A comparative study from spectral analyses of high-order methods with non-constant advection velocities (Q831247)

The response of numerical methods to flow properties dynamic changes is a key ingredient of numerical modeling calibration and the authors explore a range of spectral analyses and canonical fluid mechanics problems to compare the responses of the high-order spectral difference schemes and of the flux reconstruction methods. This is done in order to obtain more insight about the expected behavior of the numerical methods for typical engineering applications. The present work is based on three main approaches to address accuracy and robustness of a numerical scheme: (a) temporal eigenanalysis, (b) spatial eigenanalysis, and (c) non-modal analysis. All of these techniques are generalized to the non-constant one-dimensional conservation law for the nonhomogeneous linear advection equation. Numerical stability of flux reconstruction (FR) and spectral difference (SD) schemes is ensured for the linear advection equation, whereas aliasing errors could arise in the case of non-constant coefficients. The authors perform a comprehensive spectral analysis of high-order schemes and present the classical temporal approach alongwith the more recent spatial and non-modal analyses. All of these techniques are applied to the FR methods, in particular, those recovering the nodal Discontinuous Galerkin (DG) and the SD schemes for linear advection, The authors verify theoretical findings through a series of numerical experiments. The constant-velocity advection equation with a prescribed inlet condition is discretized using the SD scheme in order to verify spatial eigenanalysis results. Dissipative bubbles, typical of central fluxes, are observed in accordance with spatial eigenanalysis theory. The authors conclude that upwind fluxes are the most suitable choice for Implicit Large Eddy Simulation (ILES) computations. In spite of this, the order of approximation and the level of accuracy of the approximation play an important role and can easily lead to insufficient or excessive levels of dissipation. High-order discretizations, typically of order three to four, employing upwind fluxes are commonly used in the literature to ILES. Spectral analyses, like those presented here, can be used as a useful tool in the design of dynamical explicit subgrid scale (SGS) models which adapt the level of dissipation based on the order of approximation.