In our company we started to use the combination ICEM-tetra and STAR-CD (version 3.100). Has anyone experience with these codes applied to internal supersonic flows? Maybe it is usefull to start a STAR-CD forum if there are enough people interrested?

(1). You don't have to make an announcement or form a forum in order to ask a CFD related question. (2).My feeling is, it is going to be hard to find internal supersonic flows. (3).Internal supersonic flows will always have complex shock wave systems in it. So, mesh refinement and alignment will be important there. The shock boundary layer interaction will also create some difficult problems.

Maybe I should be more specific. A lot of commercial codes are originally for incompressible flows. To apply for instance STAR-CD to supersonic flows requires a lot of searching to find the optimal set of relaxation parameters and startup procedures. The problem is that a lot of these parameters have nothing to do with the physical problem but rather with internal numerical algorithms. I want to avoid that for each problem I have to send a geometry to the people from STAR-CD after which they come up with a set of parameters that did the job somehow. I know the most of the times it is more useful to exchange experience with people working in the same field of expertise. But You all ready know that. Still I would like to exchange information with experienced STAR-CD users.

(1). If your problem is supersonic, and if you are concerned about the operation using the pressure based formulation, then I would suggest that you look at the density based formulation for supersonic flows. (2). Actually, it is much easier to get converged solution for supersonic flows using the density based formulation. (try to avoid the low subsonic flows when using density based formulation.) (3). So, what I am saying is : the common approach is to select the method or the code based on the Mach number range. Use density based codes for flows in the transonic and supersonic (or hypersonic) range. I am sure that there are commercial codes using density based formulation. But still, the choice is yours.

If you would like to discuss about commercial code like Star-CD, then form a forum for it. Ask Jonas for help. This main forum is for general discussions only. I think there has been enough bickering with commercial code and main discussions. Let us ring down the curtain. Form a seperate forum like other commercial softwares and see how it goes. Good luck and cheers!

Geez, give the guy a break. I'd hoped that the two new forums would reduce some of this irritation. At least they have taken away the majority of the "support related" questions about commercial codes. Please try to be a bit more friendly in your responses. There is no STAR-CD forum today and Bart's question was well formulated and to the point. It wasn't even a "support question", but a more general question related to the suitability of using a pressure based code for supersonic flows. There will probably be a STAR-CD forum soon, but until then STAR-CD discussions, and any other discussion related to codes for which there is no dedicated forum, are of course welcome here.

I'm also interested in this. We are looking a bit at STAR-CD as a complement to Fluent. About supersonic flows - I don't know how well it works in STAR-CD, but I did notice that STAR-CD does not have any absorbing boundary conditions. This can be a big problem for some flows where you have shock-patterns that reach boundaries. They say that the next version will have Riemann based boundary conditions though.

Is there a particular reason why you use both STAR and FLUENT? Before we purchased a license for STAR, I tried to get some information on the comparison between the major commercial CFD codes. Except from salespersons (each code turned out to be the best for my problem?!) it is very hard to get this information.

A density based code would certainly better answer your need for transonic flows. There is only a few commercial code using this technology as it is more recent than the pressure correction method. Actually the density based code technology comes from the aerospace industry and research and allows higher accuracy and faster results.

In addition these methods have been extended to incompressible flows so that one code can run all kind of flow (incompressible- compressible - .low mach to hypersonic).

NUMECA is offering such recent technology in its FINE environments. We are, as far as we know, the only CFD vendor offering a complete CFD solution (grid - solver- visualization) based on density based code. We are today recognized as the fastest commercial code (speed -up of 3 to 10 on all other commercial CFD code) and the most accurate.

Please visit our web site www.numeca.com or contact us directly at bruno@numeca.be (bruno gouverneur.

Sure, density based methods are often better at resolving shocks etc. However, you can't generally say that they are better than pressure based methods. At low mach numbers you often see quite poor convergence with density based methods, even if you use preconditioning techniques etc. to speed up acoustic waves. You can also have these convergence problems if you have well-resolved boundary layers and use advanced turbulence models. There are things that help though, but it is not that easy. Is your code an explicit code or an implicit code?

I have recently done a lot of air-intake computations (full model of a fighter aircraft including air-intakes down to the engine). To start with I used a density based solver and it worked. However, it was a very time-consuming process to tune outlet pressures etc. to obtain desired mass-flows. Later I tried doing the same with a segregated solver and it works much better - you get a reasonable response to boundary condition mods 10 times faster than with the coupled solver. Afterwards when I had a stabilized flow it was easy to switch to the coupled solver in order to better resolve shock-structures etc.

I think both these codes have their advantages. It all depends on your particular application and of course on your personal preference and experience.

I agree that it is very difficult to get any objective benchmarks of different codes. You just have to try it on a few of your own cases before you can say if it works well for you.

What type of applications are you working with?

I am not quite sure what your application is but you may like to take a look at CFD-FASTRAN from CFDRC. This is a commercial density based code that should be suitable for your internal supersonic flows.

CFDRC's web site is at www.cfdrc.com and contains some good examples.

BTW a complete CFD solution (grid - solver- visualization) is available as standard with CFD-FASTRAN. This code has a long history and has been used in a wide range of aero/propulsion and other problems.

CFD-FASTRAN also has a (I believe) unique Chimera grid capability coupled to a 6 dof traj model for multiple moving bodies.

It might be depending on the method, schemes, acceleration technics and specificity of the density base code you use. The experience we have with our code is quite straightforward: it is faster than any other commercial pressure correction codes for compressible and imcompressible flows. For incompressible flow we use indeed preconditioning (Hirsch - Hakimi)methods and we got robust and fast convergnece (our code are progressively replacing commercial pressure correction code in the pump industry because of its speed and accuracy). A german aerospace company recently bought our code for ventillation in space cabin (low speed) as we run between 5 to 10 time faster than the competition. In addition density based code allows to treat very low speed flow within a compressible approach and not only with a Boussinesq approximation. This allows to capture more details in the flow behavior, specially for thermal driven convection problems.

I believe the speed up we obtain is due to the coupling of recent numerical methods with a very effective and strong multigrid implementation. Our code runs by default with multigrid acceleration. We have in development an additional speed up of a factor 3 to 10 due to a recent multigrid progress, which is not accessible by pressure correction code as it is linked a density based formulation.

Our code is explicit but has an implicit formulation with pseudo time stepping for unsteady simulation.

With regard to your experience, we use to work with very fine boundary layer (Y+ = 1) and with aspect ratio up to 5000 and we get very fast fast results.

Which segragate solver do you use and which coupled solver do you use ?

Marc, i think you ought to be careful about making such comments. although i think your code is very "FINE" there are quite a few density based (time iterative) commercial codes out there. CFDRC has one which has very good capabilities particularly for aerostructural applications. they also have chimera overset grids (in addition to structured and traditional tetrahedral unstructured grids). i don't know if they use multigrid (which by the way is the best and most general acceleration method) but their code is quite good. it also includes adaptive meshing which if i'm not certain your code does not. also EXA's powerflow is an unsteady code (although the technique is esoteric). there are a few other time iterative commercial codes out there although i can't quote them off the top of my head. also remember that you have to compete with the 'free' codes offerred by governments such as the codes offered by NASA

When a technical question is asked, it's normally better not to do advertisement here. The debate in the academics about the advantages and disadvantages of pressure-based and density-based methods is not conclusive at all and it has nothing to do with the speed of a CFD code. The AMG and coupled solver are also widely used in the pressure-based method, and the pressure-based method can also be marched in time both implicitly or explicitly. In turbomachinery application, one pressure-based commercial CFD code is widely used with sucesses. For pressure-based method, a lot of applications upto supersonic flow have been made, especially the CFD community from UK.

(1). I found no information about this 5 to 10 times speed up (or 3 to 10 times speed up) on your website. (2). I think, it is important to provide such technical information. What do you think?

I agree than one pressure correction code is used with success, but new technology density based code is certainly offering an alternative today towards more accuracy and speed. Our code has become the best CFD flow solver for Turbomachinery and recognized today as offering the edge in technology and we have enough reference to prove it. Most of the big engine and turbomachinery manufacturer are today turning to our code as we offer significative speed-up, better results and advanced GUI. I would like also to insist that recent multigrid progress have been achieved on density based code and as far as we know this progress can not be used by pressure correction code. We have often run benchmarks and our technology never failed against pressure correction codes but always shows speed-up and more accurate results. I would be pleased to be able to demonstrate this to you.

What about such problems which require true rotor-stator interaction? Your code does not seem to provide moving/sliding mesh capabilities. Have you sucessfully run calculations for side-channel-fan or do you limit turbomachines to axial/radial machines ?

My experience above was from Fluent. They have a very nice implicit density-based coupled solver in their new version, using algebraic multigrid etc. to speed things up. It is also very good at solving turbulent quantities. With the older explicit Runge-Kutta solver they had in Rampant it was often much more difficult to converge for example k and epsilon.

The segregated solver was also Fluent.

None of these are any wonders in terms of speed and accuracy though, but they are very robust.

Are NUMECA's codes based on the same basic models and numerics as the EURANUS family of codes? Any major differences?

Mhh, I'm not sure what I started here. A lot of sales talk from some hidden account managers maybe. The fact is, a lot of people try to convince me to use a different code for my problem (supersonic internal flows). During my PhD I wrote my own code (3D Euler with phase transition), which I considered the best for my application (how arrogant). I'm not looking for a replacement for STAR-CD. This problem is yust one of the many I want to tackle with this package. I'm not going to purchase 5 different packages. This will do yust fine. I'm only looking for people who are using STAR in the same field. I don't want no shit (sorry) about new, specialised codes, with no history, support or interface with other programs (like Pro-Engineer, ICEM, ANSYS or what ever, this is yust what we use)