Towards Implicit Subgrid-scale Modeling by Particle Methods

The numerical truncation error of vortex-in-cell methods is analyzed a-posteriori through the effective spectral numerical viscosity for simulations of three-dimensional isotropic turbulence. The interpolation kernels used for velocity-smoothing and re-meshing are identified as the most relevant components affecting the shape of the spectral numerical viscosity as a function of wave number. A linear combination of well-known standard kernels leads to new kernels assigned to the specific use for implicit large-eddy simulation of turbulent flows, i.e. their truncation errors acts as subgrid-scale model. Numerical results are provided to show the potential and drawbacks of the approach.