Contribution to the analysis of geometry effects on mixed flows

This thesis research deals with the transition from a free surface to a pressurized flow considering 2D configurations, which are often present in practice but have been poorly reported to date. The first part of the work has been performed considering a simple experimental scheme made of two rectangular cross section free surface channels connected by a rectangular cross section conduit, where the flow was pressurized. In these preliminary tests, several steady discharges have been tested considering varied conduit cross section variations. The results of these first experimental tests provided qualitative data on the flow features at the transition, enabled to generally assess the potentiality of the 2D numerical solver Wolf2D to model these flows and opened the way to the detailed study of the rectangular transition from a free surface channel to a conduit. In a second step, 14 different geometries of three main configurations have been experimentally considered to assess the effect of the conduit width, height and position along the flume axis (asymmetric and symmetric configurations) on the flow features at the transition. Whatever the geometry, a wide range of steady discharges has been tested with carefully controlled downstream boundary condition. The results analysis provided new insights on the flow characteristics at the transition and enabled to develop and validate two simple analytical expressions to predict the local head loss at the transition.Beside of the experimental investigations, numerical simulations have been performed to assess the ability of the flow solver WOLF2D to correctly model such mixed flows situations. The numerical results have been compared with corresponding experimental data. A very good qualitative agreement between numerical and experimental results has been shown. In quantitative terms, the numerical results are close to or follow the same tendency as the experimental data whatever the geometry and the discharge. However, the prediction of the local head loss is usually underestimated by the numerical model and some specific phenomena observed during the experimental tests cannot be reproduced. Finally, the computation of transient flows in some geometries selected from the previous tests has been performed. The results showed that the numerical solver is able model such unsteady situations without spurious oscillations and provides promising results. These numerical results need however to be validated considering experimental data for instance.