A benchmark for particle-laden turbulent duct flow: A joint computational and experimental study

A turbulent duct flow laden with small heavy inertial particles (Re_τ ≈ 570, St⁺≈50, d_p⁺≈0.3, St ≈ 12, and d_p/η ≈ 0.17) is studied computationally and experimentally. We examine whether a long development section can be modeled using a short periodic domain. This simplification is not valid if the development section is too short (less than 25–50 duct-height in our case), the periodic simulation is not integrated long enough (less than O(1000) large-eddy turnover time in our case), or the effective mass loading ratio is not adjusted correctly (increased by a factor of 1.5 in our case). Additionally, we show that ignoring particle-particle collisions, even when the volume fraction is as low as 3.9×10⁻⁶, produces a large over-estimation of near-wall particle concentration (turbophoresis). The necessity of tailored post-processing of simulations for a one-to-one comparison against experiments is demonstrated. Namely, the finite thickness of the laser sheet and the optically-sampled volume size should be considered when post-processing simulations to reproduce the experimental measurement of clustering statistics. Experimentally and computationally, we show that an increase in the mass-loading ratio from 2.4% to 12% has a minimal effect on clustering, slightly lowers velocity fluctuations, and diminishes turbophoresis.