[hbulus]

Hello everyone,

I am doing some validation analysis on Openfoam and fluent. When i selected same order of schemes, fluent shows enormous advantage on stability and convergence. The fluent really converged very soon according to Openfoam!
In sense, they should use same formulations or limiters, why does there a huge difference? What makes the fluent so powerful?
Do you have any opinions why the situation is like that ?

[Tobermory]

I don't know the specifics, but it is my experience that the commercial solvers are much more forgiving and robust to use than OpenFOAM, and of course require very little understanding of the numerical schemes under the hood ... that's where they earn much of their licence fee.

I wouldn't say that Fluent is necessarily more "powerful" though - it could just be that you do not have the optimum OpenFOAM settings, whereas Fluent is probably always running at near optimum settings.

[joegi.geo]

well you need to compare apples with apples, if you are using the default method in fluent which is the coupled solver, it will be orders of magnitude faster.

In my personal experience segregated solvers have the same performance, sometimes openfoam is faster. So compare SIMPLE, SIMPLEC or PISO methods.


jg

[hbulus]

Guys i know the differences in approaches. I am grateful but please give some hints that you know, not just the comments. Like;
1- Fluent always uses limiters in grad schemes or even in interpolation schemes(not sure about interpolation schemes).
2- Fluent uses nNonOrthogonalCorrectors as a default by checking non orthonality (Not sure )
3-...
..
I am looking technical advices, thanks your answers.

Have good days !

[Tobermory]

Well, JG did just give you that - he mentioned something that I had forgotten: Fluent's coupled solver. Did you use that in your test? If yes, try it again with the segregated solver. Finally, remember also that Fluent is hardly going to make it easy for you to understand their optimisation ... it is a commercial product after all.

[arjun]

Originally Posted by hbulus

Hello everyone,

I am doing some validation analysis on Openfoam and fluent. When i selected same order of schemes, fluent shows enormous advantage on stability and convergence. The fluent really converged very soon according to Openfoam!
In sense, they should use same formulations or limiters, why does there a huge difference? What makes the fluent so powerful?
Do you have any opinions why the situation is like that ?

One of the reasons why Fluent converges faster is that it is commercial code and people pay for it to use it. When people pay they do not want headaches so as a company Fluent has seen number of such customers and they have people who worked on making the software robust and accurate.

OpenFoam on the other hand is mainly developed in academics and people developing various modules move on in life once their task is done. Their goal is not to provide customers robust solvers but to finish their task.

Two things developed very differently over the years. This is why Fluent is much more robust and efficient.

Numerically it is very hard to tell as far as Fluent goes because they have tendency not to mention most critical info in their user guides. They keep people guessing about it.

[arjun]

Originally Posted by Tobermory

Well, JG did just give you that - he mentioned something that I had forgotten: Fluent's coupled solver. Did you use that in your test? If yes, try it again with the segregated solver. Finally, remember also that Fluent is hardly going to make it easy for you to understand their optimisation ... it is a commercial product after all.

It is a myth that coupled solver is most efficient way to do things. For example just last week we did calculation for drag around car using segreated solver. The mesh was 244 million cells and we ran that thing in less than 1500 iterations (kw turbulence model). I would love to know a coupled solver that pulls it off in less than this iterations (never mind the cost per iteration of coupled solver).

[Tobermory]

Originally Posted by arjun

It is a myth that coupled solver is most efficient way to do things.

I think the point is that there are some circumstances where it can be faster - conditions where the pressure and velocity are tightly coupled, and so where solving equations sequentially can hinder convergence. Absolutely right that it's not guaranteed for every problem .... your low speed car dynamics is one of them. Did you try a turbomachinery case? You might have a different conclusion.