CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Lounge (
-   -   FSI and COMSOL Multiphysics (

Caluca November 20, 2011 19:19

Major players in CFD
Okey, so we know that Ansys and CD-Adapco are major players when it comes down to CFD. OpenCFD is well known for its free and open source CFD software. Numeca (Belgian CFD company) is particularly well known by me, because I am currently doing an internship there.

However as an Aerospace Engineering student I am off-course interested in CFD (aerodynamics master track + a dozen of numerical electives), but to be more specific, I am particularly interested in FSI. So when it comes down to FSI specifically...would Ansys and CD-Adapco still be the major players?

Due to a quick search on the web, the originally Swedish company, COMSOL Multiphysics caught my attention. As their name already stipulates, they got all kinds of multiphysics simulation codes jammed into their software. Unfortunately I have not found any reference to this company on CFD-online. So my second (three part) question is: How well known is COMSOL, how good is their software and how well do they perform relatively to other CFD companies?

abdul099 November 23, 2011 18:00

I don't know COMSOL. But a lot of CFD packages can be linked to other software.

Just an example, Star-CCM+ could be linked at least to Abaqus and GT-Power and I think to some code to simulate oil & gas stuff. I don't know if there are any other options since I currently don't have the user guide opened.
This gives the CFD software the ability to deal with other physics while each software is concentrating on what it can best.

I don't know about other software like Fluent, CFX etc since I'm using only Star-CCM+. But I'm nearly sure most often they do the same: Concentrate on the one thing (or some commonly related things) the software could do the best job and extend it's abilities by linking to other software packages. (Okay, Ansys has it's one stress analysis code and an own CFD code and what else stuff ever, all linked together in a single working environment. But generally it's also just a bunch of independent programs linked together)

To have one single software package which can handle all physics one might want to combine would mean: One huge, complicated and most important, EXPENSIVE software. It would be a lot of effort to implement all stuff, or there would be some gaps.

Caluca November 23, 2011 19:11

Thank you for your reply abdul099.

For one part I agree with you. Optimising software towards one specific task, has the tendency to produce the best results in their respective fields. Thus logically, as you have already stated, combining multiple single-task optimised software to work collaboratively on a multi-physics problem could prove quite effective. However, on the contrary, the question that remains is whether the interface between the different packages (developed seperately by individual companies) is robust and/or reliable.

Naturally, companies have spend quite an amount of time to work on an interface with complementary software from other major companies, yet one must not forget that companies might not share the same philosophy or working methodology. Hence results obtained from a combined effort of their respective software might be extremely accurate, but may not have been obtained in the most efficient manner.

I am guessing that you think that COMSOL tries to solve a problem within an unified framework however that is not the case. From what I know of COMSOL (which is not much) is that they have produced several modules capable of handling single physics problems. These models are then linked to one another when dealing with a multi-physics problem. In my opinion an advantage of their software package might be (in contrast to linking several packages from different companies), that the collaboration between the modules is more efficient due to having one company, one philosophy and one working methodology behind all the individual modules.

Another argument resides within the fact that companies usually intend to make money. What I am trying to get at is that companies do not like to disclose any details concerning their source code. COMSOL naturally has access to each and every module they own and therefore are capable of exploiting the full potential of a module in the most efficient manner during a multi-physiscs simulation.

In closure, as I have stated before, I have no experience with COMSOL and the above argumentation is therefore by no means more than a deductive reasoning. Whether COMSOL does meet my implied expectancy is something I have yet to discover. If you do not agree with me, please present me your train of thought (this applies also to others).

ps: Does anyone have experience with either only COMSOL or COMSOL and several other software? If so please elaborate on your personal experience with those software.

CP85 December 5, 2011 10:55

I worked during my University-Time with COMSOL 3.5 and 4.2. I used them for several projects.

In 3.5 u needed a strong backround about the differential equations and their terms.

4.2 is a bit more userfriendly and has a completely reworked userinterface. It is very comfortable to use and has some automatic funktions. But there are still a lot of equationbased options.

Both versions have the posibility to easily enter and solve your own differential equations and boundary conditions. For Linking problems you need some Knowledge about how the equations are actually linked. But than it is a simple transfer of variables.
The main costumors of COMSOL were researchers and scientists. With version 4.2 they are trying to get more in the industrial market.

Originaly COMSOL was a FEM Module from MATLAB. It soon got to big and a separate product was created.
COMSOL is FEM based (also the CFD-Module). Due to the history it has a very strong link to MATLAB. U can link your models to MATLAB. So u can use the full power of MATLAB-scripting.

Caluca December 6, 2011 04:53

Thank you very much for you insight CP85.

Aside from not being the most user friendly software out there, how good are the results and how fast are simulations in comparison to other comparable FSI solvers?

CP85 December 6, 2011 08:57

The Results are as good as your model is ;)

Sadly I have no benchmarkdata about the solver-speed.

All times are GMT -4. The time now is 17:17.