CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Opinion about CFX4 and CFX5 (

Jan Rusås August 6, 2002 06:59

Opinion about CFX4 and CFX5
Dear all,

I have used the structured solver from CFX (now using 4.4) for many years. I have tried on occasions to switch to CFX5.* but I tend to return to the CFX4 for many reasons. I always know how to solve the problem in CFX4.4 (major issue)and I have very often found that I could not solve the problem in CFX5.*, either the right two-phase model was not available, or the appropriate combustion model was not available or even radiation, or I could not include thin surfaces they way I wanted it, etc.

I understand that CFX5.5 have many advantages compared to CFX4. , speed, coupled solver, and unstructured solver etc and I can see on this CFX forum that many people now use CFX5.

Would those of you that have experience with both CFX4.* and CFX5.* share your experience if it is worth trying to switch again to CFX5.5. Preferably if you have experience in using CFX in many fields, combustion, two-phase flow, isothermal flow in complicated structures, chemical reactions, mixing etc., but any opinion will be appreciated

Regards Jan

Dave August 6, 2002 08:43

Re: Opinion about CFX4 and CFX5
We use CFX5 in complex geometry and to be honest there is know way on earth I'd go back to CFX4. The model construction time is a fraction of what it used to be. However as a result I now spend time putting more complex geometry into my models rather than spending the time generating a better mesh (which is what I really should do!!). Where 5 falls down a little is in the data extraction. CFX4 had the user fortran routines which were well doccumented and fairly easy to construct. CFX5 now has the user fortran but this is still in its infancy and the doccumentation is by no means as clear as that of CFX4. Saying that the guys at AeA are more than helpful and can usually provide an example of what you are trying to accomplish to can get something similar together. So from a model creation point of view I don't think its an issue. Other peoples comments on the physics would be interesting ??? Dave

Robin August 6, 2002 15:12

Re: Opinion about CFX4 and CFX5
Hi Dave,

For data extraction, there is no need to use user Fortran in CFX-5. Rather, CFX-Post provides the computational facilities for extracting data.

What data have you not been able to extract using CFX-Post?


Dave August 7, 2002 04:14

Re: Opinion about CFX4 and CFX5
Robin, One are we look at is ventilation times within volumes. This is something we have found best monitored using the CFX fortran. Don't get me wrong, I wasn't putting post down. I find post very good (quantum leap over visualise view) and is excellent in its amount of user manipulation of data. May be its because I'm still only just scratching the surface of Post at the moment or may be my mind set is still turned on to the old methods of data extraction used in CFX4 ie writing a little fortran routine to extract the info to a file ??? I'm not sure. I have started to use the power syntax and have found that very usefull (and again I'm still only using it in a most basic way). Either way I think CFX5 has a great deal going for it and has great user functionalitly and I hope the guys at CFX keep up this present rate of development of the code. Dave

Robin August 7, 2002 08:30

Re: Opinion about CFX4 and CFX5
Hi Dave,

By ventilation times, do you mean the "Age of Fluid"?


Dave August 8, 2002 14:02

Re: Opinion about CFX4 and CFX5
Robin, yes, I've spoken to the guys at CFX who have been very helpful, we have tried several approaches to the problem, one using a steady state setup and scalar sources in all of the cells which we then used to calculate another user variable defined as time, relating the source build up to the time. The second approach was to fill a volume of interest with a scalar and then advect this within the model using a steady state results field. This works well (still need to do some more detailed comparisons between the 2 methods), however we needed some help from the CFX guys in setting up some fortran to automatically stop the transient simulation once the ventilation had been completed. You mentioned the age of air, did you have anything in mind ??? Dave

Robin August 8, 2002 21:09

Re: Opinion about CFX4 and CFX5
Hi Dave,

I was going to suggest the source method, but it appears you have used this already. This will give the best representation of the age of the fluid.

So what you wish to do is run a transient simulation until sufficient air exchange is acheived? i.e. define a custom convergence criteria?


Neale August 9, 2002 13:10

Re: Opinion about CFX4 and CFX5
I agree with you about user fortran in CFX-5.x It really is still in it's infancy, and the only real issue is documentation.

The CFX-4 approach was to provide hooks into the code at various points in the solver for properties, boundary conditions, initial conditions, etc... A large percentage of what you would use user fortran for in CFX-4 has been replaced by the expression language in CFX-5, which is quite flexible. And if the expression language cannot do what you need, then it is possible to write user defined fortran which you can use in the expressions. CFX-5 does have junction boxes, which can be used for various things, but you will need the help of your support people to use them. This is because documentation and training on the internal flow solver structures would be necessary for you to use them on your own.

User fortran in CFX-5 will evolve, but I think the plan is not to make it the first option to use. Most of the time in CFX-5 there is no need for user fortran because they have the expression language, which is quite flexible.


Neale August 9, 2002 13:14

Re: Opinion about CFX4 and CFX5

I'm not sure when the last time you looked at CFX-5 was, but you should check the website for information on the latest release of CFX-5, CFX-5.5.1. Many new physical models are available, including multiphase, combustion, radiation (optically thick), and so on...

As far as including thin surfaces, CFX-5 should support anything you do in CFX-4. Just build your grid in CFX-4 and CFX-5 will run it. I'm suprised you had problems.


Elteyeb M. Eljack August 10, 2002 07:13

Re: Opinion about CFX4 and CFX5
I used CFX 4.3 for long time for gas/oil combustion and two phase flow as for water bubble then I used CFX 5 release ... when I started to use CFX 5 I feel that it is complicated and not simple as 4.3 in geomitry and postprocess ... I think that 4.3 & 4.4 release are easy to work on ... 'cause of: 1. the interface of the pre/post process is clear enough to follow in 4.3/4.4 2. command file setting is easy in 4.3/4.4 3. I used to solve problem for combustion in two stage ... cold flow(for u,v,w,p,k,eps) then hot flow for T ... but I can't do that for 5.5 release. 4. while the solver running it plots the residuals on graph and there is five option for the graph ... I found them helpfull to guess if the problem going to convergence or not ... but no such options in CFX 5.5 or I can't use it yet.

I desscused this matter with other college they said that may be I used to work with CFX 4.3 so I can't go on with CFX 5.5, this really happen to me when I start to use Fluent.

Anyway AEA said that CFX 5.5 release is powerfull with physical models .... but I like to work on CFX 4.3/4.4 with more models especially for liquid fuel combustion models .... I feel very sad 'cause CFX 4.3 is going to be a history of CFX release. I hope AEA can devolp this release with it's old interface.

Regards Elteyeb

Dave August 12, 2002 05:42

Re: Opinion about CFX4 and CFX5
Hi Robin, yes, you've hit the nail on the head. We did this using the fortran. We had some teething problems, which is to be expected, but we have it working now. It just took a little time to set up and wasn't quite as easy as the old CFX4 style fortran interface with the code. I'm sure its just a time and experience thing, once I'm more confident with implamenting the fortran in CFX5 then it shouldn't be a problem. Like all things new, the best way to learn is to see examples and then take things from there. Dave

Jon Castro August 13, 2002 17:47

Re: Opinion about CFX4 and CFX5
I've been using CFX 4 for the last 6 years and have yet to use CFX 5 on a project for a client.

Having used CFX 5 before on various trial problems, the biggest difficulties I encountered were controlling the meshing and lack of user fortran for specifying boundary conditions and extracting data.

I still use my own Fortran coding (occasionally with the old Grid-gen program!!) for meshing in CFX 4. This is mainly due to the problems I've had with Build when creating large geometries with numerous constraint lines, as these split the geometry into hundreds of blocks. In addition, it is not easy to change the geometry without having to delete solids, surfaces etc and re-create them. The command file is also very handy for defining arbitrary user patches in the mesh, as to do this in Build requires the use of constraint lines whenever one wishes to define a patch in a particular location (particular cells in the mesh cannot be selected).

In summary - meshing with fortran and using the command file, in addition to the use of user fortran routines, enables much more of a "hands-on" approach and more flexibility. I would only consider CFX 5 in the event of having to define geometries having curved surfaces which would otherwise be difficult to define. This is where CFX 5 comes in handy for turbomachinery problems, for example.


Dave August 14, 2002 10:05

Re: Opinion about CFX4 and CFX5
Hi Jon, what sorts of models do you use grid gen for ? Is it not simpler to use build v5.5 ?? Or do you need to include a lot of fortran within your models ? Dave

Jon Castro August 14, 2002 10:16

Re: Opinion about CFX4 and CFX5

We use quite a lot of fortran - USRGRD, USRSRC, USRBCS, USRBF, USRTRN, USRINT and USRWTM mainly. The grids I've been developing over the last few years are mainly rectangular with a lot of constraint lines. I also tend to end up adding additional user patches in the command file which are approximated in the grid, as defining constraint lines for everything is often too labour intensive. Is it possible to specify patches on specific cells in an unstructured CFX 5 grid (without using constraint lines)?


Jon Castro August 14, 2002 10:37

Re: Opinion about CFX4 and CFX5
My comments on CFX 5 were probably more applicable to CFX 5.3, as the latest version I am informed now has user fortran available. I will have to try it out sometime.


Neale August 14, 2002 16:52

Re: Opinion about CFX4 and CFX5

Since CFX-5.3, CFX-5 has always had the expression language. Since 5.4 it can be used in a solver command file based on the new CFX Command Language (CCL). In most cases, the CEL eliminates the need for user fortan.

User fortan in CFX-5 consists of two parts: 1) Extension of CEL to handle user defined fortran functions, and 2) Junction box locations at several points in the solver startup, run loop, and back end. 1) is useful if the CEL does not provide you with what you need for boundary conditions, initial conditions, source terms, or properites. 2) can be used for more advanced things, but you will need help from support because the internal data structures of the solver are not yet well documented. User fortran in CFX-5 is really still in it's infancy, and will continue to progress, however in many cases there is little need to use it.


Dave August 15, 2002 09:36

Re: Opinion about CFX4 and CFX5
Jon Just curious though in CFX5 you don't use constrain lines ? CFX4 build is all constraint lines, but the concept in cfx5 is a little different. Or an I missunderstanding something ? Dave

Jon Castro August 17, 2002 16:28

Re: Opinion about CFX4 and CFX5
Although CFX 5 is used mostly for unstructured meshing and doesn't require you to define constraint lines, you would still define any internal structures you wanted in the geometry.

I don't use CFX 5 because I nearly always only use structured meshes for my problems. The project I'm currently on requires the use of many constraint lines in the Z-direction, so I find it easier to develop 2-D grids and then create the 3-D grid in Fortran.


All times are GMT -4. The time now is 22:51.