Dissipation additions to flux-difference splitting (Q1346552)

It is believed that the flux-difference splitting methods for solving the Euler equations are generally very robust, and no explicit dissipation is required. In this paper, it was pointed out by the author that explicit dissipation is sometimes needed in this kind of approach. A case in point is a slowly moving shock problem, and another one is related to the supersonic flow past a blunt body. For the case of a slowly moving shock, post shock oscillations have been observed with Roe's flux-difference scheme. With Harten's well-tuned dissipation parameter, \(\delta\), the post shock oscillations can be suppressed. But this parameter needs to be adjusted since it is CFL number dependent. The author suggested an improved expression for suppressing these post shock oscillations. For the supersonic flow past blunt bodies, instability has been observed for some time when Roe's and Osher's Riemann solvers were employed to examine the problem. It was only recently reported by Peery and Imlay, who also recognized the drawback of Yee's formula for artificial dissipation in the attempt to overcome the instability, and proposed a formula based upon the pressure gradient to tune the magnitude of the dissipation. Nevertheless, it was pointed out by the author that their formula is not suitable for viscous blunt-body problem when the direction with larger dissipation is normal to the body or when it is used to compute the flow with angle of attack. Another problem of their formula is that it has excessive dissipation on the supersonic side of the flow. The author proposed a scheme by retaining the form of Yee's expression, but evaluating the parameters \(\delta_k\), on the basis of the pressure gradient, so that the dissipation on the supersonic side of the flow is reduced, and large dissipation parallel to the shock is implemented to assure the stability of computations. The merits of the author's suggestions are supported through his reported results of computations of both cases.