IMPROVED PREDICTION OF LABYRINTH SEAL PERFORMANCE THROUGH SCALE ADAPTIVE SIMULATION AND STREAM ALIGNED GRIDS

摘要

The accurate prediction of cavity flows is of importance to the turbomachinery design process. However, cavity flows are complex. It is known, that RANS models tend to struggle with the prediction of cavity flows and the flow phenomena associated with them. At the same time, scale-resolving methods are more accurate and give a more detailed view on the turbulent structure of the flow. This is accompanied by an inherent dependency on the computational grid, the timestep, and the size of the domain. Therefore, an experimentally validated comparison of RANS, URANS and SAS simulations for a stepped labyrinth seal is given in the paper at hand to demonstrate the individual methods capabilities, limitations, and requirements. It was shown that an alignment of the grid with the local flow direction can save about 40% of computational resources, while simultaneously reducing the discretization error by 25%. RANS and time averaged URANS results in comparison to measurements showed that the swirl development in the cavity is overpredicted and the cavity vortex is underpredicted. A distinct grid dependency was noticed for the SAS-SST turbulence model. The intermediate grid enhances the results in comparison to RANS and URANS. URANS-SST and SAS-SST simulations capture the same dominant frequencies of the velocity spectra, when the same sector size is used. Furthermore, the prediction of dominant frequencies depends strongly on the circumferential size of the domain. The time-averaged results are more sensitive to the grid refinement and turbulence model than to the size of the domain.