CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   CFX pressure based (

clifford bradford June 16, 2001 05:54

CFX pressure based
is CFX Tascflow pressure based? Is the same solver used in CFX 5? the CFX web page is not clear.

John C. Chien June 17, 2001 02:40

Re: CFX pressure based
(1). From the user's point of view, what's in the black box is not important. (2). The important thing is whether it can handle the user's problem or not. (3). You can try to read the user's guide to get additional information. I had used the TASCflow quite extensively before, and I still think, the method used is not of interest to me at all. (4). From the mesh point of view, the multi-block structured mesh is the right approach for accuracy (relative to other unstructured code I had used before). But the real solution accuracy is another story. For turbomachinery applications, I think, there is still a long way to go. (5). I can't tell you the source of problems, but I think, it is common to most commercial codes, namely, the mesh arrangement, the solution algorithm, the turbulence model. (6). In terms of convergence rate (if it does converge at all), it is much better than other code I have used. So, the advantage is: better mesh approach, and better convergence rate. After having said that, I must say that it is not uncommon to have the solution diverge in the first few iterations, for some turbulence models. The issue is quite involved, so I would say that it is common to many commercial codes. (7). The advantage of using this code is that it has done some home work exercises in the turbomachinery applications. (and the turbogrid mesh generation code is one good example.) (8). In terms of true solution accuracy, it is not good enough for my applications based on my standard.(like most commercial codes)

clifford bradford June 17, 2001 14:04

Re: CFX pressure based
Yeah John, one of the difficulties i've had with some comercal codes is their reluctance (?) to properly describe their computational scheme. i've read manuals where they make 1st order upwind seem like the best thing since sliced bread.

i will say though that one advantage of some commercial codes is the meshing. What i mean is that some codes give a wide variety of different types of meshes to use. Of course the skill of the user is most important. I do like the multiblock structured codes because I think you mix the advantages of both the structured and unstructured codes.

my experience with commercial codes is nil though so I can only talk from what I've read about them. I also have no experience with pressure based schemes.

Turbomachinery is difficult for the commercial codes mainly because you need special post-processing. Also I think turbomachinery people want a lot of automation so that you can do parametric studies and optimization etc. things are getting better now though. A lot of people depend on commercial codes because they don't have resources to make their own codes. if the code companies are responsive (and the users are knowledgeable) then you can get good results I think. What is your experience?

Question: can you have non-reflecting boundary conditions with pressure based schemes?

John C. Chien June 17, 2001 14:56

Re: CFX pressure based
(1). I don't use the pressure-based approach in my own codes. Are you ready to make that 30-year U-turn into the right direction? I mean, it will take another 30 years for people to realize that it is not the best approach. (2). My experience? I have been laughing here for the last three years. (3). CFD is "war". In the war, many will die. Becuase they are not prepared for it.

Geethakamal,Nallan Chakravarthy June 18, 2001 11:54

Re: CFX pressure based
Hi, I don't have experience with commercial codes. I'd used CFX-4 for a very short period(a week) by verifying incompressible,laminar test cases. I guess CFX-5 is the modified version of CFX-4. I didn't try any turbulence models. I don't think any commercial code is "the solution" to both compressible and incompressible flows. I don't think it can do justice to both in all aspects. I'd worked quite extensively with compressible flows(generating my own code)using various flux-difference and flux-vector splitting and hybrid splitting schemes. I know some softwares which handles incompressible flows accurately but does a poor job with simplest compressible test cases(where shock is weak!),let alone strong shock test cases like blast wave. But they claim that their software can handle compressible test cases. You don't even see where the shock is sitting and in how many zones it's captured. CFD is surely like a war. I think we have a long way to go in terms of developing numerical models to represent actual physical problems. People who use commercial codes should have a fair knowledge of the code's subtleties instead of using it like a blackbox! Vendor's manuals should be clear too about the code. If there is a good rapport between the two maybe we can do better in getting good results.

Dan Williams June 19, 2001 02:02

Re: CFX pressure based
You guys are speculating alot, and I have experience with my own codes and commercial codes, so I thought that I'd point out some facts (most of this information is in the user manuals, or available from tech support, if you don't mind reading/digging a bit):

CFX-5 is not "modified" version of CFX-4. This is a completely false statement and anyone who tells you otherwise does not know what they are talking about. CFX-4 is a cell centered multiblock structured mesh code with a lot of physical models implemented.

CFX-5 is an evolution of CFX-TASCflow. Just like CFX-TASCflow, it uses a finite volume, coupled implicit, pressure based solution technique (i.e., solves for pressure and velocity at the same time in the same A matrix). Pressure and velocity are co-located, so p-v decoupling is dealt with using a Rhie-Chow approach (with alot of experience thrown in to help ensure convergence and robustness).

Strong conservation is ensured on all control volumes, as it is a finite volume code after all. Algebraic multigrid (AMG) with an ILU smoother is used to accelerate the linear solution. Other than a different implementation of the discretisation for advection, which it has to be to also allow unstructured grids, CFX-5 is essentially identical to TASCflow in it's basic approach. In some cases, especially if there a many grid blocks, CFX-5 might even do better because CFX-TASCflow cannot agglomerate control volumes (for AMG) across the grid blocks. If you are wondering where all this came from simply look for CFD papers in the literature from people working at the University of Waterloo on AMG (there are many of them).

I've mentioned before that CFX-5 captures complex shock structures in ~3 "zones" as you call them, and contact discontinuities in ~4-6 "zones". I claim this and I will gladly send all of you pictures if you like. What's nice is it does this as well as the more traditional density based schemes such as FCT, ENO, PPM, MUSCL, Roe's scheme, LeVeque, etc... (there are so many these days I've lost count, or stopped counting as the case might be). Although I've never tried it, CFX-5 would probably also do blast waves as well. As long as you ran a timestep which decently resolves the wave propagation I don't see why there would be a big problem. You can always do better on a "particular" problem if you write a specifically tailored code. This is not good business because CFD vendors have to appeal to a variety of clients, so they need to be fairly general. Pressure based schemes are simply more generalisable and that's one of the reasons that most vendors use them.


joseph June 19, 2001 07:42

Re: CFX pressure based
hai dan,

are you saying that if i solve compressor problem in tascflow ,then I use cfx5 then will I get the same results?

regards , joseph

John C. Chien June 19, 2001 13:08

Re: CFX pressure based
(1). I don't know. But I think, it's unlikely. (2). I guess, a few years back, there were two companies and there were two codes, CFX and TASCflow. (3). Then one company disappered and the code became CFX-TASCflow. And there were still two codes, CFX-4 and CFX-TASCflow. (4). Then the TASCflow technology was used,along with un-structured mesh, in a new code. Since there was one company left, they called it CFX-5. (5). Now, there are CFX-4, CFX-TASCflow, and TASCflow derived technology code called CFX-5. I think, it would be easier to understand if they use the name something like "CFX-TASCflow-5".

clifford bradford June 19, 2001 19:31

Re: CFX pressure based
Hmmm...why is it then you think that the pressure based schemes are so popular in commercial codes? My experience with density based schemes is that if you can't get convergence that you can find out why not (even if you're not then sure how to fix it) but with pressure based codes I've seen experienced people flumoxed. the problem with that is that most people who use CFD in the industrial world use commercial codes. I wonder how they get by. I've noticed that a lot of the commercial companies are trying to move into areas that traditionally have been "ruled" by density based schemes and a lot of people are taking them seriously and buying their programs.

Perhaps this is not the best forum ;-) for this discussion though.

How do you like using CFX relative to other commercial CFD codes? what do you like/dislike about it?

Robin Steed June 19, 2001 19:46

Re: CFX pressure based

If you import your TASCflow mesh into CFX-5 and apply the same boundary conditions, turbulence models, discretization etc. you will get the same result.


John C. Chien June 19, 2001 20:03

Re: CFX pressure based
(1). The market is free market, I hope. (2). If the buyers are happy with the commercial codes, then they must be very happy to ask commercial codes related questions here. And this is also a free forum. (3). So people said that it is not good to watch TV or use video games because something bad can happen. (4). The forum is a place like a free market, where different opinions are presented. (5). It is against the principle to say "this is the right approach or the right answer", even though there is something called the "truth". (6). I am always amazed that there were six million jewish people killed in world war II. Popular is relative. And In the 19th century, the potential flow theory was very popular. When they couldn't predict the drag, they simply call it a "paradox". (7). So, shall we say that it is a paradox. (8). To buy or not to buy, it is the question. In the movie of the "Schlesinger's List"(?), he was very sorry to cry that he could have bought more jewish people, but he didn't. (9). I think, human brain and thinking are not perfect. The realization of this is very important. (not the realization of what is the truth.) (10). So, even if there are six million copies sold, it is still hard to say that it is a popular code. And there was a time when almost everyone in China has a copy of Mao's small red book. (11). The important thing is: CFD is War, and to survive, you must have your "Own" thinking. (you must be able to think, independently, like mesh independent solution.) And if you don't have your own opinion about when the solution is converged solution or mesh independent solution, then there is a likelyhood that sooner or later you will join the list of the dead people.(even if your body is still moving) (12). Awareness is the key issue, not the true solution or approach. For that, you will have to make your own decision, becuase the analysis is going to be "your analysis".

John C. Chien June 19, 2001 20:11

Re: CFX pressure based
(1). Well, that can happen, only when CFX-5 is identical to the CFX-TASCflow. (2). More or less like the new packaging of the old soda. Could someone check it out with a simple test case? I don't have the right answer.

Dan Williams June 19, 2001 23:03

Re: CFX pressure based
Yes, you should get almost exactly the same results.


All times are GMT -4. The time now is 22:56.