Abstract
This study aims at providing evidence of the importance of using high-fidelity methods for accurate loss estimation during initial ORC turbine design. The case of a supersonic stator vane in both linear and annular cascades is used to exemplify the divergences in loss between flows where crucial features including the boundary layers, the wake and secondary flow are disparately modeled. Specifically, well known strategies, namely Reynolds-Averaged Navier-Stokes (RANS) modeling, Delayed Detached Eddy Simulation (DDES) and Large Eddy Simulation (LES), are used. In the linear cascade, the misprediction of the boundary layer properties leads to large deviations in trailing edge loss attributed to momentum deficit. On the other hand, resolving the unsteady wake dynamics provides substantial improvement of the base pressure estimation, which is the strongest contributor to overall trailing edge loss. In the annular cascade, fair agreement is found with experimental pressure measurements and numerical data from the literature. Stagnation pressure deficit through the wake is well reproduced by the DDES, while the RANS solution features a wider and slightly displaced wake. Overall, the present study shows that, although the computational cost varies over orders of magnitude, accurately resolving the complex flow phenomena that exist in a turbine vane is a crucial step for improved ORC turbine design.
