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.
[COMSOL V4.0beta] Who has tried it, and what do you think ?
Posted 15.10.2009, 02:49 GMT-4 8 Replies
Please login with a confirmed email address before reporting spam
Hi
Whom has tried out COMSOL V4.0Beta so far, and what are your feelings ?
What about telling COMSOL developpers too so that the final version is even better ?
I'm rather convinced, I'm always sceptical to large changes, VISTA was/is still a nightmare fo me (I'm getting old ?) but once the COMSOL people had given their 1h introduction here at the coference in Milano, I'm getting more convinced, and I reall like the sequencing, it's some ideas similar to the "processes" in GIBBSCAM a very interesting CAD/CAM software for CNC machins. The "process" technique is that you group (user set-up, tailored to your need) a certain number of sequences, steps, and save that as a "process" then you apply this to a given geometry, and finalise the settings. COMSOL is not foully there but almost I beleive. I'm not yet shure that you can set up a process with some physics and your preferred plots, predefined as a "process". Still plying with it.
On thing more to implement is having perhaps a postprocessing function: external call: i.e. you define a set of boundaries or domains to work on, by a selection and you do a call to an external function (matlab, C++ java ...). The question is how to get the plots and data back (interface: a selection of boundaries out,m a GUI window in ?) in a JAVA window that you can "incorporate into the COMSOL GUI and save into the report. This could allow users with specific need to write a short postporcessing function, more complex than an analytical function (i.e. surface deformation decomposition into a large orhogonal polynomial base: type Zernicke or Legendre ...) and have this as standard part of COMSOL for the user. Probably this will be in the API "to come" functionalities.
Once you catch the first 3-4 sequences in V4 to set up your model, and understand to try out the right click (COMSOL is really going from never the right click, to everything in the right mouse click, hope its not also political ;) ten I would say V4 is rather "as ususal".
So what do you think ?
Looking forward to read you soon
Ivar
PS: I'm not COMSOL, just a happy user, seeing finally, 30 years after have finished my physics studies, how I can extensively use PDE's to do physics engineering and not only nice mathematics ;)
Whom has tried out COMSOL V4.0Beta so far, and what are your feelings ?
What about telling COMSOL developpers too so that the final version is even better ?
I'm rather convinced, I'm always sceptical to large changes, VISTA was/is still a nightmare fo me (I'm getting old ?) but once the COMSOL people had given their 1h introduction here at the coference in Milano, I'm getting more convinced, and I reall like the sequencing, it's some ideas similar to the "processes" in GIBBSCAM a very interesting CAD/CAM software for CNC machins. The "process" technique is that you group (user set-up, tailored to your need) a certain number of sequences, steps, and save that as a "process" then you apply this to a given geometry, and finalise the settings. COMSOL is not foully there but almost I beleive. I'm not yet shure that you can set up a process with some physics and your preferred plots, predefined as a "process". Still plying with it.
On thing more to implement is having perhaps a postprocessing function: external call: i.e. you define a set of boundaries or domains to work on, by a selection and you do a call to an external function (matlab, C++ java ...). The question is how to get the plots and data back (interface: a selection of boundaries out,m a GUI window in ?) in a JAVA window that you can "incorporate into the COMSOL GUI and save into the report. This could allow users with specific need to write a short postporcessing function, more complex than an analytical function (i.e. surface deformation decomposition into a large orhogonal polynomial base: type Zernicke or Legendre ...) and have this as standard part of COMSOL for the user. Probably this will be in the API "to come" functionalities.
Once you catch the first 3-4 sequences in V4 to set up your model, and understand to try out the right click (COMSOL is really going from never the right click, to everything in the right mouse click, hope its not also political ;) ten I would say V4 is rather "as ususal".
So what do you think ?
Looking forward to read you soon
Ivar
PS: I'm not COMSOL, just a happy user, seeing finally, 30 years after have finished my physics studies, how I can extensively use PDE's to do physics engineering and not only nice mathematics ;)
8 Replies Last Post 23.10.2009, 02:40 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi Ivar,
Thanks for posting this thread, which is the first one concerning the new version of COMSOL.
I really appreciated the new release of COMSOL, especially the new features and the functionalities like the parallel computing and new CAD applications.
Of course the GUI is new and it will take time to get used to it, but it facilitates the construction of the model and also the post-processing. I was also fascinated by new tools emerging from the combination of Matlab with COMSOL.
The possibility to create API with COMSOL was also a new and useful feature, and also the possibility to create its own Java code and to use it for post-processing.
It was really a successful conference and I hope that the next one will be more fruitful.
I don’t know if you were there or no, but during the talk of Niklas Rom and in one of the slides, it was written “we don’t want swap…”, and personally I thought about your remark concerning the use of hard drive when COMSOL is computing the solution…It seems to me that the hard drive activity that you noticed was related to the load of something else (loading solvers for example). During the computation everything is stored in the RAM.
Concerning the test of the version 4.0 Beta:
I have just installed it in WindowsXP 64bit and Ubuntu 64bit.
For people who are using Linux and especially Ubuntu, they should install additional OpenGL library, otherwise COMSOL will crash during the starting.
*** Herein the library: LibGL1-mesa-SWX11 and its debugger.
Another point is that you cannot load a model constructed with the previous versions. You should wait for the new release. To test the new version, you should build a new model and run it, in order to compare.
Hope it helps
Cheers,
Faycal
Thanks for posting this thread, which is the first one concerning the new version of COMSOL.
I really appreciated the new release of COMSOL, especially the new features and the functionalities like the parallel computing and new CAD applications.
Of course the GUI is new and it will take time to get used to it, but it facilitates the construction of the model and also the post-processing. I was also fascinated by new tools emerging from the combination of Matlab with COMSOL.
The possibility to create API with COMSOL was also a new and useful feature, and also the possibility to create its own Java code and to use it for post-processing.
It was really a successful conference and I hope that the next one will be more fruitful.
I don’t know if you were there or no, but during the talk of Niklas Rom and in one of the slides, it was written “we don’t want swap…”, and personally I thought about your remark concerning the use of hard drive when COMSOL is computing the solution…It seems to me that the hard drive activity that you noticed was related to the load of something else (loading solvers for example). During the computation everything is stored in the RAM.
Concerning the test of the version 4.0 Beta:
I have just installed it in WindowsXP 64bit and Ubuntu 64bit.
For people who are using Linux and especially Ubuntu, they should install additional OpenGL library, otherwise COMSOL will crash during the starting.
*** Herein the library: LibGL1-mesa-SWX11 and its debugger.
Another point is that you cannot load a model constructed with the previous versions. You should wait for the new release. To test the new version, you should build a new model and run it, in order to compare.
Hope it helps
Cheers,
Faycal
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
I have not tried 4.0 beta...but personally I would like it if COMSOL gave a better API interface. Complicated problems are difficult to do using the GUI, however I dont think the API provided at present is good enough. I would definitely like it if it were possible to do more low level programming within COMSOL.
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Dear Faycal:
Thank you for posting this information about the opengl library. I have a Ubuntu system with the nvidia drivers (which install opengl drivers for their chips) and comsol v4-beta worked fine on that system. I have another laptop with ubuntu installed that uses drivers for the intel chips on the laptop. comsol would not run on that system unless I issued the command "comsol -3drend sw". However, with your help in installing the package "libgl1-mesa-swx11", it will now run without this extra switch on the command line.
Another very interesting piece of news from the conference for me is the upcoming release of the CFD module.
Also, many new features associated with the GUI that provide increased functionality not available in prior versions.
Thank you for posting this information about the opengl library. I have a Ubuntu system with the nvidia drivers (which install opengl drivers for their chips) and comsol v4-beta worked fine on that system. I have another laptop with ubuntu installed that uses drivers for the intel chips on the laptop. comsol would not run on that system unless I issued the command "comsol -3drend sw". However, with your help in installing the package "libgl1-mesa-swx11", it will now run without this extra switch on the command line.
Another very interesting piece of news from the conference for me is the upcoming release of the CFD module.
Also, many new features associated with the GUI that provide increased functionality not available in prior versions.
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi
For me there areso many new goodies in the V4beta (which is to be tested, not yet to be used for serious work) that it will be difficult to talk about them, I do not know where to start. The API interface was announced for the middle of 2010, currently it's only a Java I/F but this was not enhanced as the solution for interfacing to the GUI, here we need to be somewhat more patient.
On the other hand here are really many new concepts how to sort data (selections, new name of groups) add operators (replacing most functions and integration variables), sequncing geometry and meshing with parametrised geometry that you can change at will, enhanced muti entry solving ad postprocessing with different sequencing.
It was 200% worth to come to the conference, to discuss with the COMSOL people and users. It took me some hours to understand the new approach, the explanations at the beginning of the conference were very helpfull, but then the possibilities this gives are outstanding, specially compared with today.
Clearly COMSOL needs a few months to clean it up and stabilie better the software, but I'm really entousistic about the new approach, and eager to use it for real.
You can find more official info on Nikla's Multiphysics Blog
For me there areso many new goodies in the V4beta (which is to be tested, not yet to be used for serious work) that it will be difficult to talk about them, I do not know where to start. The API interface was announced for the middle of 2010, currently it's only a Java I/F but this was not enhanced as the solution for interfacing to the GUI, here we need to be somewhat more patient.
On the other hand here are really many new concepts how to sort data (selections, new name of groups) add operators (replacing most functions and integration variables), sequncing geometry and meshing with parametrised geometry that you can change at will, enhanced muti entry solving ad postprocessing with different sequencing.
It was 200% worth to come to the conference, to discuss with the COMSOL people and users. It took me some hours to understand the new approach, the explanations at the beginning of the conference were very helpfull, but then the possibilities this gives are outstanding, specially compared with today.
Clearly COMSOL needs a few months to clean it up and stabilie better the software, but I'm really entousistic about the new approach, and eager to use it for real.
You can find more official info on Nikla's Multiphysics Blog
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
I was at the COMSOL Conference in Boston recently where I saw V4 in action. Although I haven’t test driven it myself much yet, I think this is a significant step forward that adds crucial new capabilities.
Among the many improvements I especially like
1- the sequencing concept where you can set up the workflow and modify and re-run it with ease
2- ability to define what used to be coupling variables as more general operators (e.g. you don’t need to decide in advance which variable the operator will operate on; likewise you don’t need to define multiple coupling variables that uses the same operator).
As for feedback, I look forward to improvements in cluster computing capabilities. In addition a very important capability for me (and fortunately almost trivial to add) is to define user-customizable hot keys so that all the common operations that can be done through menus / icons can also be called up through simple key strokes possibly in combination with mouse clicks/moves. For example, it would be very useful to have such shortcut keys to do “zoom extends, box select, rotate, pan, zoom, bring up global definitions or expressions etc”
Regards,
Ozgur
Among the many improvements I especially like
1- the sequencing concept where you can set up the workflow and modify and re-run it with ease
2- ability to define what used to be coupling variables as more general operators (e.g. you don’t need to decide in advance which variable the operator will operate on; likewise you don’t need to define multiple coupling variables that uses the same operator).
As for feedback, I look forward to improvements in cluster computing capabilities. In addition a very important capability for me (and fortunately almost trivial to add) is to define user-customizable hot keys so that all the common operations that can be done through menus / icons can also be called up through simple key strokes possibly in combination with mouse clicks/moves. For example, it would be very useful to have such shortcut keys to do “zoom extends, box select, rotate, pan, zoom, bring up global definitions or expressions etc”
Regards,
Ozgur
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Hi Paul
I partially agree with you, having run Ansys & Nastran/Patran for a couple of decades I agree these are "faster" to get straight to the solution for a classical structural/thermal FEM model for an engineer (compared to an academic need) but once you want to do a little more, add omething a little more fancy, or going into mems, I, today, prefer largely COMSOL.
And I beleive from what I have seen so far from the V4beta that the new goodies is going to enforce his, for the engineers (still improvements there to be done) and certainly for the academics.
As a development/system engineer I strongly beleive that COMSOL has today, an unique position, that I, in no way, would miss to follow and to participate to, from an user point of view, bring to a good engineering level. I'm only missing a few things in structural, and a few items/scripts in the formal PA/QA approach to verify andf validate my models in a systematical way.
For the MATLAB link, I do not beleve COMSOL will leave it out, at best I believe they will add a more general API llink, but it is also clear for me, we (in our compan) have a tought fight with MATLAB, they are good, they have an excellent product, but we have very serious problems to pay the amount they want and get a decent ROI on that investment. MATLAB cannot put a "patent" on mathematics, the way they are trying to. The good thing: they are giving gifts to their compoetitors, to develop good compeeting products.
Thanks for the remarks so far, I hope for more
Ivar
I partially agree with you, having run Ansys & Nastran/Patran for a couple of decades I agree these are "faster" to get straight to the solution for a classical structural/thermal FEM model for an engineer (compared to an academic need) but once you want to do a little more, add omething a little more fancy, or going into mems, I, today, prefer largely COMSOL.
And I beleive from what I have seen so far from the V4beta that the new goodies is going to enforce his, for the engineers (still improvements there to be done) and certainly for the academics.
As a development/system engineer I strongly beleive that COMSOL has today, an unique position, that I, in no way, would miss to follow and to participate to, from an user point of view, bring to a good engineering level. I'm only missing a few things in structural, and a few items/scripts in the formal PA/QA approach to verify andf validate my models in a systematical way.
For the MATLAB link, I do not beleve COMSOL will leave it out, at best I believe they will add a more general API llink, but it is also clear for me, we (in our compan) have a tought fight with MATLAB, they are good, they have an excellent product, but we have very serious problems to pay the amount they want and get a decent ROI on that investment. MATLAB cannot put a "patent" on mathematics, the way they are trying to. The good thing: they are giving gifts to their compoetitors, to develop good compeeting products.
Thanks for the remarks so far, I hope for more
Ivar
Please login with a confirmed email address before reporting spam
Posted:
2 decades ago
Version 4.0 will include a LiveLink to MATLAB providing full associativity between the environments. Users of MATLAB will be able to call COMSOL Multiphysics functions directly from the MATLAB prompt, Simulink and the Toolboxes .
COMSOL is a MathWorks Connections Program partner and we are committed to maintain a smooth interface to MATLAB.
For more information see
www.mathworks.com/products/connections/product_detail/product_35492.html
COMSOL is a MathWorks Connections Program partner and we are committed to maintain a smooth interface to MATLAB.
For more information see
www.mathworks.com/products/connections/product_detail/product_35492.html
(...)
For the MATLAB link, I do not beleve COMSOL will leave it out, at best I believe they will add a more general API llink, but it is also clear for me, we (in our compan) have a tought fight with MATLAB, they are good, they have an excellent product, but we have very serious problems to pay the amount they want and get a decent ROI on that investment. MATLAB cannot put a "patent" on mathematics, the way they are trying to. The good thing: they are giving gifts to their compoetitors, to develop good compeeting products.
Thanks for the remarks so far, I hope for more
Ivar
(...)