Prediction of pressure fluctuations on sloping stilling basins

Abstract

Various types of hydraulic jump occurring on horizontal and sloping channels have been analyzed experimentally, theoretically, and numerically and the results are available in the literature. In this study, artificial neural network models were developed to simulate the mean pressure fluctuations beneath a hydraulic jump occurring on sloping stilling basins. Multilayers feed a forward neural network with a back-propagation learning algorithm to model the pressure fluctuations beneath such a type of hydraulic jump (B-jump). An explicit formula that predicts the mean pressure fluctuation in terms of the characteristics that contribute most to the hydraulic jump occurring on the sloping basins is presented. The proposed neural network models are compared with linear and nonlinear regression models that were developed using considered physical parameters. The results of the neural network modelling are found to be superior to the regression models and are in good agreement with the experimental results due to relatively small values of error (mean absolute percentage error).Résumé : Divers types de ressauts hydrauliques se trouvant dans les canaux horizontaux et inclinés ont été analysés expérimentalement, théoriquement et numériquement et les résultats sont disponibles dans la littérature. La présente étude développe des modèles de réseaux neuronaux artificiels pour simuler les variations moyennes de pression sous les ressauts hydrauliques le long des bassins d'amortissement inclinés. Un réseau neuronal à couches non bouclé muni d'un algorithme d'apprentissage à rétropropagation est utilisé pour modéliser les variations de pression sous de tels types de ressauts hydrauliques (B-Jump). Une formule explicite qui prédit la variation moyenne de pression en termes des caractéristiques contribuant le plus au ressaut hydraulique dans les bassins d'amortissement est présentée. Les modèles de réseaux neuronaux artificiels proposés sont comparés aux modèles de régression linéaire et non linéaire développés en utilisant les paramètres physiques arrêtés. Les résultats de la modélisation des réseaux neuronaux s'avèrent supérieurs à ceux des modèles de régression et présentent une bonne corrélation avec les résultats expérimentaux en raison des valeurs relativement petites de l'erreur (pourcentage d'erreur moyenne absolue).Mots clés : réseaux neuronaux, variation de pression, ressaut hydraulique, bassin d'amortissement incliné, formulation de réseau neuronal explicite, analyse de régression.[Traduit par la Rédaction]

Multi-Output Neural Networks for Estimation of Synthetic Unit Hydrograph Parameters: A Case Study of a Catchment in Turkey

“…NNs are relatively stable with respect to noise in data and have a good generalization potential, to represent inputoutput relationships [11]. Once a NN model is trained for its generalization properties, it can be demonstrated that the trained model represents the physical process of the system.…”

“…The knowledge acquired for the problem domain during the training process is encoded within the NN in two forms: (a) in the network architecture itself (through number of hidden units), and: (b) in a set of constants, or weights [12]. Although there are several attempts in other scientific branches that have shown, that useful information could be obtained from trained neural networks (see Yao [12] for relevant references), limited research has been done for water-engineering applications [11,13]. * corresponding author; e-mail: agunal@gantep.edu.tr This study focuses on modelling of three unit hydrograph parameters, namely, the peak discharge q p , the time to peak discharge t p , and the time base of unit hydrograph t b , based on the most relevant hydrological and geomorphological variables.…”

Gene Expression Programing for Estimating Suspended Sediment Yield in Middle Euphrates Basin, Turkey

“…), which are especially attractive for modeling processes. However, adequate knowledge about the physics of these techniques is limited [5,6].…”

An explicit neural network formulation for evapotranspiration

“…The question is: How can one apply this kind of model in any other study, while the model has not been formulated? Recent studies, such as Khorchani & Blanpain (2005) and Guven et al (2006) in the civil engineering literature, have been seen to overcome this problem by providing the explicit formulation on which the neural network system is based.…”