A collision model for grain-resolving simulations of flows over dense, mobile, polydisperse granular sediment beds (1608.03566v3)

Abstract: We present a collision model for phase-resolved Direct Numerical Simulations of sediment transport that couple the fluid and particles by the Immersed Boundary Method. Typically, a contact model for these types of simulations comprises a lubrication force for particles in close proximity to another solid object, a normal contact force to prevent particles from overlapping, and a tangential contact force to account for friction. Our model extends the work of previous authors to improve upon the time integration scheme to obtain consistent results for particle-wall collisions. Furthermore, we account for polydisperse spherical particles and introduce new criteria to account for enduring contact, which occurs in many sediment transport situations. This is done without using arbitrary values for physically-defined parameters and by maintaining the full momentum balance of a particle in enduring contact. We validate our model against several test cases for binary particle-wall collisions as well as the collective motion of a sediment bed sheared by a viscous flow, yielding satisfactory agreement with experimental data by various authors.

• The paper introduces a refined collision model for grain-resolving simulations of flows over dense, polydisperse sediment beds, enhancing time integration and enduring contact handling.
• The model effectively handles polydisperse particles, offering a more realistic simulation of natural sediment beds and providing new insights into particle size variance effects.
• Validated against experimental data, the model reliably predicts sediment bed height and velocity profiles, demonstrating its robustness for environmental and engineering simulations.

Overview of a Collision Model for Grain-Resolving Simulations of Flows

The work presented in this paper introduces a sophisticated collision model for the phase-resolved Direct Numerical Simulations (DNS) of sediment transport, specifically focusing on flows over densely packed, mobile, polydisperse granular sediment beds. The model capitalizes on the coupling of fluid and particles using the Immersed Boundary Method (IBM), refining existing approaches to achieve consistent results for particle-wall collisions. This paper extends the methodologies of previous researchers by enhancing the time integration schemes and incorporates novel criteria for handling enduring contact, which is vital in various sediment transport phenomena.

Key Contributions

1. Improved Time Integration for Consistency: The authors have refined the temporal integration approach to mitigate inaccuracies in simulating particle-wall collisions, crucial for capturing correct rebound characteristics. The adaptation involves transitioning from a conventional Forward Euler scheme to a more accurate predictor-corrector scheme intrinsic to the RK method.
2. Polydisperse Particle Handling: The model goes further than previous literature by addressing the complexities introduced by polydisperse spherical particles. By doing so, the model can more realistically simulate environments where particle size distribution plays a critical role, such as in natural sediment beds.
3. Resolution of Enduring Contact: Through a methodical approach, the authors introduce enhancements that handle situations of enduring contact by maintaining the momentum balance accurately. This includes the incorporation of full hydrodynamic and buoyancy effects on particles within the sediment bed, ensuring that the Shields parameter remains a reliable prediction tool for sediment transport.
4. Validation against Experiments: The collision model's efficacy is demonstrated through rigorous validation against experimental data for both binary particle-wall collisions and sediment motion driven by laminar flow, showcasing its reliability and precision in replicating experimental outcomes.

Numerical Results and Implications

Numerical simulations conducted using the refined model show remarkable alignment with experimental benchmarks, revealing its robustness in handling sediment transport scenarios. Particular emphasis is placed on quantitatively validating the model against the experiments by Aussillous et al., a benchmark in the domain. These validations underscore the model's capability to predict sediment bed height and velocity profiles accurately under varied flow conditions, with numerical results such as fluid height and Shields number aligning closely with empirical data.

Moreover, the model's adoption of a polydisperse approach provides new insights into its implications on sediment transport dynamics, potentially enriching our understanding of how variance in particle sizes affects bed mobility and stability. This model opens avenues for more sophisticated analyses of phase-resolved simulations, crucial for environmental and process engineering applications where sediment transport is a primary concern.

Future Directions in Artificial Intelligence

While this paper is rooted in mechanical engineering and computational fluid dynamics, its methodologies intersect with AI research, particularly in machine learning applications for predictive modeling. The meticulous collision model can serve as training data for AI models aiming to predict sediment transport under various conditions, potentially enhancing the rapid simulation capabilities of AI systems in environmental modeling. Moreover, the approach of using IBM alongside DNS provides a testbed for AI-driven optimization algorithms aiming to refine numerical methods for high-fidelity simulations, thereby pushing forward the boundaries of this interdisciplinary research field.

In conclusion, this paper offers a rigorous and validated approach for modeling granular sediments in fluid flow, distinctly positioning itself as a valuable resource for computational simulations of sediment transport. Its implications extend into AI, setting the stage for future research at the intersection of numerical simulations and intelligent predictive analytics.