Session Q08: Particle-laden Flows: Simulations

Comparison of Particle Tracking Methods in Large Eddy Simulations and Direct Numerical Simulations

Nanoparticles are found in many useful products, such as sun protection creams, wood preservatives, and tires. However, nanoparticles can have several undesirable characteristics as well. For example, the combustion of hydrocarbons can produce nanoparticles which cause health problems and adversely affect the environment. Because of these deleterious effects, it is important to track them in numerical simulations for predictive modeling. There are two main methods to track particles as they move through a flow field. The first is the Lagrangian approach, in which the motion of each individual particle is tracked. The second is the Eulerian approach, in which the particle phase is treated as a continuum. In terms of flow physics, Direct Numerical Simulations (DNS) are the most accurate, but due to their high computational cost, Large Eddy Simulations (LES) are often used instead as a compromise between accuracy and efficiency. Although there are studies regarding the particle tracking and fluid simulation methods separately, there is still a lack of a comprehensive study on the combination of the two areas. In this work, DNS and LES are conducted on a cube of homogeneous isotropic turbulence with particles tracked using both the Lagrangian and Eulerian methods. Special attention will be paid to the effect of the LES filtering on the transport of the particles.    

SPARSE Model: Closed!

The direct Eulerian-Lagrangian simulation of turbulent, particle-laden flow through the Navier-Stokes equations combined with the tracing of a large number of particles is computationally expensive for large-scale problems. To reduce computational cost, small scale turbulence is often modeled and groups of physical particles are amalgamated into clouds, whose average location is traced. Typical Lagrangian models (such as Particle-Source-In-Cell and Cloud-In-Cell models) assume that the average motion of the cloud is governed by only the average interphase momentum difference between the carrier and disperse phases, neglecting subscale perturbations.  In [Davis, Jacobs, Sen Udaykumar, “SPARSE—A subgrid particle averaged Reynolds stress equivalent model: testing with a priori closure”, Proc. Roy. Soc. A, 473, 2017] we presented a Cloud-In-Cell (CIC) formulation for particle-laden flows that accounts for cloud size and subgrid-scale stresses using averaging techniques, and for cloud deformation using methods from continuum mechanics. By expanding the particle drag correction factor to include fluctuating terms and Reynolds averaging the full particle momentum equation, the so-called SPARSE model accounts for the effect of particle variances within the cloud. Here, we propose a closure of the SPARSE model using expansion techniques at the averaged center location of the particle  cloud. The model is shown to accurately predict particle dispersion of clouds for a number of test cases, including a stagnation flow, an ABC flow and an isotropic turbulence flow.   

A DNS study of dispersion and deposition of particles in turbulent particle-laden channel flow – Effect of grid resolution

Direct numerical simulations of airflow with different grid resolutions and Lagrangian particle tracking were performed to investigate the particle dispersion and deposition in a turbulent channel flow. Different test cases with particles Stokes numbers varying from 2 to 100 were simulated.  The upward and downward flow configurations were considered for each case. The Saffman lift force is implemented in the in-house Eulerian-Lagrangian solver, and the importance of shear lift on particle dispersion and deposition was assessed. Fluid and particle statistics include the mean and fluctuating velocity profiles for fluid and particles, the average concentration of different size particles, and the particle deposition velocities were presented. The particles deposition velocities for different cases were validated against the available experimental data.  Particular attention was given to the influence of grid resolution on particle deposition.   

Not Participating

The concentration and agglomeration of uncharged particles in turbulent flows is a fundamental scientific problem that was investigated for decades. Recently, we developed beyond the state-of-the-art numerical methods to capture in detail the influence of electrostatic charges and the flow field on particle dynamics. Using the new simulation tool pafiX, we discovered the influence of charges carried by the particles on their preferential location in wall-bounded turbulent flows. In particular, we found that for most flow conditions charges of the same polarity counter-act the tendency of particles to accumulate close to the walls, which is called turbopheresis. However, for high charge levels, the concentration at the walls decreases. The concentration peak detaches from the walls and moves toward the bulk of the flow. Moreover, electrostatic charges attenuate the vortical motions of particles which are typical for square duct flows. In this talk, our latest discoveries will be reviewed. Additionally, we will present new results regarding the effect of bipolar charging on particle concentration.   

Numerical simulations of polydisperse particles in multiphase flows using population balance models and computational fluid dynamics

In this study, we investigate the formation of hydrate particles and their influence on the multiphase flow dynamics from a meso- and macro-scale perspective via a population balance model (PBM) and computational fluid dynamics (CFD) simulations. The formulation of the PBM accounts for nucleation, diffusion-induced growth, and Brownian and turbulent aggregation. The PBM is solved by means of the quadrature method of moments. Analyses of the evolution of characterising parameters of the dispersed phase, such as the reconstructed particle size distribution, number of particles, and volume fraction, are presented together with the solution of the flow field. We elucidate the explicit connections of the problem formulation with the molecular scale and the potential opportunities to link molecular dynamics simulations with higher-scale models. Finally, the extension of this PBM-CFD framework to investigate polydisperse multiphase flows in other applications (e.g. manufacturing of silver nanoparticles) is highlighted.   

Effect of the good solvent nature in flash nano-precipitation via population balance modelling and computational fluid dynamics coupling approach

The effect of the good solvent nature on polymer nano-particles (NP) formation in flash nano-precipitation is here investigated through a combined population balance model-computational fluid dynamics approach (PBM-CFD). Four good solvents are considered: acetone, acetonitrile, tetrahydrofuran and tert-butanol and the different resulting mean NP size is predicted in terms of mean radius of gyration via the Flory law of real polymers. Good solvents effects are here modelled in terms of solute–solvent interactions, using the Flory–Huggins theory and the Hansen solubility parameters. In this way, kinetics and thermodynamics are intertwined in a unique modelling tool. Our results show that the proposed methodology is able to predict the role played by the different good solvents, analysing single factors at the time. More specifically, the dynamics of mixing is decoupled from the dynamics of aggregation achieving a deeper insight into the fundamental fluid properties which affect the final NP size, pointing out the main mechanisms involved and showing a good agreement with experimental data.   

Stickslips in steady shearing as a Siganature of Solid-Fluid transition

We report in our recent work the occurrence of shear-rate dependent stickslips as a distinctive signature of solid-fluid transition for granular flows [10.1103/PhysRevLett.126.128001]. With additional measurements on the tribology between particles, we identify the speed-dependent lubrication to be one fundamental cause for such transition, and propose a new dimensionless number as the control parameter in describing the change of behaviors. Further investigations in more generic cases, including the use of a velocity-weakening friction in numerical experiments, provide further confirmation on such picture, and allow us to visualize the development of stickslips that have not been predicted by existing theories.