The researches published so far on flexible net barriers focused on the control of debris flows and intercepting falling rocks. Few research studies focused on describing the interaction between a flexible barrier and large wood (LW) flows. Due to the limited articles published on this subject, there are a number of issues that have not yet been addressed, which have an effect on the behaviour and design of a net. This knowledge gap is a strong obstacle to the use of these cheap and light structures. Therefore, this research is aimed towards studying some of these: LW-related head losses against flexible barriers and release conditions of LW overtopping flexible barriers. Several experiments were carried out on a scale model in order to analyze the effect of increasing discharge, as well as the amount of wood volume introduced into the channel, on the water depth at the structure and the eventual release of logs. The flexible barrier was printed on a 3D printer and scale factors were determined so that the barrier deforms consistently with the load it experiences at real scale. Results could then be applied to the prototype scale with more confidence. To the best of our knowledge, this thesis is the first work performed on a scale model of a flexible barrier using a method to rigorously follow the barrier mechanical similitude. The results revealed that the reliability of flexible barriers as a method of controlling LW flow is quite high. Tests on rigid barriers performed in a previous work showed that overtopping and massive release of LW was nearly systematically observed. Conversely, no overtopping was observed on the same experimental apparatus with flexible barriers: they extremelly permeable and our pumping power was not high enough to submerge them. To find an overtopping condition, it was necessary to place some blockage that favors sufficient increase in water depth and, as a consequence, the release of the LW flow through the dam crest. With the first series of experiment (without obstruction), a new dimensionless approach for the calculation of head losses based on the ratio between buoyancy and drag force and the dimensionless solid volume of LW was developed and compared with the water depths obtained in the experiments. It showed better performance than different equations that have been proposed for the estimation of head losses related to the interaction of LW and rigid barriers. It was found that these equations overestimated head losses in flexible barriers. The complementary experiments (with artificial obstruction of the net) enabled to observe seven massive overtopping events. Consistently with the past work on rigid barrier, it was observed that overtopping appear when water level above the net crest is typically about 2-5 times the LW mean diameter.