A coupled approximate deconvolution and dynamic mixed scale model for large-eddy simulation

TL;DR

This paper introduces a novel coupled approximate deconvolution and dynamic mixed scale model for large-eddy simulation of complex flows, validated against direct numerical simulations and showing high accuracy in capturing flow statistics.

Contribution

A new blended subgrid-scale model combining approximate deconvolution with a dynamic mixed scale approach, validated using spectral element methods for complex flow regimes.

Findings

Model achieves excellent agreement with DNS statistics.

Coupling with spectral element method reduces numerical dissipation.

No clipping needed for dynamic parameters, ensuring stability.

Abstract

Large-eddy simulations of incompressible Newtonian fluid flows with approximate deconvolution models based on the van Cittert method are reported. The Legendre spectral element method is used for the spatial discretization to solve the filtered Navier--Stokes equations. A novel variant of approximate deconvolution models blended with a mixed scale model using a dynamic evaluation of the subgrid-viscosity constant is proposed. This model is validated by comparing the large-eddy simulation with the direct numerical simulation of the flow in a lid-driven cubical cavity, performed at a Reynolds number of 12'000. Subgrid modeling in the case of a flow with coexisting laminar, transitional and turbulent zones such as the lid-driven cubical cavity flow represents a challenging problem. Moreover, the coupling with the spectral element method having very low numerical dissipation and dispersion builds a well suited framework to analyze the efficiency of a subgrid model. First- and second-order statistics obtained using this new model are showing very good agreement with the direct numerical simulation. Filtering operations rely on an invertible filter applied in a modal basis and preserving the C0-continuity across elements. No clipping on dynamic parameters was needed to preserve numerical stability.