Experimental and Artificial Neural Networks Modeling for Rivers Bed Morphology Changes near Direct Water Supply Intakes

摘要

In this research, the problem of sediment movements near direct intakes was investigated from the river bed morphology point of view rather than that concerning the effect of sediment withdrawal by the intake to the water treatment plant. As expected the river bed morphology will be affected b y the intake operation, and when the pumping stops, the river will tend to recover this effect by its natural flow, hence a model is required to relate the rate of river bed morphology recovery to the variables that are expected to be relevant, such as the pumping rate, the geometrical variables and time of operation to time of non-operation ratio. A physical model was built. Experiments were conducted to create a data base for these input -output variables, which were used to find an (ANN) model, for the representation of this relationship. Image processing technique is used in this study to analyze the scour and deposition photos from which the volumes of the scour holes after intake operation time and that after intak e non-operation time were found, which allo ws the estimation of the rate of recovery. The results obtained from the image processing o f these photos had prevailed that these volumes can be approximated as a half ellipsoid. An ANN.-factorized back propagation model was fitted to the data base with a correlation coefficient of (0.843), which was considered acceptable according to Smith (1986) criteria. Comparison between the outp ut values(rate of recovery) obtained using this (ANN) model and those obtained experimentally for cases that are not included in the data base, indicates high compatibility with a maximum percentage difference of (7.15% and 5.07%) for overestimating and underestimating respectively