We are looking at CFD packages for analysing radial turbomachines, turbine and compressor (i.e. rotor + radial stator vanes + volute/housing), so we need a code that can mesh blade rows and 3D housings as quickly as possible, as well as having robust solver methodologies. TASCflow, FINE/Turbo and FLUENT are in the running - are there any other codes that we should be giving serious consideration to?

(1). The radial turbomachinery is quite different from the axial turbomachinery. The complete geometry is highly complex, and the 3-D flow field always has flow separation in it. (2). The relatively thin leading edge always creates 3-D flow (with separation at off-design conditions, and even at the optimum design condition), not mentioning the complex 3-D scroll, 3-D volute flows. (3). Whether the current turbulence models can handle it or not is unknown. Both Fluent and CFX-tascflow can handle this types of flow problems, but the results depends strongly on the users experience. I have used both codes, I think both are useful. What you need is more experience with the code and the radial machines. Codes alone will not solve your problem, not even a starting geometry.

We presented a sample problem for a number of the commercial based codes to solve and based our selection on that. The willingness (or lack) of the individual CFD code suppliers, should help your choice, too. You might add CFDRC to your list, as well as others, but their results on your sample problem might help your choice. Several of the companies backed out after investigating our problem... and only one of the group tested was close to laboratory results - we bought this code after a trial period.

Which code was the best in your case ?

Numeca's Fine Turbo was not in our competition at that time, so the closest for our application was TASCflow. It was about 4 years ago and I had not heard of them (Numeca)...

So far we are pleased with TASCflow, but no commercial code is perfect. Would be nice to have several, but ...

We have an ideal setup in that our labs can verify/challenge our CFD results, which help with modeling practices, etc.

it's a question of ease of problem setup (geom import, mesh gen, post processing, availability of moving boundaries etc) specific to turbomachinery as much as accuracy. a code with turbomachinery specific features (i think CFX has a turbomachinery specific version as does numeca)-look in turbomachinery magazines, would take away a lot of headache. erich also has a good idea give your candidates a test problem (have them use your CAD files etc)and make them put up or shut up. also make sure they can customise their software to your computers. a lot of these companies want you to pay big $ to use their program then tell you to get new computers to run it. go for the company that going to give you accurate, usable software and is going to kiss your ass too! it's your money