The proposed paper deals with modelling of turbomachinery performance while using advanced CFD methods based on either 3-D CFD or 1-D CFD. The main focus is put on turbine (radial one, which is typically applied in automotive turbochargers), however topic of compressors is also considered. Different turbomachinery was available for both experiments and simulations. This concerns classical fixed geometry radial turbine, variable geometry turbine (VGT) and twin-scroll turbine. Regarding the compressors, classical automotive one was available. The presented work was supported by different turbomachinery manufacturers, hence all necessary data were available to create detailed 3-D CFD models of the above-mentioned turbomachinery. Some of them were also converted into 1-D CFD models as well. The proposed paper concerns mainly the modelling part of the work, hence the experimental data are used as reference ones. All the 3-D turbomachinery models were calculated in CFD SW tool AVL FIRE and the results were compared with reference data. Sensitivity studies of selected parameters were carried out (e.g., mesh parameters, turbulence models) – this also includes different calculation set-ups in terms of rotor modelling (fully movable rotor, frozen rotor approach). Regarding 1-D CFD turbomachinery models, they need a lot of calibration using either experimental data or 3-D CFD ones to match mass-flow/efficiency performance. However, they are much faster than 3-D CFD ones, hence they can be applied in 0-D/1-D CFD codes (e.g., GT-Power) to run the complete engine model simulations. As they are based on 1-D CFD, they are much more predictive than standard models based on steady-state maps. This brings significant improvement in calculation accuracy under transient operating conditions or for un-conventional turbomachinery design(s) (e.g., twin-scroll turbine). The main conclusions are the following: • 3-D CFD: o Calculations are still very time demanding – this also concerns steady-state operation. o Such models can usually predict mass flow rate very well. Concerning the efficiency, the prediction is at least qualitatively correct. o It is recommended to apply fully movable rotor approach, however it is clearly slower than frozen rotor approach. o Concerning the turbulence modelling, advanced RANS models (based on more than 2 turbulence transport equations) are sufficient as a correct prediction of boundary lawyer is important. It seems that advanced approaches (LES, PANS) do not bring any significant advantage. • 1-D CFD: o It needs a lot of calibration effort. Either experimental data or 3-D CFD ones are sufficient for that. o It is sufficient to calibrate the model under steady-state conditions (turbomachinery test bed). Due to its predictive ability based on 1-D approach, the model works fine under transient operation (e.g., pulsations under ICE steady operation, ICE transients) without any need to additional calibration. o It can match data from 3-D CFD under transient operation, however there is still some room for improvement. o Such models are ready for application in system level modelling (0-D/1-D tools), however they are significantly slower than the standard ones (based on maps).

Dr. Oldrich Vitek, Czech Technical University in Prague, CZECH REPUBLIC Prof. Jan Macek, Czech Technical University in Prague, CZECH REPUBLIC Dr. Zdenek Zak, Czech Technical University in Prague, CZECH REPUBLIC Dr. Vit Dolecek, Czech Technical University in Prague, CZECH REPUBLIC Dr. Radek Tichanek, Czech Technical University in Prague, CZECH REPUBLIC Dr. Bohumil Mares, Czech Technical University in Prague, CZECH REPUBLIC