Closing parameters of morphodynamical models without trial-and-error

Meurice, Robin;Soares Frazao, Sandra
(2023) M.S. Yalin Memorial Colloquium 2023 — Location: Palerme (26.January.2023)

Files

Presentation_Robin_Meurice.pptx
  • Open Access
  • Microsoft Powerpoint XML
  • 6.02 MB
Download

Details

Authors
Abstract
The widely used shallow-water equations are insufficient to model appropriately complex river flows highly laden with sediment such as dam-flushes. We propose a 2D finite-volume two-phase/two-layer model able to do so and to cope with the sediment concentration. The objective is to use this model as a predictive and optimization tool for future dam flushes, to avoid environmental disasters, among other things. The problem is that this model is more complex than a simple shallow-water model. It has more parameters to calibrate. For some of them, no closure formulation exists in the literature and a river-specific trial-and-error calibration based on past-obtained data must be achieved. Hence, the model cannot be used as a predictive tool. To circumvent this trial-and-error calibration, we propose a closure procedure combining a numerical experiment and machine learning. That procedure can theoretically be used to close the last unclosed parameter of any morphodynamical model. The model, with its parameters closed by the presented procedure, gave satisfactory results for two different dam-break test cases. Yet, its performance in the case of dense suspension still needs to be addressed.
Affiliations

Citations