Generalized characteristic relaxation boundary conditions for unsteady compressible flow simulations

Abstract

We develop numerical boundary conditions for the compressible Navier–Stokes equations based on a generalized relaxation approach (GRCBC), which hinges on locally one-dimensional characteristic projection at the computational boundaries, supplemented with available information from the flow exterior. The basic idea is to estimate the amplitude of incoming characteristic waves through first-order one-sided finite-difference approximations which involve the value of the reference flow state at the first exterior (ghost) point. Unlike other characteristic-based relaxation methods, the present one requires minimal user-supplied input, including the reference flow state, which may be totally or partially known, and in general may vary both in space and time. Furthermore, it can be applied to any type of computational boundary, either inflow or outflow, either subsonic or supersonic. The method is theoretically predicted to convey reduced reflection of waves at computational boundaries compared to other ones, and to have better properties of frequency response to injected disturbances. Numerical tests confirm the improvement of the nonreflecting performance, and demonstrate high degree of flexibility, also for problems with non-trivial far-field boundary conditions (e.g. flows in rotating reference frames) and for the artificial stimulation of subsonic turbulent boundary layers.