up their sleeves to join those who are trying to fix these problems. Although unlike others who simply turned their backs to us, you're complaining about it. Honestly, I appreciate that a lot! But I would also appreciate if you're also willing to poke and push and help the community in the right direction.

Originally Posted by BrendaEM

The tutorials jump from a hello-world extruded 2d case, to the motorbike, which has linked case set-up files.

OpenFOAM's documentation is space. I would like to see comments on every line in all tutorial case files.

I very vaguely remember seeing this discussed in the past here on the forum in at least 2 other threads and some 2-4 years ago. The project "foam-extend" as it is today was born with the idea to solve at least part of these issues, but has had a lot of problems gaining the necessary momentum.

Originally Posted by BrendaEM

I have seen valid results from Open Foam. I've crashed it too, setting timesteps up, throwing untrapped, cryptic errors.

The non-orthognal issue is an issue.

OK, then I think that there are a few things that need to be clear here:

1. For free, there is a lot of constraints on what people can do, simply because "for free" usually means doing it on one's own free time. Very rarely is someone being paid to help on the forum, wiki or anywhere else.
2. Many open source projects have no commercial company behind it. OpenFOAM has a few, albeit only one is able to directly affect OpenFOAM's source code. If any of them are hired for fixing this and any other problem, they will.
3. Learning how to use OpenFOAM is something that has always been a mind boggling issue. From my calculations, training one person in using OpenFOAM, can cost as much or more than a one year license for a commercial CFD software, may it be in real money or in equivalent time spent learning how to use it.
4. The people that try to help here on the forum try to have one clear objective: if each individual who has experience and has enough spare time to help and solve a few questions/posts/threads, then it's one more solution per issue that is out there in public and can be seen by all. In other words, information that can be Googled.
5. On the other hand, if those who post the questions are not sensible to the limitations that the "community" of people that have the expertise and enough time to help out, then the cycle repeats itself many times over:
   1. Inexperienced person asks for help, but does not follow the guidelines: http://www.cfd-online.com/Forums/ope...-get-help.html
   2. Experienced person has to decide whether to answer back to that new person that he/she should give more details. If he/she spends time with those who did not ask the questions with enough detail, then someone who did ask the question properly will not see an answer soon enough.
   3. And then there are several possible outcomes... many of which result in the inexperienced person feeling utterly helpless and not getting the help he/she needs... and the experienced person tries to cope with the fact that there are 30 or more people per month that need help and that are not able to properly follow the guidelines... and so on and so on...

... all of this to say:

1. I ask again: Are you willing to not only point fingers, but also begin to help in working in solving this and many other problems? Pointing fingers and finding problems is good, but without sufficient information, the issues will unlikely get solved any time soon.
2. For this particular problem, can you provide one or more example cases of the exact meshing/simulation problems you're complaining about? Because since the 5th of November I have not seen any one of the posts here with a practical test case.
3. Have you searched for any OpenFOAM-related courses free and/or commercial and try them out?
4. Have you looked for and/or at other CFD open source projects? Gerris and SU2 come to mind.
   Oh and Caelus-CML is a recent fork of OpenFOAM, with a lot more documentation... albeit missing a lot of features as well : http://www.cfd-online.com/Forums/ope...-openfoam.html

Best regards,
Bruno