CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   Comparison: COMSOL, Fluent, CFX (

glennfulford November 21, 2009 20:26

Comparison: COMSOL, Fluent, CFX
Hi, I am fairly new to cfd (I have more of an applied maths fluid dynamic sbackground).

My PhD student has been using COMSOL to solve a problem involving pesticide-fumigant dispesion in a porous media: solving Laplaces equation for the pressure and velocity and and advection-diffusion equation for the concentration of fumigant.

We have had a lot of troble getting COMSOL to give solutions for some 1D test problems, non-vergencemoscillations and negative concentrations are problems. We have a fairly high Peclet numbers, several hundered or thousand (on the scale of the problem, net the cell size). In COMSOL there seems to be few options for controlling upwinding algoritms for handling the advection term (or am I missing something?)

We have avaiable both CFX and FLUENT, and we are thinking of using one of them instead. Can anyone advise me if these are likely to be more successful (they are both finite -volume based, unlike COSOL which is finite element, so I have been told that this is better). TDo they provide options to select have the advection terms are handled (flux limiter options, or better yet, allow one to implements an algorith rather than just choosing from a menu)?

I would appreciate any advice on this.
thanks very much, Glenn.

Ahmed November 22, 2009 00:17

Finite element based solvers suffer from this oscillatory behavior, the general practice to suppress this behavior is the use of what is called "artificial viscosity" e.g. if nu is the viscosity of your fluid, you start the solution with 1000 nu (this number is not a rule of thumb but just to give you an idea, you can even go for higher numbers depending on your problem, i.e you need to experiment a bit) then you gradually reduce this nu after a hundred or so iterations, that is to say when you see a stable solution (again, a bit of experimenting) till you reach the true value of nu.
I would not decide for a change till I discuss the situation with the technical support of the programme you are using, anyway, If you decide for a change, do not expect what you are asking for in a commercial programme, you can select between first order, second order ....etc..
If you want to implement your own scheme etc.., you have the open source numerical tool kit known as OpenFOAM (R), but be advised, the learning curve is very steep.
Good Luck

andy_ November 22, 2009 03:05

CFX and FLUENT are commercial CFD codes with bits added. COMSOL was a small MATLAB FE structures code with bits added. Given that you have a CFD problem that is causing issues the former are likely to be a better bet.

A straightforward structural FE implementation does not work for CFD at high Peclet numbers and needs to be modified. The modifications (Petrov-Galerkin, mass lumping, etc...) drive it towards a FVM method (which is a particular form of FE method anyway). Whether something is labelled FE or FV is not important if it has been well developed to follow the physics of a particular class of flow. But it is more relevant for general codes.

The CFD codes will have advection-diffusion equations for scalars where the value remains between the physical bounds because it is a requirement for many turbulence models. However, this will be achieved by introducing unphysical diffusion in regions where the grid is too coarse to resolve the gradients.

All times are GMT -4. The time now is 11:28.