Gentlemen,

I would like to choose a CFD package to simulate simulaneously the flow from inlet, to vane, to turbine rotor, to diffuser, and to volute, in our turbocharger. Does anybody have any such experience? Perhaps somebody can give me a suggestion to choose the most appropriate CFD package to do this work. Thanks,

John

Hello,

You should have a look at CFX tascflow from AEA, they are the market leaders for turbo machinery.

Regrads

David

(1). It's a very aggressive goal. (2).That's the future goal of every turbomachinery company. (3). Every piece of the component you mentioned is currently a research subject. Just to mention one: a separated diffuser flow with high inlet swirl. (4). I have used Fluent, TASCflow and several other codes extensively in turbomachinery for the last several years. My experience is: these codes can handle turbomachinery components, but there are problems in convergence and accuracy of the results. (5). From my point of view, as a long time promoter of CFD, your participation in trying out these apporaches (meshing, solver, pre-, post-processing) will keep the field moving. Thus improve the chances of reaching the goal. (6). But don't bet your life on it. Not even your job, unless you are paid to use the codes.

Hi John

I suggest you to look at NUMECA's software (IGG/Autogrid, FINE/Turbo, cfview) with the turbomachinery tools available in the Turbo version.

Kind regards Arrigo

As JCC points out this is a *very* ambitious target to try to achieve from (what sounds like) a standing start. Even those of us who use CFD in radial turbomachinery every day and have years of experience would think twice (thrice?) about attempting such a large computation all in one go. Of course, the commercial code vendors will tell you that you can achieve anything with their packages, whether you belive them or not is up to you. A more workable approach would be to do the analysis on a component by component basis. This will allow you to build up experience and isolate problems which may (will!) arise.

John,

I've been involved with this type of calculation. It can be a lot of work, especially the meshing of all these components. What are you hoping to get out of the solution? Can this be obtained by running only some of the elements?

One of the fundamental questions to ask when considering what CFD package to use is how the package will handle the coupling between the static elements (vanes, diffuser, volute) and the rotating elements (the turbine rotor). An unsteady solution would probably too expensive for you, so the CFD package must do some averaging for you between the rotating and non-rotating elements. Make sure this averaging does not wash out whatever you are looking for in the results. Each CFD package approaches this somewhat differently.

With so many components involved, I would consider first to obtain, in a rather straightforwrd manner, an overall picture of the global flow field, i.e. why not running a throughflow/axisymmetric/meridional computation first?

In such codes, the blade are modeled to turn the flow. Furthermore, the meridional computation will also be of help for starting your big 3D steady(unsteady?) computation. Believe me, with several components involved, it is far from evident to get a good initial solution.

I think that Numeca has a throughflow module in its code based on a time-marching finite volume method, i.e. suitable for compressible flows.

Regards,

I would recommend that you consider Numeca Inc.'s Fine/Turbo code. I have been using this code for multistage analysis of both high and low pressure turbines and I am very pleased with the results. Thanks to its AutoGrid utility it is a matter of an hour or less to come up with a decent, orthogonal, 3D, periodic mesh around multiple blade rows. Also, the convergence is very robust and fast thanks to the multigrid accelaration. I have carried out a comparative study where I compared the results from Fine/Turbo to results from a viscous, 3D, government code and found out that the convergence rate and levels of residuals were much better with Fine/Turbo and the accuracy was as good, if not better. Also, I have simulated some test cases from the literature and found out that the loadings are in excellent agreement with the data. I am also aware of a study where Fine/Turbo was evaluated against some other commercial and government codes and it was found to be more suitable for multi blade row analysis. Finally, their consulting engineers are highly knowledgeable in CFD and are very helpful. You can find out more about the product from the consulting engineer that I consult to at ALGrosvenor@az.rmci.net.

Good luck and best wishes.

Hakan Aksoy, Ph.D.

(1). I think, the question was about the simulation of the whole turbocharger (radial machine I guess), not just the multi-stage flow of axial machines. (2). For your study of multi-stage flow, what were the typical mesh size used for the whole problem (number of stages and the total number of grid points or cells) ? What was the turbulence model used in your study?