Integration of artificial neural networks into TELEMAC-MASCARET system, new concepts for hydromorphodynamic modeling. (June 2019)

• Record Type: Journal Article
• Authors: Kaveh, Keivan; Bui, Minh Duc; Rutschmann, Peter
• Is Part Of: Advances in engineering software. Volume 132(2019)
• Page Start: 18
• Page End: 28
• Publication Date: 2019-06
• Subjects: Artificial neural network -- Sediment transport -- Hydromorphological modeling; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2019.03.011
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: Two new schemes were proposed for hydromorphological changes in fluvial rivers, where artificial neural network models have been integrated into a hydromorphological model system. Two complex hydro-morphological models including the evolution of a 3D isolated bump in a straight channel and the evolution of the bed in a 180° channel bend were used to test the efficiency of the proposed schemes. The novelty of the proposed models is that they reduced the computation costs significantly in the prediction of both hydrodynamics variables and morphodynamics. It can be concluded that applying the developed schemes would reduce the computational costs and simplify the computational procedure, as they did not require the shear stress and sediment transport rate calculations. It is also proved that these schemes could perform well for long-term time series prediction, even though the ANN models were not trained for that length of time. Abstract: In this study, two new calculation schemes were proposed for hydromorphological changes in fluvial rivers, where artificial neural network (ANN) models have been integrated into a hydromorphological model system. For this purpose, the open-source finite-element system TELEMAC-MASCARET has been applied to simulate two complex hydro-morphological models including the evolution of a 3D isolated bump in a straight channel and the evolution of the bed in a 180° channel bend. The simulated results were used as input-data in ANN models to obtain ANN-based approximator for the new proposed schemes. The novelty of the proposed models is that they reduced the computation costs significantly in the prediction of both hydrodynamics variables and morphodynamics. To evaluate the prediction qualities of the proposed models, a comparative study has been carried out for these models by estimating several parameters that describe the errors associated with the model in terms of statistical measures of goodness-of-fit between the estimated bed change and TELEMAC-MASCARET simulation.