Profile loss and turbulence modelling in radial outflow turbine cascades

Abstract

Radial outflow turbines (ROTs) offer an alternative for single stage high-work axial turbines in steam and organic Rankine cycle applications. Their main benefits are lower Mach number levels, improved design and off-design performance and improved capability to handle high volume flow rates. In addition, radial outflow turbines have been recently shown to have potential to replace radial inflow designs in supercritical CO2 applications. The design process of a ROT follows the typical axial turbine procedures by replacing the axial flow component with the radial one. From geometry point of view, the main difference between these turbine types is that the pitch-to-chord ratio changes from blade inlet to outlet. Previous studies have shown that the axial turbine loss correlations tend to overpredict the losses in comparison with the numerical data. However, only part of the most used loss correlations has been studied so far. In addition, the effect of turbulence modeling has not been widely covered in public literature regarding ROTs. Therefore, this study examines the accuracy of Traupel’s loss correlation on ROT cascade profile loss prediction by comparing it with numerical simulations which are validated against experimental ROT data. Also, the effects of different turbulence models are studied in detail by comparing seven different modelling approaches. As a result, it is shown that the Traupel’s model tends to over predict the profile loss compared to all turbulence models. The turbulence modelling approach was found to have significant effect on the results, mainly originating from the boundary layer and wake predictions, which derive from the differences in turbulent kinetic energy prediction capabilities. The results, in general, conformed also with previous axial turbine findings. It was found that the specific dissipation rate and Spalart-Allmaras turbulence models provide highest conformity of the profile loss predictions, they do also agree best with the experimental isentropic Mach number data inside and outside the cascade. The dissipation rate-based models, however, agree least with the experimental data and they predict, for example pressure side flow acceleration phenomena near the trailing edge, which is not seen with specific dissipation rate-based models. In addition, the Reynolds stress models did not justify the computational effort invested since they did not provide improved prediction capability compared to the best performing two-equation and even the one-equation model.