Skip to main content
To KTH's start page To KTH's start page

Aerodynamic Design and Aeromechanical Analysis of Mixed and Radial Flow Turbines

A study on meanline method, stator tilting endwall design and forced response analysis

Time: Mon 2021-09-27 09.00

Location:, Stockholm (English)

Subject area: Energy Technology

Doctoral student: Yang Gao , Energiteknik, Turbomachinery & Propulsion

Opponent: Prof. Damian Vogt, Inst of Thermal Turbomachinery and Machinery Laboratory, University of Stuttgart, Germany

Supervisor: Prof. Andrew R. Martin, Kraft- och värmeteknologi; Doc. Jens Fridh, Kraft- och värmeteknologi

Export to calendar


In this energy transition era, turbocharging is still an important technology for the automotive industry to reduce fuel consumption and lower emissions in its vehicles. This importance can be seen from both conventional fossil-fuel powertrains, and emerging applications, such as increased utilization of biofuels along with hydrogen fuel cells. For automotive turbochargers, the turbine has mainly two alternative types, i.e., mixed flow turbines (MFTs) and radial flow turbines (RFTs). These devices are mature and commercially available yet still have significant potential for improvement towards ensuring higher performance, more robust operation, and lower cost. With this in mind the overall aim of this study is to improve the aerodynamic design and the aeromechanical analysis methods for MFTs and RFTs. Specifically, the investigation covers three research topics: meanline method, stator tilting endwall design, and forced response analysis. 

A meanline method tool is newly developed to predict the performance curves. For RFTs, the results present a generally good agreement between the predicted performance and experimental data. However, for MFTs, two limitations of loss models used in the meanline method have been identified: spanwise variation of incidence at the rotor inlet is neglected; and performance variations at different speeds cannot be captured by the investigated passage loss models. To overcome the first limitation, a multi-section incidence loss model is proposed. For the second limitation, more research work is suggested to investigate the effect of mixed-flow features at the MFT rotor inlet.      

As a contribution to investigate the mixed-flow feature at the MFT rotor inlet, different stator tilting endwall designs are numerically evaluated with computational fluid dynamics (CFD) tools. An MFT with well-documented experimental data is selected as the baseline and used to validate the CFD method. Performance improvement has been seen from those designs with a sharp turning on the shroud-side endwall just before the rotor leading edge. The optimal design in this study has a -45° tilting angle of the shroud-side stator endwall. It achieves approximate 1%-point higher efficiency than the baseline design over the 100% and 50% speed lines. Detailed aerodynamic analyses of the internal flow field contribute to the understanding of the performance change. 

After the aerodynamic design, aeromechanical analyses are necessary steps to achieve the mechanical robustness. In this part of the study the accuracy and the computational cost of different CFD methods are compared in the forced response analysis of an open-geometry RFT using three CFD methods – full annular, phase-lag, and non-linear harmonic (NLH)– for forcing predictions paired with time-domain and harmonic balance (HB) methods for the aero-damping predictions. It is found that for the stator-induced forcing, all three CFD methods predict the same pattern of forcing distribution. Taking the full annular method as the reference, the maximum blade displacement predicted by the other two methods has less than 15% deviation. However, for the volute-induced forcing, the NLH method is excluded due to increased computational cost. The phase-lag method predicts a distinct forcing distribution to the reference full annular method, leading to approximate 50% difference of the maximum blade displacement. When predicting the aero-damping, the reference time-domain and the HB methods predict similar log decrement values with less than 4.6% deviation. In terms of computational effort, the harmonic methods, namely NLH and HB, reduce the effort by a factor of 42 and 6 respectively for the forcing and aero-damping predictions compared with the reference method.