Session: 32-01 High Fidelity CFD

Paper Number: 127676

127676 - High-Fidelity Predictions of Aerodynamic Losses Through a Supersonic Stator Vane 

Organic Rankine Cycle machines (ORCs) have been under intensive research in the recent years as they show very good promise to increase waste energy recovery. A key component to ORC machine efficiency is the expander,  and efforts have been made to understand and reduce losses associated with the flow acceleration through the vane (Baumgärtner et al. (2021), Galiana et al. (2015)). Optimal blade shapes for a specific target outlet pressure have been proposed in the frame of non-ideal gas behavior (Wheeler & Ong (2013)), and help approach the isentropic flow limit. However, sources of loss such as the low-pressure region behind the trailing edge (TE), or entropy generation in the boundary layer must be addressed. A first step in addressing these mechanisms is to accurately predict them, and the commonly used Reynolds-Averaged Navier Stokes (RANS) methods are known to fail in separated or transitional flows. On the one hand, some models assume that the boundary layer is fully turbulent, especially in ORCs, where the typically high density of the fluid leads to very high Reynolds numbers. The overestimated boundary layer height near the blade TE affects the flow separation controlling the base pressure, resulting in poor performance predictions. On the other hand, RANS models that explicitly treat the transition may provide more satisfactory results (Marconcini et al. (2016)) but are still incapable of providing reliable estimates of wake thickness and total pressure loss. In this study, we produce both low- and high-fidelity simulations of a dense gas flow in a supersonic stator vane designed and experimentally investigated by Baumgärtner et al. (2021) to showcase the crucial importance of the accurate prediction of the boundary layer state.

First, the flow around an annular ORC turbine cascade is simulated with the RANS equations supplemented by the Spalart-Allmaras turbulence model. Comparisons are made with experimental measurements of the time-averaged pressure inside the vane in both air, treated as a perfect gas, and a dense vapor, the refrigerant R134a. Fair agreement is obtained for the pressure distribution, but discrepancies occur in regions of interest such as near the TE or across the shock waves and wake behind the blades. Then, a higher-fidelity method, namely, Delayed Detached  Eddy Simulation (DDES), is considered: the computations are ongoing. Further details on endwall effects as well as secondary radial flows within the vane are expected to improve the predictions. Finally, wall-resolved Large Eddy Simulations (LES) are performed on the same geometry, but in a linear cascade configuration with spanwise periodicity conditions to reduce the computational cost. Results clearly show that the flow near the blade wall showcases complex phenomena, such as boundary layer re-laminarisation, shock-induced laminar to turbulent transition, and supersonic vortex shedding at the TE. Companion RANS and DDES simulations of the same simplified configuration are systematically compared with the LES. Large differences in the loss estimates between RANS and LES are observed, due to the inability of RANS to predict the boundary layer state and the low pressure behind the TE. Overall, the study provides strong evidence of the importance of using models of increased fidelity for accurate ORC turbine design.

Presenting Author: Camille Matar Institut Jean le Rond d'Alembert, Sorbonne University

Presenting Author Biography: BEng in Mechanical Engineering at the University of Southampton, UK.
MSc in Aerodynamics at Imperial College, London, UK.
Currently a PhD student in Sorbonne University in Paris, France.

Authors:

Camille Matar Institut Jean le Rond d'Alembert, Sorbonne University
Xavier Gloerfelt DynFluid, Ecole Nationale supérieure d'Arts et Métiers
Paola Cinnella Institut Jean le Rond d'Alembert, Sorbonne University