Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Straw Poll: Who would like to see the model gallery documentation overhauled???
Posted 2013年1月25日 GMT-0500上午9:26 9 Replies
I thought I'd post this here to get the community opinions before suggesting it to support.
If I want to learn how to use an aspect of Comsol I work through example models and I visit the forums, and often the advice from the forum is to work through a particular example model. This is, in my opinion, the best way of learning how to use any technical software.
I believe that this process is quite seriously hampered in Comsol compared to other products like Labview/Matlab or even things like Adobe Creative Suite. I think it is hampered because of the structure of the model documentation.
Basically, I think it is near enough useless to give users a step-by-step instruction sheet for how to build an example model without telling them WHY they are doing what they are doing. In almost all cases, when I try to adapt what I have learned from working through the example models to my actual applications, there comes a point where something fails and I have no way of figuring out which aspects of the example models are crucially important and which aspects should be changed because I often don't know what they were doing in the example model or why.
One clear example is in the application of boundary conditions to particular physics (which BCs are necessary for the physics to solve, which are problem-dependant).
Another example of the type of problem is that the built-in material AIR cannot be used with the electric currents interface without changing its conductivity from 0 to some small positive value. I realise the physical reasoning behind this, but this must be such a common problem that I can't believe there isn't some up-front info in the model library examples that explicit state this. It's the kind of problem that can hold new users up for days/weeks and it's not the only one: a user might get past this only to run into another similar problem in the very next step of his/her model.
The way the documentation is written, you might come across a line that says something like "change the value of conductivity from 0 to 1 S/m in the material properties". But there is no explanation of why you are doing this, or whether this is generally important or only important to the specific problem in the example. This means that you could be searching around the documentation for an answer to your problem only to read right over it again and again.
Basically I think the barrier for entry into COMSOL is needlessly high because of the way the documentation is handled. With a relatively tiny amount of work on the part of Comsol (hell, it could even be done by the community) the documentation could be updated and thousands of new users would be able to answer their own questions instead of raising hundreds and hundreds of forum and support tickets. All that is needed is the inclusion of one or two explanatory sentences after each section of the step-by-step instructions.
This is not a rage post, it doesn't relate to a specific problem I am having right now, I genuinely think that Comsol are wasting an opportunity to grow their community and broaden their user-base.
I'm interested to hear your thoughts.
intersting topic !
I can remember I felt it also like that a few years ago when I started to use COMSOL, but I must admit that today, I have other arguments. Some of the examples you are giving come from the fact that you have NOT gone through the model on a quick paper sketch, stating which variables to solve for, on which domain, how many dependent variables do you have, which level of derivation is required, hence how many BC Direchelet or Neumann are required, made the list and controlled that the count is fully there.
If you choose EC you solve for "conduction" and if air is defined as an isolant (0 conductions) the physics you are applying are wrong (out of the main hypothesis = conduction) For isolants you should add a simple Ampere law node and define your air like that, or "cheat" and give it a little conduction.
But then do not forget that values of material properties in a ratio of more than 1:10^6 to 1:10^8 will not be correctly solved, because our computers today have 64 bit FPU's and even in Double precision 128 bits, they do not represent the real number scale with enough precision to handle too large differences (check the value of sqrt(eps))
For BC's you must understand your equations enough to count the number of fixed points you give to get an unique solution, as most PDEs give many equivalent solutions, each offset by some constant or slope, and COMSOL cannot decide which are "physical" and which are not, we end users must do that.
Then the question boils down to: should COMSOL doc (and examples) revrite all physics ?
I have some 500 books on differrent aspects of physics (and chemistry) in my bookshelf in my office, they do not even cover the full range of topics that are covered by COMSOL, in addition COMSOL has managed to unify all these physics in such a way that you can mix them all with COMSOL. But this is by far not so for most books: they consider at most 1-2 physics and do not care if the way to present the subject is compatile or not with all other existing physics.
So today, I believe COMSOL is already doing it a great job in unifying all their physics, and to document it, densly I agree, you need to read the 4-5000 pages of docs a few times.
So, to do better they must open a "Faculty of COMSOL use of Physics" but that could also be done by any University today. So shold we users pay more to use COMSOL and to finance such a Faculty ? I prefer not to, and would rather like to see more universities and high schools teach physics with COMSOL, and there one could discuss these issues of what is required, how many BC's, what does this mean.
Have fun COMSOLING
NB: I'm not from COMSOL, just a "paying" user, as most of you out here ;)
Anyone else have an opinion?
The documentation contains no worked examples but is just a text description. It would be nice to see simple setups in the documentation. And for the worked examples posted online, the newer versions (for ver4.2 +) tend to explain more of the underlying reasons for what is going on in the background. For them to cover all of the operating principles would be a major task though. I will circle back and say that the program is definitely meant for science/tech backgrounded people. And for someone to even attempt to model using PDEs or ODEs, that requires some education as well. It is not a simple program to use and it can be if the problem is not complicated. Most problems that need to be solved are complicated.
I also think that their intent is to have you buy the program and then you will take their classes to learn how to use it. One could look at that in a couple different ways. You either look at it as part of the operating costs or you look at it as an added cost that takes time from you. If you paid a lot of money for the program, you would expect it to be "self teaching". But it is not.
I would just like to add that I am a PhD student myself. My undergrad and masters were in physics but now I work in biology. Comsol is used heavily in our group and even the post-doctoral researchers have the same 'baby-steps' problems you mentioned. It doesn't even apply only to people new to the program, my collegues and I have been working in Comsol for years, and we each primarily use different modules which we've become fairly adept with. If the situation arrises that I want to use a new module though, I feel like a complete beginner again.
So I'd like to stress, that this problem applies even to the "PhD types" and academics. I think the only people who have a headstart are people that have a specific FEA-based background (like some engineers perhaps). I think many people come to Comsol for the first time with no prior FEA background and while the UI is amazing and makes the model building intuitive, it's all out the window when you click that 'compute' button and it throws up an error.
I honestly don't mean to underestimate the importance of users really trying hard to learn the principles of the software and the physics behind their problems - this is what Ivar has always stressed and I think he's right to do so. I just think the documentation could make it considerably easier for those without an FEA background to do exactly that.
Thanks for adding to the discussion, Dennis.
By the way, Amanda Brunton from Comsol kindly emailed me to express interest in this issue, and she's told me that she'll keep an eye on this thread so you opinions would be most welcome.
I have another comment here: as I'm not so sure COMSOL is so "easy" for a FEM specialist coming from "older" FEM programmes.
I had done quite some N&A programming (two well know programmes used by different engineering communities), actually I followed closely and interpreting the results, as at that time we had and older collegue with many years of FEM experience. When he retired, we switched to COMSOL (because of the multiphysics) and I had to take over the FEM analysis for the group, I must say I had a hard time, because my mind was turned in the "old way, concetrating first on the mesh, applying BC to the nodes, and you loose the nice approach of COMSOL: thinking physics first. And before, we were limited to structural and some thermo-structural analysis, and a detailed model took us months to build, run and validate!
Now, that I have finally understood the COMSOL approach, I'm rediscovering physics and really enjoying it, as I can mix physics as never before, and even the FEM math aspect is taken over by COMSOL, under the hood, so now you can really think physics first, and you discover you have forgotten a lot, and that most simplification we use "traditionally" are really drastic, detailed effects show up, and the nice animated images from COMSOL lighten up your presentations, even if pages of equations, to verify the same calculation, impresses the audience, the "wow" effect comes from the images to illustrate even simple models.
Concerning prices, I must admit that today COMSOL is far cheaper in purchase (for an industrial) than the other tools I know and can mix far more physics in a simple way (but lets not tell COMSOL, they might change that ;).
Finally for the training, all (serious) software providers proposes courses, what I see with COMSOL is that the quality of the engineers giving the courses (at least the one I meet here in Europe) is well above average, something I havent really felt for other tools and courses I have taken and prices are rather mid leve (but I live in a "rich" country, what about places with many young enginners but far less budgets for their work ?l.
Nevertheless, training should ideally be done in the high schools, technical schools/univeristies and the univeristy/enginnering level. But this require a change in the mentality (or of generation?) of our professors/teachers I believe. It's a subject I often exchange with those of my study colleages that are now teaching: how to learn to younger students the essence of COMSOL Multiphysics programming? whithout having to learn them all the subtilities of FEM math theory, but still get them to "feel" when the results are fundamentally wrong ? I havent found the answer yet, suggestions highly appreciated ! And for class kits, the COSMOL price is really competitive, I find.
"COMSOL is only for PhD's": I do not believe so, but I agree, for me as engineer, it's still too "science prone", and some house-keeping actions take too long to set up, repetitively. I would like to have more built in V&V (verification and validation) but here I see also a challenge, as the traditional rules do not all apply to COMSOL, and the multiphysics aspects make it complex to define some simple rules.
For me the main questions remain:
- how to verify that the geometry has ben correctly/created imported, units, sizes, scale ...?
- are all material at the right place ? total mass, inertia ...
- are the correct BC's set up nicely ? any overconstarints?
- are the main conservation laws respected ? integration over relevant BCs
- are the mesh density corret w.r.t. dependent variable gradients, are these correctly resolved ? relevant plots ?
But all this applies differently for each physics and physics combination, quite a subect to treat ;)
But this is essential to make the tool even more user friendly or less error prone (due to typos or user misunderstandings)
Comments from the academics teaching physics with COMSOL highly appreciated :)
Why shoudl'nt our young science/math students, all over, that today are writing fancy programmes or playing advanced computer games, not have as much fun simulating physics with COMSOL (as I do) ?
PS: COMSOL had one interesting webinar (in German), last year, commenting experiences in using COMSOl for the student training in the technical universities in Germany it's probably recorded somewhere
My feature suggestion only goes as far as a (relatively) small documentation update. Not to devalue the training days, but I think I speak for a lot of people in that I like to have the resources to work through problems on my own if possible.
I know it's a very different program, but I like the way Mathworks does the Matlab documentation. It probably wouldn't suit Comsol docs to have every seperate function layed out in separate pages as in Matlab, but it would be nice to see a better connection between general theory (in the help files) and worked examples (in the model documentation). As it stands, they are completely separate and it is not straightforward to connect a given part of the general theory to a specific aspect of your model. This is my main problem with Comsol documentation.
Some of the examples are very well done while others are just a sequence of steps with no explanation whatsoever. Obviously there is a room for improvement here.
But the biggest problem are the help files, containing a lot of useless information like the following: "The value in the Maximum element size field specifies the maximum allowed element size."
We are in the process of rewriting the product documentation entirely. This is a major project that may take up to a year to complete. We will make available new versions of the various books as they are finished.
A new version of "Introduction to COMSOL Multiphics" has been made available today for download separately at www.comsol.com/shared/downloads/IntroductionToCOMSOLMultiphysics.pdf .
It is also bundled with the product download starting today.
As always, we welcome feedback.
just to "bump" this one:
I would be very happy to read a bit more about how COMSOL is teached in the different technical schools/universities aroudn the world, to give me some hints how this can be done in the most efficient way.
How do you deal with the documentation, recommend it, give fully you own lectures ...
In particular how can we learn to younger students the COMSOL and PDE solving via FEM techniques, without learning tem absolutely all the details of all FEM math principles ?
Dear professors, and students, how is it done at your place ?
Learn by doing, by studying (what the doc, the examples books, courses ...)
Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.