Prediction of the Characteristcs of Free Radial Hydraulic B-jumps Formed at Sudden Drop Using Anns

Pure theoretical models seeking for the length of jump is extremely difficult (if not impossible). For this reason, empirical prediction equations are normally developed using statistical multiple line regression (MLR). These models may be satisfactory enough for few numbers of variables but may predict unrealistic values when many variables are involved. Fortunately, the recently developed computational tool, the artificial neural networks, can be used to develop computational models to predict the length of jump more accurately than MLR can do. In this research paper, the ANN was used to develop a prediction model to predict the length of free hydraulic B-jumps formed at sudden drop in radial stilling basins provided with end sills. The data needed for modeling were collected from large series of experiments conducted in a laboratory flume under wide variety of test conditions. Results from both MLR and ANN for the length of jump were compared with the experimental data. Also, the available MLR and theoretical models for the depth ratio and the energy loss ratio were compared to similar ANN model developed for the same ratios. It is concluded that the ANN is more efficient than other prediction models in most of the cases.