Machine learning-enhanced PIV for analyzing microfiber-wall turbulence interactions

Abstract

A machine learning-based approach, RAFT-PIV, is used to measure with single-pixel resolution the flow field around a microplastic fiber in a turbulent channel flow at a Shear Reynolds number of 1000. The results reveal the interaction of the fiber with a hairpin vortex. The fiber rotation rate is correlated with slip velocity distributions along the fiber length, demonstrating higher rotation rates with increased slip velocity gradients. The fiber's alignment with the spanwise direction during its trajectory is explained through its progressive alignment with the head of a hairpin vortex, characterized by the swirling strength, shear strain rate, and local flow velocity. Higher fiber rotation rates were found likelier in the presence of a vortical structure. These findings highlight the potential of machine learning-enhanced PIV techniques to deepen our understanding of fiber-turbulence interactions, essential for applications such as microplastic pollution mitigation.