Improvement of the scale-adaptive simulation technique based on a compensated strategy

Abstract

The original scale adaptive simulation (SAS) method can be implicitly divided into two computational regions, i.e., Reynolds-averaged Navier–Stokes (RANS) region in the inner layer and large-eddy simulation (LES) behavior region in the outer layer. In the inner layer, the small-scale fluctuations captured by RANS are suppressed. Similar as the RANS/LES hybrid methods, the excessive turbulent dissipation from RANS model may suppress the generation of LES content in the region near RANS/LES interface. To address this problem, an improved SAS technique has been proposed in the present study, i.e., the compensated SAS (CSAS) approach. The CSAS method can be also divided into two different computations. One is still the original SAS calculation in the outer layer region, and the other is the constrained LES computation in the inner layer. To achieve CSAS technique, Reynolds stress and turbulent heat flux in the inner layer are calculated by a specially treated subgrid-scale (SGS) model, i.e., both SGS stress and SGS heat flux are split into the mean and fluctuating parts. RANS conditions are naturally satisfied in the mean part, and the remains (the extra mean and fluctuating parts) can be treated as the compensated terms. Just as the original SAS, the interface of compensated and non-compensated regions is also determined by the local turbulence scales. To assess the prediction performance of CSAS technique, three test cases have been investigated, including two massively separated flows and one attached flow. Two massively separated flows are the incompressible and compressible flows past a circular cylinder. A spatially evolving supersonic turbulent boundary layer is chosen as the test case for the attached flow. The predictions of mean flow and turbulence statistics show that some improvements are obtained by CSAS.