Author(s): Minh Duc Bui; Petr Penz; Peter Rutschmann
Linked Author(s): Peter Rutschmann
Keywords: Rtificial neural network; Contraction scour; Sediment transport
Abstract: The paper presents an artificial neural network (ANN), which could accurately estimate maximum equilibrium contraction scour depth. The designed feedforward network includes one hidden layer and seven nodes within that layer. Its hidden neurons use a hyperbolic tangent sigmoidal transfer function. The learning algorithm is based on Levenberg-Marquardt backpropagation in batch modus. ANN was implemented using the MATLAB software package. The data set used for this work has been conducted by Dey&Raikar (2005) for long contractions under clear-water conditions in the Hydraulic and Water Resources Engineering Laboratory, Indian Institute of Technology, Kharagpur. A dimensional analysis was done to reduce the input variables for the ANN and to achieve a better representation of these input variables. The importance of the individual input parameters was tested with a sensitivity analysis. This revealed the contraction ratio to be the by far most sensitive parameter, followed by the densimetric Froude number. The calculated results show that the selected ANN estimates equilibrium maximum scour depth under clear-water conditions within the range of the used data set significantly better than other conventional methods.