LAUSR

Abstract The performance of several turbulence closure models in stagnating turbulent flow is investigated and their deficiencies identified. The simulations are compared with new measurements of the two-dimensional stagnation flow on a flat plate in a wind tunnel obtained using stereo-particle image velocimetry. Numerical solutions to the Poisson equation for the instantaneous pressure allowed all pressure-velocity correlations to be estimated and then separated into their linear (mean strain rate dependent) and nonlinear contributions. Reynolds-averaged Navier–Stokes (RANS) turbulence models were investigated because of their widespread use in, for example, urban wind flow simulations, where impinging flow is common. The models tested are the k − ϵ and some variants as well as a common Reynolds stress model with and without wall reflection. As is well-known from previous studies, the k − ϵ model over-predicted k. Durbin’s modified eddy viscosity reduces, but does not correct this over-prediction. The dominance of the turbulent diffusion in the turbulent kinetic energy equation near the wall was not reproduced by any of the gradient diffusion models, suggesting major deficiencies in the modeled terms.