Aeroacoustic Simulation of Human Phonation with the Wale Sub-Grid Scale Model

The paper reports on an aeroacoustic model of voice generation in human larynx, based on Large Eddy Simulation with the Wall-Adapting Local Eddy-Viscosity (WALE) sub-grid scale (SGS) model. The simulation uses a three-step hybrid approach, with an incompressible finite volume CFD computation providing the filtered velocity and pressure, evaluation of the aeroacoustic sources, and simulation of the sound propagation by finite element discretization of the Acoustic Perturbation Equations. The WALE SGS model is used to overcome the limitation of the classical Smagorinski SGS model, which overpredicts the SGS viscosity in regions of high shear, especially within the boundary layer in the glottal constriction. Results of the 3D CFD simulation, location of the aeroacoustic sources and the spectra of the radiated sound for two vowels are presented. finite volume method with central differencing scheme for the spatial discretization of the diffusion term. The upwind-based schemes bring high numerical diffusion and hence they are not recommended in LES studies. The spatial discretization of the convective term is realized using the total variation diminishing scheme, combining benefits from first order schemes to keep stability and second order schemes for high accuracy. The temporal discretization is realized using the two-step second-order backward scheme. The simulation was run in parallel on 16 processors of a computational cluster for twenty periods of vocal fold vibration, i.e. T = 0 . 2 s, and took about 480 hours of walltime. The timestep was adjusted by an automatic algorithm to keep the Courant number below a predefined limit, here Co < 1. _______________________________________________________________________