Experimental characterization of two-phase liquid-liquid dispersions in rotating flows through neural-network assisted PTV and PIV techniques

This work investigates the drop dynamics in two-phase oil-water dispersions subjected to a rotating flow. In this complex condition, the flow behavior is influenced by drag, lift, centrifugal, Coriolis forces, which cause the drops to rotate, deform and break-up. The motion of both phases was studied with flow visualization. In the first tests, images acquired with high-speed camera (HSC) enabled an analysis on the physics of o/w dispersions. The emulsion formation was observed during transient tests by varying oil fraction. Then, a particle tracking velocimetry (PTV) tool was developed in the scope of machine learning and convolutional neural networks. It was applied to track drops and compute diameters and velocities at different flow rates. In the second set of tests, particle image velocimetry (PIV), together with processing routines in MATLAB and Python, generated velocity and turbulence fields related to the water flow. These results further explained the drop dynamics, which is influenced by characteristics of the water phase. These findings are relevant since they can validate hypotheses and help improve models or numerical simulations for multiphase flows in rotating devices.