This thesis employs three-dimensional simulations combined with a hybrid front-tracking/level-set method to elucidate the complex interplay between flow dynamics and interfacial behaviours in a practical operational unit. Specifically, it examines the mixing of immiscible oil and water by a pitched blade turbine in a cylindrical vessel. The study focuses on the effects of impeller rotation frequency, the presence of surface active agents (surfactants) and their physiochemical properties on the evolution of interface and mixing performance, including interfacial area, dispersed drop number, and drop size distribution within the system.

Furthermore, due to the high computational cost of high-fidelity simulations, several data-driven models based on recurrent neural network are constructed using the simulation data and evaluated as an alternative approach for predicting dispersion performance metrics over multiple future time-steps based on their initial observations.

The results from this thesis provide a detailed picture of the underlying physics and interfacial dynamics within mixing systems as influenced by impeller speed, surfactant elasticity and solubility. This work bridges the gap in understanding of liquid-liquid dispersion process. Moreover, this thesis examines the applicability of several popular machine learning techniques in the field of multiphase flow mixing, highlighting the challenges and critical considerations involved in this process.