Author(s): Adrien Bourgoin, Sylvain Guillou, Jérôme Thiébot, Sofiane Benhamadouche, Riadh Ata
Linked Author(s): Riadh Ata
Keywords: Turbulence, numerical modelling, large eddy simulation, free surface flows, Telemac
Abstract: For the purpose of improving turbulent processes modelling for environmental flows, a large eddy simulation (LES) approach is being developed in TELEMAC-3D (Hervouet, 2007). Although it is still not widely used, LES is increasingly applied for this kind of flows, thanks to the growth of computation resources. RANS modelling, generally associated with k ? ? model, remains favorable for numerical modelling of natural flows, and is by the way the most popular approach for turbulence modelling. Nevertheless, as this approach neglects the transient flow characteristics, it cannot be used to analyze the turbulence induced by complex seabed morphology (abrupt bathymetric variations, macro-structures, bed roughness etc. ). The development step involves the implementation of several LES turbulence models, such as the dynamic Smagorinsky (Germano et al. , 1991; Lilly, 1992), and several numerical tools needed for the implementation of a LES. For example, the turbulent inlet boundary condition is achieved by a Synthetic Eddy Method (Jarrin et al. , 2006) which produces a fluctuating and periodic boundary condition in order to initiate the turbulent processes. Moreover, as TELEMAC-3D uses prismatic meshes that can be strongly anisotropic, the turbulence model has to be modified by introducing two length scale filters (instead of one). An important part of the developments has been achieved. The chosen validation case is a in an open channel (Handler et al. , 1993). Although first results are encouraging, they revealed lots of issues related to this kind of models (scheme order, mesh quality, mesh anisotropy, CPU time, boundary conditions, periodicity etc. ). At this step, several issues have been fixed (boundary conditions, periodicity, mesh anisotropy). However, the remaining issues which can explain major gaps between numerical an experimental results (mainly scheme order and numerical diffusion) are ongoing works. Preliminary attempts show promising improvements and will be detailed in future publications
Year: 2017