• DBSCAN algorithm was adapted for spatiotemporal reconstruction of bubble trajectory. • Bubble motion in dilute swarms was studied. • Methodology was validated via volumetric mass conservation. • Immediate and collective values of bubble size, shape and velocity are described. • Bubble velocities in swarms exceed single-bubble theory due to liquid turbulations. Accurate characterization of bubble trajectories in swarms remains a significant challenge in multiphase flow research. This study introduces a tracking methodology optimized for high-speed imaging data based on the DBSCAN clustering algorithm (Density-Based Spatial Clustering of Applications with Noise), treating detected centroids as a spatiotemporal point cloud (x, y, t). The algorithm’s neighbourhood radius, ε , is physically derived from maximum instantaneous rise velocities, while a weighted normalization procedure ensures spatial sensitivity over extended capture times. The methodology was validated using new experimental data including measurements for water and two other aqueous solutions through volumetric mass balance and statistical convergence tests. Results indicate that even at low gas holdups (<0.5%), bubble rise velocities in clusters systematically exceed single-bubble predictions, likely due to wake-induced acceleration. This shows that even under conditions of very diluted swarm, the traditional correlation between bubble shape and rise velocity does not apply. It follows that trajectory-averaged dynamics are dominated by local turbulence and path instabilities, which exert a greater influence on bubble mobility than static shape-drag dependence alone.
Shirokov et al. (Fri,) studied this question.