Expander design and optimisation for Electric Turbocompounding and Organic Rankine Cycle systems in waste heat recovery applications

Abstract

This thesis investigates the effect 3D blade features and non-radial blading on the performance of radial turbines for waste heat recovery applications. To address the problem, two design methodologies were developed and used for the analysis: a low-order meanline model accounting for non-ideal gas effects and a 3D parametric model coupled with CFD.

The design of expanders for waste heat recovery systems considered in this thesis are electric turbocompounding (ETC) and Organic Ranking cycle (ORC). These are challenging compared to traditional radial turbomachinery as ETC operates at a low-pressure ratio with direct waste heat and ORC operates at a high-pressure ratio with refrigerants.

Meanline modelling was able to predict the efficiency and mass flow of radial turbines both with air and refrigerants as working fluids with a relative root mean square error (RRMSE) of less than 1% between meanline and CFD results. This accuracy was achieved after calibration of the loss coefficients with CFD. Moreover, optimum radial turbine designs were obtained for the ORC and ETC applications achieving total-to-static efficiency of 79.98% and 82.86%, respectively. However, meanline modelling has limitations in predicting losses accurately and capturing the effect of 3D geometry modifications on performance. To overcome the loss prediction limitation, a second calibration of the loss coefficients is suggested by minimising the error between the loss breakdown in meanline and the loss breakdown in CFD. Although the error in loss distribution prediction meanline and CFD decreased after this second calibration, the RRMSE in efficiency increased up to 4.4\% and 3.6\% in ORC and ETC applications, respectively.

The 3D parametric model coupled with CFD was used to address the meanline model limitation to predict the effect of 3D geometry modifications. For ORC application, the effect of cone angle of the rotor meridional profile was evaluated, while the effect non-radial fibre blading was assessed for the ETC application. The geometry modifications were introduced after finding optimum baseline geometries with the 3D parametric model for the turbines. The efficiency of the optimum designs obtained by 3D parametric model and the meanline approach was similar, showing a 1.3pp increase only for the ORC turbine.

The modification of the cone angle of the ORC rotor for the same meridional profile (same radii, blade angles and blade heights) led to a maximum efficiency difference of 2pp, while the meanline model predicted no difference. The non-radial fibre assessment concluded that lower incidence angles can improve efficiency. The optimum non-radial fibre blade design beta_blade,4=20 deg showed an increase of 0.3% in efficiency in single passage simulations compared to the baseline ETC design, which was radial fibred. The improved performance was also demonstrated experimentally at Imperial College's test rig, showing a maximum efficiency increase of 2pp at design point.