I can make any complex grid I like, using CFX-5.4.1. And I am very pleased with the Meshcontrol options. OK, sometimes strange things occur but these are always solved, never a dead end... There are many routes to Rome.

Gert-Jan

Hi Robin,

CFX 5.4.1 is a nice solver, but I very often obtain the meassage FINMES. Could you elaborate on this message? What does FINMES stand for? Is it an abbreviation?

Astrid

As I said, ProAM was not available. As I understand it, ProAM is the convergence of ProSTAR and SAMM. Unless you paid extra for SAMM you did not have it.

Robin

FINMES is a linear solver error that can occur when you use autotimestep. If you half your timestep through physical timestep, thisn should converge.

I am actually surprised about these comments. Clearly one can have different views about different codes particularly influenced by one's employers. However what is surprising is that the very basics are totally incorrect. For example the claim that STAR is based on staggered grids is indication that the writter does not even know the most basics about STAR-CD. From ther firsdt version of STAR in 1987 it did not use staggered grids. In fact STAR was the first general purpose unstructured CFD code in the market which clearly used co-located cells. Any way I don't think I need to go on!

Hi star-cd user,

(1) I stand corrected.

(2) Do you use AMG?

(3) How robust is MARS when solving on a tetrahedral grid? Do you get the same results?

(4) No doubt it is popular with Star users, do you have another option?

Robin

Howdy,

I know nothing about STAR myself.

But, if were getting into a bit of a contest about who had the "first commercially available co-located code", CFX-TASCflow has been co-located, and using coupled AMG since 1984 or 1985 (somewhere in there anyways, but I don't want to split hairs). So, CFX was the first.

Dan.

(1) Wrong. CFX was the first. See my other post below.

(2) CFX-TASCflow has been using coupled AMG since the beginning (1985). Other options were available at one time I believe, but the coupled solver is so much superior. Note that CFX-5 is based on CFX-TASCflow technology.

(3) The High resolutions scheme does this in CFX-5 without the fancy acronyms. These non-linear advection schemes are a dime a dozen now days, and they all essentially doing the same thing. CFX-5's scheme is basically a modified version of MUSCL (by vanLeer). I've never heard of MARS, what is it's basis? I saw a recently STAR brochure with a shock tube calculation. The result looked pretty good, but that calculation has been beaten to death so many times by 1000's of researchers worldwide. STAR, it's just no big deal.

(4) User coding is available in CFX-5.5. It is possilble to do many things, and you don't have to recompile the solver. Everything is handled with shared libraries. Note that user source code has been in CFX-TASCflow and CFX-4 for many years. This has been an especially important feature for CFX-4 users, and it will be important for CFX-4 users who decide to switch to CFX-5. I guess the main point is that in CFX-5 it is usually not necessary to use user subroutines. Quite a bit of powerful functionality is available without writing your own source code.

Dan.

Sorry for you but TASCFLOW is multiblock-structured and not unstructured.

Using autotimestepping with 'tmax' will give you an increasing timestep. Then your solution is not so straightforward.

FINMES can also occur using physical timestepping. Same solution, halving the timestep?

Astrid

OK fine, add some more details to turn it around. If I recall correctly, we were only discussing which comercial codes were co-located. Nothing was mentioned about what types of grids could be run.

Dan.

1. Does this mean that you understood the correction? 2. Yes 3. MARS is the truly bounded second order scheme regardless of mesh type and quality 4. STAR-CD has other options as well as a flexible structure to allow user coding

Hi Star-CD user,

Yes. I understand the correction. I apologize for this error, it was not my intention to mislead anyone.

Best regards, Robin

As a former employee of adapco, I know that they used to offer university licenses at a far cheaper cost (but you will also receive less support). Also, adapco did offer trial periods (usually around 1 month if I'm not mistaken) where you can test the code personally to see if you like it. I suggest that you contact a local sales engineer for more information to see if these policies are still in effect. I'm not sure if AEA Technology has some similar policies.

I definitely think that it would be in your best interest to demo both of the codes for your given application.