Its difficult to say because it depends on your case complexity, computational power (such as High Performance Computer), meshing and the equations that you want to solve. Then you can compare the experimental data or correlation from other researcher.

In my opinion , FLUENT is better than COMSOL if we see in number turbulence model on each software.

But, Comsol is more user friendly. If your case is uncommon you can use User Defined Function in fluent (c language base) but if you are solving by COMSOL you can directly put the equation without writing in C such as FLUENT.

Dear Alaa Temimy,

you can be fairly certain, that all commercial codes available on the market are free of coding and algorithmic errors, i.e. they implement a numerical method for the solution of your problem correctly.

However, the numerical method itself, inherently, includes sources of error. These come primarily from space and time discretization (computational grid density and time step) as well as from which method has been used and terms up to which order have been neglected.

Furthermore, the mathematical-physical model, employed to describe the process you are simulating, is usually also an approximation of reality. For example, we know fluid viscosity is temperature dependent, but we choose to neglect that in our simulations, etc.

If the problem simulated is non-linear, a user-prescribed criterion is usually used to set the desired lowering of the error norms, which also contributes to the overall error.

My suggestion is to first identify all sources of error and try to keep them of the same order.

Secondly, one should always perform a nodalization analysis (and possibly a Richardson extrapolation) to examine, how the solution changes when different grids and time steps are used.

A short answer to your question, it is impossible to give a percentage to estimate the accuracy of a simulation. And beware, the accuracy does indeed depend on the expertise of the engineer doing the simulation.

Thank you Teguh Hady Ariwibowo

Thank you Horimek Abderrahmane

Thank you Jure Ravnik

Fluent, Comsol etc are software tools to solve energy and fluid flow problems. Especially in order to do in fluent, a three dimensional model has to be prepared by Gambitt or Auto CAD and should be imported to Fluent to solve. However, basic equations like continuity, momentum and energy equations are inherent in software. While solving a certain problem, certain assumptions and flow constraints should be given. Hence in any software, the assumptions should be taken precisely as near as to the problem defined so that results will get accurately. However, the results got from this software is not also correct but varies by definite percentage because this depends on assumptions, drawing and flow constraints as mentioned above. This should be validated for any one set of value experimentally or by theoretical modelling and later the entire work can be done by software. However, the exact matching is not possible. The deviation may be permitted up to plus or minus 20%. Complicated problems can be attempted through softwares since in some cases becomes unavoidable and can be trusted after testing with simple problems.

Thank you Pisipaty Srinivas Kishore