Comparing turbulence models for flow through a rigid glottal model

Abstract

Flow through a rigid model of the human vocal tract featuring a divergent glottis was numerically modeled using the Reynolds-averaged Navier-Stokes approach. A number of different turbulence models, available in a widely used commercial computational fluid dynamics code, were tested to determine their ability to capture various flow features recently observed in laboratory experiments and large eddy simulation studies. The study reveals that results from unsteady simulations employing the k-omega shear stress transport model were in much better agreement with previous measurements and predictions with regard to the ability to predict glottal jet skewing due to the Coanda effect and the intraglottal pressure distribution or related skin friction coefficient, than either steady or unsteady simulations using the Spalart-Allmaras model or any other two-equation turbulence model investigated in this study.

Ozone uptake during inspiratory flow in a model of the larynx, trachea and primary bronchial bifurcation

“…Additionally, the S-A system has shown the capability of accurately recreating jet flow, of simulating adverse pressure gradient effects, of exhibiting laminar-to-turbulent transitions—all necessary components for laryngeal simulation. Although work by Suh and Frankel found the S-A turbulence model incapable of simulating the Coanda effect in two-dimensions34, our work shows that such an assertion does not apply when simulations are conducted for three-dimensional flow. Furthermore, the use of the S-A turbulence model in the current study reduced computational costs relative to the k-ω or other, more conventional, two-equation systems.…”

Validation of a flow–structure-interaction computation model of phonation

“…The time-integration scheme in the rigid VF flow solver is first-order implicit, which is lower-order (in discretization) compared to other numerical formulations (Scherer et al, 2001, 2010; Suh and Frankel, 2007). The first-order scheme is the only possible option in the computational FSI model because of limitations on the communication between the flow and structural solvers in the ’segregated’ implementation presented here.…”

“…1). The mesh is similar to that in Suh and Frankel (2007) where the smallest edge length is of the order 0.002 mm. The VF surface does not deform at any point in time.…”

Computational aeroacoustics of human phonation

“…These include studies with fixed vocal folds (e.g. [12,13,14,15]), forced vocal fold oscillation [16,17,6,18,19,20,21,22,23] and models with the airflow fully coupled to elastic tissue oscillations [24,25,26,27,28,29,30]. Only a few of these computational studies [24,12,14,15,22] solve the flow field in 3D, and most of them on a static geometry.…”

“…Numerical computation of such transitory, highly unsteady airflow with massive separation is problematic. The Reynolds-Averaged NavierStokes (RANS) turbulence modeling is used rather marginally [17,12]. Some recent papers [37,14,15] claim that RANS approach for modeling of human phonation is inadequate and employ Large Eddy Simulations (LES).…”