Water management requires a good understanding of sediment dynamics. Dam management, sediment mining, channel training works are all anthropogenic activities that deeply affect this dynamics. In many areas of the world, their level is such that fine-grained sediment fluxes have been, or are being, modified significantly in magnitude and rate. This results in a complex evolution of the river bed and banks that may alter living conditions for fishes and might be detrimental for human activities Fine sediment infiltration particularly affects the river bed structure and permeability. Clogging depends both on hydraulic conditions and bed sediment grain size distribution. The hydraulic conditions for the deposition or erosion of fine and coarse sediment mixtures are poorly known. A better river management requires further investigations, particularly on flushing event efficiency and impacts in terms of restoration. Flushing operations are already launched in the Durance River but still needs to be optimized. In this paper, several experiments are presented which aim at better understanding the removal of a fine sediment clog by flows restoration. Four experiments have been conducted to flush previously infiltrated beds in the tilting flume of Irstea Lyon-Villeurbanne (HH-Lab). Gravel bed (d50=6.8 mm) was initially clogged with sand (d50=0.813mm) and small glass beads (d50=0.066mm). To obtain such bed, fine sediment-laden flows were recirculated over an initially clean gravel bed. Fine sediment were in homogeneously distributed along the vertical in the bed as predicted by (Herrero & Berni 2016). The same amount of water was released in the flume for each experiment but the shape of the hydrograph was different Bed shear stress was estimated using water depth and velocity measurement. As a reference, inception of movement of particles of a gravel bed starts for a flow discharge approximately equals to 35L/s. Pictures of the bed were taken at the beginning and at the end of the experiments. The evolution of fine sediment deposits (shape of the dunes, fine content within the bed) was obtained from pictures analysis. It was found that initial beds were not exactly the same for each experiment as the infiltration process was undertaken under different conditions. Initial bed of experiment 1 presented the largest ripples all along the flume whereas ripples were quite flattened for experiments 3 and 4. A laser scanner was used to obtain bed roughness along the flume. Bed form roughness was estimated via the standard deviation of a bed elevation profile of 10cm and grain roughness via the standard deviation of a 3.5cm detrended profile. As a reference, the roughness of a flat gravel bed is about 3mm. For experiments 1 and 2, the ripples were washed out but the small grain roughness attests that fine sediment mainly remains at the bed-surface. Hydrographs, and especially the experiment 3, better removed fine sediments (higher grain roughness, closer to the one of gravels). Both hydrographs slightly eroded the bed, similarly to the experiment 2 but more than the experiment 1. From these experiments, it appears that with a given amount of water, the best strategy to wash fine sediment from the surface with a limited mobilisation of the coarse matrix is a rising hydrograph. An additional experiment has shown that higher discharges do not better clean the bed. None of the experiments results in a gravel bed free of fines over a larger depth than one gravel grain. However, if the bed needs to be entirely preserved, lower discharges can partially clean the surface. French Water agency (AERMC) is thanked for its support.