Modelling practices for axial compressors - assessment of contemporary enhancements of k−ω shear stress transport turbulence model

Abstract

A Reynolds-Averaged Navier-Stokes (RANS) analysis requires turbulence models, such as the Shear Stress Transport (SST). Their limitations are particularly evident in compressor applications due to the inherent flow complexity. Several modifications were proposed to enhance the SST predictive capability, but only a handful of validation examples exist for axial turbomachinery. Furthermore, the reproducibility of the available work is frequently restricted by proprietary geometrical data that cannot be published. The aim of this study was to present a comprehensive account of the Computational Fluid Dynamics (CFD) process to analyse an axial compressor. Particular attention was paid to validating the recent enhancement of the SST model and their applicability to axial compressors. Best practices for selecting an adequate numerical setup were evaluated and discussed in detail, including aspects of discretisation, modelling errors, and challenges encountered in the post-processing phase. RANS simulations of a single blade passage of NASA Rotor 67 (R67) were performed at design speed. The turbulence model study was conducted using the SST model along with five enhancements: increased shear stress limiter (SST-a1), streamline curvature correction (SST-CC), reattachment modification (SST-RM), the γ -Reθ laminar-turbulent boundary layer transition (SST-T), and a blend of the latter two (SST-RM-T). The study showed that while the analysed turbulence models gave reasonably consistent results at stable compressor operation, they exhibited marked differences towards the stall. Combining the transition model with reattachment modification allowed for the best agreement with the experiment across the entire operating range. Thus, it was recommended for transonic rotor simulations. Further findings reiterate the importance of careful and consistent development of the CFD methodology and can serve as helpful guidance for similar applications. We demonstrated that mesh distribution is of primary importance to achieve gird independence in sensitive regions such as hub corner stall and tip gap clearance. It was also found that inadequate flow sampling and averaging can significantly distort the simulated compressor characteristic.