Development and Validation of Turbulence Closures for Three-Dimensional Reynolds-Averaged and Partially-Averaged Navier-Sotkes Application to Open-Channel Flow in Bends and Meanders

Abstract

Development and validation of turbulence closures for three-dimensional Reynolds-Averaged and Partially-Averaged Navier-Stokes. Application to open-channel flow in bends and meanders Understanding and being able to predict curved and meandering flow behaviour is key to river engineering. This work analyses this kind of flows using three-dimensional, non-hydrostatic computational models. Given the ubiquitous presence of turbulence in environmental flows and its crucial importance, different turbulence closures are applied to three curved and meandering open-channel flow scenarios: a single 270° bend, a two-bends meandering channel and an infinite meander. The analysis particularly focuses on the structure of the secondary flow, the mechanisms of generation and modulation of coherent structures, and their influence on the shear stresses. The modelling approaches utilised during this research fall within three fundamental families, URANS, PANS and LES. The fundamental difference among them is their approach to solve or model the turbulence stresses. The predictions provided by these models were tested, compared and validated. The influence of several modelling parameters – besides the turbulence closure itself - on their performance is also analysed, with a special emphasis on the discretisation scheme for the convective term and the inflow conditions. The results show that some configurations of PANS remarkably match the available experimental and LES datasets regarding the prediction of primary and secondary flow and turbulence structure. URANS combined with a k-ε turbulence closure provides a very robust and consistent forecasting of the scenarios under investigation, particularly the primary flow, while lacking on the prediction of some of the mechanisms driving the secondary motion and the turbulence structure. Non-linear eddy viscosity models were tested with irregular results, and overall failing to improve k-ε performance. Turbulence development and the memory of prior bends in meandering channels seem to be key to the secondary flow structure. It was found that coherent structures of consecutive bends interact with each other, which has important repercussions to sediment and pollution transport in environmental flows. It was also found that the turbulent fluctuations within the bends are strongly anisotropic and cannot be well described by models reliant on isotropic assumptions. Future lines of work based on this research could provide dynamic PANS models and new turbulence closures that will generate quick, reliable, and accurate three-dimensional tools for river engineering modelling.