Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : November 29, 2018 - November 30, 2018
The prediction of viscoelastic-fluid turbulent flows within the RANS (Reynolds-Averaged Navier-Stokes) computational framework is required for engineering applications. For such flows the modelling of an additional non-linear term Λij, correlated with the viscoelastic stress dissipation in Reynolds-averaged constitutive equation, is of decisive importance for the flow prediction accuracy. In this paper, a Reynolds-stress model (RSM) was adopted as a background turbulence model within the RANS framework for computing a drag-reducing turbulent flow over a backward-facing step (BFS). Two different model formulations for Λij were investigated by computing three BFS configurations differing in terms of geometry expansion, i.e. expansion ratio (ER). The turbulent intensities were overestimated within the separated shear layer just behind the step edge for the case with ER = 1.25, while they were slightly underestimated for the case with ER = 1.75, irrespective of the Λij modelling. As a result, the reattachment length predicted by RANS became rather short exhibiting an opposite dependency on the ER-parameter compared to the complementary DNS data. In summary, the overall models’ predictive performance in terms of presently investigated Λij-term formulations differs only marginally within the separated shear layer.
