Evaluation of Performance and Code-to-Code Variation of a Dynamic Hybrid RANS/LES Model for Simulation of Backward-Facing Step Flow

Computational fluid dynamics (CFD) simulation results are presented for the canonical test case of flow over a backward facing step (BFS). The BFS case exhibits complex physics including turbulent separation, reattachment, and boundary layer restart. Results are obtained using two different turbulence models as representative examples of two classes of Modeling: Reynolds-averaged Navier-Stokes (RANS) and hybrid RANS-LES (large-eddy simulation). The specific models used are k-omega SST and dynamic hybrid RANS-LES (DHRL). The objective of the study is to compare the performance of both turbulence models as implemented in three different flow solvers (FlowPsi, Loci-CHEM, and Ansys FLUENT) and using three different methods for numerical discretization of the convective terms in the governing equations. Results are compared to experimental data for validation purposes. Results show that both k-omega SST and DHRL models are capable of reproducing the mean flow physics with reasonable accuracy. The differences due to solver algorithm and convective discretization scheme are apparent for both models, but the DHRL model shows more sensitivity, as expected. Overall the results highlight the importance of considering all integrated aspects of a turbulent CFD simulation to ensure that an optimum combination of model and numerical method are employed.