Professor National Taiwan University, New York, United States
Abstract Submission: Modeling suspended sediment transport in turbulent flows requires accurately capturing the randomness of particle trajectories. Traditional Lagrangian Stochastic (LS) models, such as the Random Displacement Model (RDM) and first-order Langevin approaches, face limitations due to the non-differentiability of Brownian motion and difficulties in validating models with available Eulerian concentration data. To overcome these challenges, we propose a novel multivariate, differentiable Lagrangian Stochastic Process (LSP) that incorporates turbulent velocity structures directly into the model’s random term. This new approach allows for more accurate simulations of sediment transport and improves the integration of Direct Numerical Simulation (DNS) data into model validation. Our results show that the multivariate LSP enhances the prediction of sediment trajectories, providing a more reliable framework for sediment transport modeling. This development offers significant improvements in the accuracy and applicability of LS models for real-world hydraulic systems.
