Planejamento de experimento computacional e aprendizagem estatı́stica aplicados em simulação euleriana refinada da colisão gota-partı́cula

Abstract

The droplet-particle collision is a common phenomenon in nature and industry, covering processes from the fluid catalytic cracking to the cleaning of dusty gases. Recently, it has brought more attention from the scientific community because of the growing demand for process control and optimization. In this way, this doctoral dissertation covers in two steps the research on droplet-particle collision; both use the Basilisk code for numerical simulation. The first step simulates the impact of a droplet onto a spherical particle in an initially quiescent environment. With the assistance of the design of computer experiment and machine learning, this step aiming at characterizing the lamella formed due to the collision, analyzing the interfacial flow by a mechanical energy viewpoint, and building a predictive statistical model as a function of the flow dimensionless parameters. In agreement with our mechanical energy analysis of interfacial flows for the analyzed parameters range, the Reynolds and Weber numbers show a negative and a positive effect on the lamella area, respectively. The second step of this work prospects the bidimensional simulation of the droplet collision against a cluster of particles. It objects to build a broad database as a basis for the phenomenon analyses, notably the particles arrangement effect over the liquid surface area and particles wet area, using deep learning models.