OpenFOAM based Evaluation of PANS Method with Non-Linear Eddy Viscosity Closure for Separated Turbulent Flows

At high Reynolds number (Re), the traditional Reynolds averaged Navier-Stokes (RANS) approach based on the Boussinesq hypothesis is incapacitated in two aspects: indiscriminate averaging of flow variables and linear eddy-viscosity model (LEVM) closure. Despite satisfactory performance of LEVM+RANS in several flows of practical interest; at high-Re turbulent flows - it fails to capture essential flow features. Partially averaged Navier-Stokes (PANS) methodology, being a scale resolving bridging method, is inherently more suitable than RANS for simulating turbulent flows. However, PANS equations, derived from RANS, continue to inherit the inadequacies from the parent RANS model based on LEVM closure.

To enhance PANS’ capabilities for simulating separated flows, we attempt to address the shortcomings of the LEVM closure. Our study is categorized into three steps: (i) Extraction of PANS version from non-linear eddy viscosity model (NLEVM) based RANS, (ii) Asymptotic analysis of the model and influence of filter-width control parameter on turbulent stresses in the homogeneous turbulence environment and (iii) Implementation in OpenFOAM and evaluation of the model (in predicting mean and instantaneous flow features) in separated non-homogeneous high Re flow past square cylinder.