NAVIER-STOKES STALL PREDICTIONS USING AN ALGEBRAIC REYNOLDS-STRESS MODEL

The paper presents results of a computational study of low-speed flow over an airfoil in stall. The code that solves the mean-flow equations is a rather standard explicit Runge-Kutta time-marching cell-centered finite volume technique using central differencing. The Baldwin-Lomax model failed to predict stall, and a standard k-epsilon transport model underpredicted the separation region in comparison with experiment. Only an algebraic Reynolds-stress model produced good agreement with the observed stall. The numerical treatment of the turbulent transport equations is novel. The k and epsilon equations are calculated implicitly using hybrid central/upwind differencing. A tridiagonal matrix procedure solves the resulting discretized linearized equations in both coordinate directions. This method for solving k and epsilon has proved to be very efficient and much more stable than the explicit solver used for the mean-flow equations. The combined approach therefore is semi-implicit. The influence of the explicit adding of the fourth-order numerical dissipation in the mean-now equations is investigated, and it is shown that it has negligible effects on the calculated results.