A river diversion barrage during its operable life has to be safeguarded against unpredictable local scouring due to uncertain flow conditions and gate openings, which may affect its structural stability. Further, sedimentation in the shallow reservoir (often called the pond) behind a barrage may produce adverse conditions that may affect its normal performance. This thesis studies both these aspects of barrage operation and reports the summary of research work carried out on these two broad topics and discusses the conclusions drawn from each. The first part of the thesis attempts to explore the extent of scour on the downstream of a barrage due to different hydraulic conditions and combinations of gate openings. The hydraulic conditions include both controlled (that is, gate operated) as well as uncontrolled conditions and free or submerged flows as dictated by the tail water level. The gate opening patterns take into account both normal operation conditions and extraordinary situations like the breakdown of one gate. With the help of an artificial neural network (ANN) model, trained with a large number of experimental observations, the worst possible situation of downstream scour is identified. A method to model the profile of downstream scour by ANN is demonstrated that would help to understand the three dimensional variation of the scoured bed under any combination of flow situation. Finally, the application of one commercially available computational fluid dynamics software package, FLUENT, is used for successful numerical simulation of experimental and prototype flow and scour conditions. In the second part of this report, results and conclusions drawn from studies on sediment flush-out from the pond of a barrage are presented. With the help of an ANN model trained with the observed data obtained from experimental runs, the best possible combination of gate opening area, pond water depth and river discharge are identified that would effectively flush-out sediment from the pond. Another ANN model was developed to decide the best pattern of gate opening of a barrage that would flush-out a sediment mound from the pond. Keywords: Barrage, Scour, Sedimentation, Flush-out, ANN, Numerical simulation.