A near-wall domain decomposition approach in application to turbulent flow in a diffuser

Abstract

Turbulent flow in an asymmetric, two-dimensional diffuser is studied using a near-wall domain decomposition method and a k–ε turbulence model. A one-dimensional boundary layer equation is used to transfer the boundary conditions from the wall to an interface within the flow. The boundary conditions applied to the fluid velocity and turbulent kinetic energy are of Robin type. They are mesh independent and can account for arbitrary source terms. This approach avoids the computational expense of fully simulating the turbulent boundary layers. For the first time, the technique has been applied to modelling a separated flow with an unstructured code. It is shown how the interface boundary condition on the turbulent kinetic energy allows the recirculation region in the diffuser to be captured. In contrast, the standard wall function approach, based on the log law, fails to predict any recirculation region. The only parameter required to apply the domain decomposition method is a turbulent viscosity profile across the boundary layer. Three different profiles are used in this work. It is shown how making the turbulent viscosity a function of the pressure gradient improves flow predictions for the diffuser. The results demonstrate that the method is an efficient way to simulate the boundary layers in engineering problems that include complex geometries or separating flows.