Hi all,

just for fun....go over to the CFX forum and ask people how they like the documentation for the user Fortran in CFX-5. Fluent has an entire manual and many examples. It is hard to document stuff these things....believe me you are better off with Fluent in this instance...believe me. ;-)

Bak_Flow

Hi,

yes there are cases that Fluent is not so robust but that can be said of any other solver with the same technology.

However one would have to do a 1-1 comparison to be fair...which I have done. People usually use more complex models in Fluent for combustion and radiation for example. CFX does not have some of these models. If you run 1 step mixed is burned reaction with P1 radiation in both codes...CFX will do better. If you want the right answer say it takes EDC and DO radiation....CFX does not have the models????

Most people are doing more than flow in a duct these days...fluent simply has more models!

That being said I hear Fluent is going to introduce some new solver technology including a pressure based coupled solver and some faster density based solvers for high Mach flows...that should prove interesting!

Regards,

Bak_Flow

Hey Neale,

sounds like a bit of an advert!

So a bit of the stuff in CFX is bad....but keeps on improving??

Feature matrix is far far far fuller??

Ok just to level the playing field.....

Would you call a code that could not do profile boundary conditions till its 7th release, still can't do transient lagrangian particles, adaption in one fluid zone, periodic elements with DP, DT, and many others a full feature matrix??

Why is Fluent so successful....it is not about the height and breadth of the feature matrix or a nice look and feel....it is about getting the features people really need working!

Regards,

Bak_Flow

More models doesn't mean anything, if those models don't work or their solution is not accurate.

Any user with a good CFD background, is able to implement a model, the problems come when you have to solve it an accurate way.

A robust solver in the hands of expert people is a good instrument, a good set of model with a poor solver is not.

Regards, ap

well i used fortran with cfx and i never found any problem, and in the documents its well explained how to use, it infact gives more freedom control than udfs (i can't say much though, since udf is one thing i could not make much out of it, thanks to fluent's docs), the irony is i do lot of programming with c/c++ where as my knowledge of fortran is limited.

another part of hard to doc these things:

in my opinion the arguement that some thing is bad with another software does not mean that this software should not be improved. so even if we assume that cfx docs are bad and cfx is worse compared to fluent. it doesn't mean that fluent should not improve things. fluent is such a big company if they wish they can improve the quality of their docs, otherwise how it happens that your fluent support knows lot of things that you can not learn from documentation. who tells them and where do they learn from. they must learn from some documents or training. so if such things exist in docs or training, why they are not documented for normal users like you and me. i guess the license fee is pretty hefty for fluent, can't we expect even this. and when i say cfx docs are better i mean in this regard that they have tried to put as much as possible into docs. for example fluent tutorial has around 15 examples max, compared to cfx4.4 it has more than 80 examples (even more may be i am roughly saying) , and cfx4.4 is almost outdated compared to cfx5 series. the question is can you summerise you vast featured fluent solver in just 15 tutorials. if there are features we as users like to learn. and yes, we (at least me) are not so intellegent as fluent might wish we shall be.

Hi I was talking about CFX-5.

CFX4 is a dead code now. I agree the user Fortran there was useful. CFX-5 is a whole other story.

Regards,

Bak_Flow

yes, cfx5 is entirely new thing and i can not say anythign about it,

but i shall honestly admit, i never faced any convergence problem with fluent or cfx (except for EDC which gives cold flow after convergence), and i strongly believe convergence is mostly in the hand of user, if he makes grid properly and apply model properly things can go easily. so if one understands cfd he can work with naything, there are people who get even better results with their own codes. And software comparision is very debatable issue, in the end it boils down to person likes (sometimes we just like things)

Why FLUENT doesn't give a full documentation?

The answer is easy. FLUENT offers consultancy services too ;-)

Yes ap, and when you try to ask them something about scheme or UDF they say that service is not in your support package.Luca

I know that answer very well Luca,

"UDF are not included in your support package. But you can always buy a package of support hours or attend a course"

This is what was answered to me. Not a good answer, considering FLUENT cost.

I'm new to Fluent, could zxaar please explain what: "execute a command on tui" means?

(from posting: zxaar , Thu, 28 Apr 2005, 3:13 a.m)

Thank you.

Well, I guess the best answer to solve all these problems in time, is to wait for a FREE GPL implementation of CFD code. I frankly do not see that happening anywhere in the near future, although I do believe that this will definitely happen.

What about OpenFOAM? It is exactly what you want.

http://www.opencfd.co.uk/index.html

see if you facing negative volumes after creating interface you can try this:

before making interfaces execute this command on TUI (text user interface or fluents command prompt)

(rpsetvar 'nonconformal/cell-faces 0)

after this if you define the interface the mesh check usually do not fail, at least in my case has not failed so far.

Did not know about it. Neat

Hi,

in vertex based schemes the flow variables are stored at the vertices of the mesh elements.

In cell centered schemes the flow variables are stored at the centers of the mesh elements.

In the first, control volumes, for which the conservation equations are satisfied are built up from pieces (quadrants for quad meshes in 2D, for example) around the vertex. In the end these control volumes form a mesh as well which is called a dual mesh which has the same fundamental purpose as in the cell centered scheme.

For quad/hex meshes there is very little difference between them except a bit at the boundaries. For tri/tet meshes there is a huge difference in the resulting number or degrees of freedom. Usually a factor between 3 and 4 less for the vertex based method. This makes codes hard to compare with the same mesh.

Good discussion in Blazek's book...see book section of CFD-Online!

Regards,

Bak_Flow

I have used both codes, so I was just trying to give an objective opinion rather than show a bias one way or another. There is no question that both are great products and there is no question that Fluent and CFX both have strengths and weaknesses.

I guess I did not clearly explain what I mean by feature matrix as you have misunderstood.

What you have mentioned about the various features that exist in Fluent and not in CFX is simply a feature "tick box". i.e. Why does CFX not have feature A, feature B, feature C, feature D, etc... (you give examples of transient particle tracking, periodic BCs with delta P, mass flow, etc..). For sure CFX also has useful features, which Fluent does not have, that both work well and that people really need, otherwise no one would use CFX at all. However, that being said, I don't want to get in a tick box war so I immediately concede... CFX clearly looses the "tick box" battle.

Rather than a "tick box", what I meant is "feature interaction". eg: feature A is great but does not work with feature B, feature C, feature D, etc... An example in Fluent of this is the periodic condition you mentioned where you specifiy the mass flow rate. This feature does not work with combustion and only works in the segregated solver. There are many other examples like this in Fluent.

If you pick any one feature that both Fluent and CFX have, you would be able to run it with more models activated in CFX than you would with Fluent. It's just the way it is.

I think CFX had profile BCs since 5.3, which was the second release. How one looks at this may depend on your definition of a profile BC though.

Neale

There is still quite a difference in hex meshes, since in a vertex scheme, integration points are located on each sub-face (the faces of each element sector around a node), whereas a cell-centered method only has integration points at the cell faces. For a hex mesh, this means CFX has 24 integration points per control volume, as opposed to 6 integration points with Fluent.

-Guy

Hi Neale,

yes I agree with your point. A massive level of interations between models is nice. I am however an Engineer....somone brings me a problem to solve if there is one thing that is missing in the code for what I need....I go to the next code or figure out how to add it myself. All I am saying is that my experience has been that there is a lot missing in CFX-5...except if you want to do turbo work!

So taking your example, periodic element sections with combustion....

1. Does anybody ever need this...i am not sure...I have seen flames through tube-banks but I think the properties are varying so strongly with temp and species variation that the flow may not be periodic??

2. What would be the Engineering solution in CFX 5?

Not to put too fine of a point on it...but in versions prior to 5.7 there was cloud of points interpolaton which one could use if:

1. you did not care to get the correct profile

2. you only had a few points because it was tedious to cut-up your variables into separate files and arduous to load

come-on...........it was a pathetic!

Looked like a "tick box" feature to me! ;-) I don't know that feature would be the correct term for detailed boundary conditions. Hey there is an idea for marketing people around the world...start listing all the boundary conditions options, fluids, etc.

Regards,

Bak_Flow

Yep,

but I was not really addressing the Fluent vs CFX implementations just the guy's question about centered and vertex schemes in general. ;-)

On that note there is nothing preventing a Cell centered scheme from using more integration points than 1 on the faces. Formally, 1 integration point/face and 4 are both 2nd order accurate on regular grids. In practice I believe more is better but I have not read any good studies on this.

Bak_Flow