A systematic methodology for robust identification of droplet breakage kernels for emulsification processes

Accurate modeling of emulsification processes within Population Balance Models (PBMs) for the prediction of the droplet size distribution (DSD) requires reliable identification of the breakage frequency kernel. This study investigates the identifiability and sensitivity of PBM parameters, model selection and dataset selection for emulsification, under a wide range of operating conditions characterized by Reynolds of the dispersed phase number and the Weber number. Frequentist and Bayesian optimization approaches are employed to estimate the parameters. The Bayesian approach permits also to quantify uncertainty distributions. Sensitivity and identifiability analyses are then conducted. Using a dataset based on fractional factorial experimental design is found to be satisfactory to identify parameter subsets that are robust and widely generalizable. The methodology also allows discrimination between the available breakage kernels based on their description of the experimental observations. This work provides a systematic methodology for ensuring reliable PBM application for emulsification processes.