An investigation of interface conditions inherent in detached-eddy simulation methods

The interfaces where the RANS modeled areas match the LES resolved regions are comparatively investigated with regard to the four popular detached-eddy simulation (DES) variants; namely, one-equation Spalart-Allmaras (SA) and two-equations Menter's SST background DES methods (SA-DES and SST-DES), as well as their respective delayed versions (SA-DDES and SST-DDES). The comparisons are aimed at further interpretation of their performance differences under various flows. Although all four DES variants can consistently predict results in fully separated circular cylinder flow, the SST-DES interface is like the SA-DDES interface around the wall, which indicates that, in this case, the shielding function f(d_cor) of SST-DDES is redundant. Moreover, the recalibrated f(d_cor) for SST-DDES is found to preserve double the boundary-layer thicknesses in the flat-plate flow, and shown to be too conservative to resolve the unsteady vortex in the cavity-ramp flow. On the other hand, SA-DDES with the shielding function f(d) shows an advantage by properly balancing the need of reserving the RANS modeled regions for wall boundary layers and generating the unsteady turbulent structures in detached areas.