Towards the digital extraction column: Online-monitoring and analysis of fluid dynamics in liquid-liquid extraction columns

• Successful application of multi-object tracking to drop swarms in extraction columns over image sequences • Online measurement of the dispersed phase hold-up and the drop size distribution in extraction columns • Column monitoring system enabling the early detection of column flooding • Extensive experimental study on fluid dynamics in pulsed sieve tray extraction columns • Parameter estimation for a drop swarm sedimentation model An effective monitoring system for liquid-liquid extraction columns must evaluate key fluid dynamic properties such as Sauter mean diameter, the hold-up of the dispersed phase, and the drop sedimentation velocity to accurately estimate the available mass transfer area and the solvent residence time. However, while many studies have focused on investigating the hold-up and the drop size distribution (DSD) studies on drop sedimentation remain scarce, often leading to its estimation based on the remaining fluid dynamic properties. In this work, we introduce a column monitoring system that enables a holistic assessment of the column operation based on all three fluid dynamic properties. For this purpose, we used the differential pressure method to determine the hold-up, and two telecentric camera setups to determine the Sauter mean diameter, and the drop sedimentation velocity. The camera images were processed by YOLOv8 for drop detection and the ByteTrack algorithm for drop tracking, achieving high accuracy on unseen data. In an extensive experimental study, we investigated the interdependency of the fluid dynamic properties at different operating conditions including flooding in a DN50 pulsed sieve tray extraction column. The obtained experimental data was used to parametrize a drop sedimentation model. Our findings indicate that assuming a constant swarm exponent in the model is inadequate, particularly at lower liquid loads.