Inertial particle velocity and distribution in vertical turbulent channel flow: a numerical and experimental comparison

TL;DR

This paper compares numerical simulations and experimental measurements of inertial particle behavior in vertical turbulent channel flow, highlighting agreements and discrepancies at different volume loadings and emphasizing the need for further model refinement.

Contribution

It provides a direct comparison between two numerical models and experiments for inertial particles in turbulent flow, revealing areas for improvement in modeling particle clustering and wall interactions.

Findings

Numerical models agree with experiments at low volume loading.

Discrepancies increase at high volume loading, especially near walls.

Particle clustering is overpredicted in simulations compared to experiments.

Abstract

This study is concerned with the statistics of vertical turbulent channel flow laden with inertial particles for two different volume concentrations ($\Phi_{V} = 3 \times 10^{-6}$ and $\Phi_{V} = 5 \times 10^{-5}$) at a Stokes number of $St^{+} = 58.6$ based on viscous units. Two independent direct numerical simulation models utilizing the point-particle approach are compared to recent experimental measurements, where all relevant nondimensional parameters are directly matched. While both numerical models are built on the same general approach, details of the implementations are different, particularly regarding how two-way coupling is represented. At low volume loading, both numerical models are in general agreement with the experimental measurements, with certain exceptions near the walls for the wall-normal particle velocity fluctuations. At high loading, these discrepancies are increased, and it is found that particle clustering is overpredicted in the simulations as compared to the experimental observations. Potential reasons for the discrepancies are discussed. As this study is among the first to perform one-to-one comparisons of particle-laden flow statistics between numerical models and experiments, it suggests that continued efforts are required to reconcile differences between the observed behavior and numerical predictions.