Direct numerical simulations of particle-laden density currents with adaptive, discontinuous finite elements

Abstract

Abstract. High-resolution direct numerical simulations (DNSs) are an important tool for the detailed analysis of turbidity current dynamics. Models that resolve the vertical structure and turbulence of the flow are typically based upon the Navier-Stokes equations. Two-dimensional simulations are known to produce unrealistic cohesive vortices that are not representative of the real three-dimensional physics. The effect of this phenomena is particularly apparent in the later stages of flow propagation. The ideal solution to this problem is to run the simulation in three dimensions but this is computationally expensive. This paper presents a novel finite-element (FE) DNS turbidity current model that has been built within Fluidity, an open source, general purpose, computational fluid dynamics code. The model is validated through re-creation of a lock release density current at a Grashof number of 5 × 10 6 in two and three dimensions. Validation of the model considers the flow energy budget, sedimentation rate, head speed, wall normal velocity profiles and the final deposit. Conservation of energy in particular is found to be a good metric for measuring model performance in capturing the range of dynamics on a range of meshes. FE models scale well over many thousands of processors and do not impose restrictions on domain shape, but they are computationally expensive. The use of adaptive mesh optimisation is shown to reduce the required element count by approximately two orders of magnitude in comparison with fixed, uniform mesh simulations. This leads to a substantial reduction in computational cost. The computational savings and flexibility afforded by adaptivity along with the flexibility of FE methods make this model well suited to simulating turbidity currents in complex domains.

Application of the adjoint approach to optimise the initial conditions of a turbidity current with the AdjointTurbidity 1.0 model

“…The experimental evidence cannot yet validate all of these theories. Computer models, along with laboratory experiments, have been useful tools in improving our understanding of the dynamics of turbidity currents (Talling et al, 2007a;Kneller and Buckee, 2000;Parkinson et al, 2014). However, computer models have not often been directly applied to recreating deposits found in the field, despite their capacity to do so.…”

Application of the adjoint approach to optimise the initial conditions of a turbidity current

“…DG element types are known to be particularly suitable for advection dominated problems (Peraire and Persson, 2008). They are good at preserving discontinuities as they produce stable discretisations without the need for diffusive stabilisation strategies, such as streamline upwinding (Peraire and Persson, 2008). These are important features in shallow water particle-laden density current models.…”