Multiscale data-driven modeling of transient deposition physics of dense granular flows

Abstract

Geophysical flows involving granular masses exhibit complex dynamics with transient mass and momentum changes due to deposition. Geophysical flows are typically simulated using depth-averaged (DA) models, which rely on empirical closures for deposition. However, these models typically overlook the detailed grain-scale physics involved in deposition, treating the flow as an equivalent fluid at the macro-scale. This study introduces a multiscale framework to integrate grain-scale deposition physics into macro-scale DA models without relying on empirical closures. The framework utilizes a surrogate model, trained on discrete element modeling (DEM) datasets, to capture changes in effective flow depth. This surrogate model is integrated with a DA model to create a multiscale approach, improving the deposition physics within an efficient computational framework. The effectiveness of the proposed multiscale framework is assessed by studying how a granular mass, initially in motion, settles when the slope angle is suddenly reduced to zero. Predictions from the multiscale model of effective flow depth (i.e., not including deposited material) and DA velocity are compared with DEM results. It is demonstrated that the proposed framework has potential to streamline upscaling simulations and facilitate field-scale hazard assessments in the future.