Search

This paper presents and discusses the results of the 2022 International Computational Fluid Dynamics Challenge on violent expiratory eventsaimed at assessing the ability of different computational codes and turbulence models to reproduce the flow generated by a rapid prototypical exhalation and the dispersion of the aerosol cloud it produces. Given a common flow configuration, a total of 7 research teams from different countries have performed a total of 11 numerical simulations of the flow dispersion by solving the Unsteady Reynolds Averaged Navier-Stokes (URANS) or using the Large-Eddy Simulations (LES) or hybrid (URANS-LES) techniques. The results of each team have been compared with each other and assessed against a Direct Numerical Simulation (DNS) of the exact same flow. The DNS results are used as reference solution to determine the deviation of each modeling approach. The dispersion of both evaporative and non-evaporative particle clouds has been considered in 12 simulations using URANS and LES. Most of the models predict reasonably well the shape and the horizontal and vertical ranges of the buoyant thermal cloud generated by the warm exhalation into an initially quiescent colder ambient. However, the vertical turbulent mixing is generally underpredicted, especially by the URANS-based simulations, independently of the specific turbulence model used (and only to a lesser extent by LES). In comparison to DNS, both approaches are found to overpredict the horizontal range covered by the small particle cloud that tends to remain afloat within the thermal cloud well after the flow injection has ceased.

Previous Scale-Resolving Simulation studies of flow over urban-like obstacles uses Large Eddy Simulation and course grids to reduce computational cost. However, the coarser the mesh the more reliance on the subgrid scale model to accurately account for scales associated with high wavenumbers. Furthermore, when high-resolution simulations are of importance, such as the transport of urban contaminants, mesh refinement becomes necessary. Often clustering of mesh cells produce errors at grid-refinement interfaces, mainly on the fine side of the mesh when it is located upstream of the coarse one. Three scale-resolving turbulence models, the One-Equation Scale-Adaptive Simulation (One-Eq.-SAS), the Shear Stress Transport-Improved Delayed Detached Eddy Simulation (SST-IDDES), and the Algebraic Wall-Modelled Large Eddy Simulation (WMLES) models are utilized to assess their effect on the accuracy of the results when applied on both coarse and mesh-refined grids. The selection of these models was first based on the computational cost where the WMLES is the cheapest to solve since it involves no partial differential equation, while the SST-IDDES model is computationally the most expensive. Simulations are carried out on a relevant and complex test case of flow through a periodic array of cubes. The results reveal that models that do not inherent grid scale parameters in their formulation perform best in flows with global instabilities.