In this study, attempts to suppress numerical viscosity in incompressible smoothed particle hydrodynamics (SPH) computations are reported. Two-dimensional computations are performed for inviscid and viscous flows to evaluate the effects of numerical viscosity suppression. The first approach is to reduce numerical viscosity at the wall by considering only the wall-normal components of the forces between fluid particles and wall particles. The second approach is to reduce numerical viscosity within the flow field by employing elliptic kernel functions whose major axes are aligned with the local mean flow direction. It is found that special treatment of the wall radically reduces the numerical wall friction. Using an elliptic kernel function is found to work reasonably well in reducing numerical viscosity.
ASJC Scopus subject areas