Detecting Lagrangian Coherent Structures in Realistic Experimental FLuid-Flow Data Sets
Detection of coherence in sparse Lagrangian particle tracking (LPT) data is a crucial step towards increasing our understanding in a vast number of biological, engineering and geophysical flows. This thesis explores two Lagrangian-invariant approaches for coherent-structures detection in fluid flows using LPT data. The first tested technique is the Coherent-Structure Colouring (CSC) algorithm (Schlueter, 2017). The performance of this Lagrangian approach is assessed by comparing CSC-coloured tracks with baseline vorticity fields of tracks past an Ahmed reference body and in a swirling jet. The effects of two normalized parameters on the identification of vortical structures were defined and studied: the mean track length; and the mean inter-particle distance. It was found that the CSC algorithm yielded accurate detection of coherent structures when mean inter-particle distances were smaller than 15% of the characteristic length scale of the flow, whereas results quickly deteriorate for sparser Lagrangian data. In the second study, a robust coherent-structure detection framework based on Voronoi tessellation and techniques from spectral graph clustering was developed. In this novel approach, the neighbouring times of two particles, defined as the period of time Voronoi cells of two particles remain connected by a Voronoi edge, is adopted as a metric for coherence. Particles for which Voronoi cells share a common Voronoi edge for longer periods of time present a smaller distance in the higher-dimensional eigenspace, hence presenting coherent motion. The proposed approach was shown to be successful at identifying coherence with realistic, sparse LPT data from large wind tunnel experiments. Specifically, coherent structures are identified for the first time with inter-particle distances on the order of 100% of the characteristic length scale of the flow.