Numerical investigation of corner separation flow using Spalart-Allmaras model with various modifications

Three-dimensional corner separation is an inherent flow feature in axial compressors which blocks flow path and worsens the performance. To find an effective turbulence model for such flow, numerical simulations of the corner separation flow are conducted in the prescribed velocity distribution (PVD) cascade using the Spalart-Allmaras (SA) model with various modifications. Turbulence models studied here include the SA model, the SA-neg model, the SA-noft2 model, the SA-R model, the SA-RC model, the SA-Helicity model, the SA-KL model, the SA-S model and their combinations with the quadratic constitutive relation (QCR) corrections. Compared with the experiment, the SA-Helicity model predicts more accurately than SA model with other commonly used modifications which indicates the importance of reasonably considering turbulent non-equilibrium transport characteristics in large-scale vortices. With the QCR corrections for considering turbulence anisotropy, the blending models perform better than the original SA model but most of them still overpredict the separation region. Among them, the SA-HelicityQCR2013 model which provides a similar result as the SA-Helicity model performs the best. In further validations based on the transonic compressor NASA Rotor 67, the SA-Helicity and SA-Helicity-QCR2013 models can still give more accurate predictions, which is reflected in the stall margin as well as the spanwise distributions of total pressure and total temperature at both the near peak efficiency and the near stall conditions. The universality of the helicity correction is proved to a certain extent. In future turbulence modeling studies of turbomachinery internal flows, it is necessary to reasonably consider turbulence non-equilibrium and turbulence anisotropy in large-scale vortices. (C) 2022 Elsevier Masson SAS. All rights reserved.

Automated summary

This study conducts numerical simulations of corner separation flow and compares the predictive capabilities of various turbulence models. The results show that, with the modifications considering turbulent non-equilibrium transport characteristics and turbulence anisotropy, the SA-Helicity and SA-Helicity-QCR2013 models have more accurate predictions.