Numerical models of multiphase flows with sticky particles

Abstract

Plugging caused by adhesive particles remains an important problem in multiphase fluid dynamics for decades and needs extensive theoretical research. This work aims to investigate this problem using a coupled Computational Fluid Dynamics-Discrete Element Method (CFD-DEM) approach, which allows for detailed tracking of particles and their interactions. The study is methodically divided into four main stages, each corresponding to a research paper.

The initial phase introduces the application of CFD-DEM to simulate the hydraulic transportation of non-cohesive glass beads in a pipe. The simulations were conducted after experimental data using the commercial software Star-CCM+. This step is essential to evaluate how well the model predicts changes in the flow regime and validate these predictions against experimental results.

The primary goal of this research is to simulate plug formation in a multiphase flow with sticky ice particles. Before progressing to this complex stage, the study’s second phase focuses on the cohesive collisions of individual particles. For this purpose, sticky ice particles immersed in a subcooled oil phase are simulated. New data on the collisional dissipation of energy and the ice particle coefficient of restitution were obtained using three methods for velocity measurement: high-speed experimental video recording, Positron Emission Particle Tracking (PEPT), and numerical simulations. The cohesive collision process was simulated by considering particle cohesion, size, and shape, providing information on the mechanical properties of particles for the following research steps.

The third stage holds key importance, applying the CFD-DEM model to simulate plugging in a slurry of ice in decane, using the foundational information gained from the previous stages. This part of the research includes a detailed parametric evaluation, focusing on the changes in the flow system’s characteristics, particularly the Reynolds number, particle concentration, and surface energy. This section also compares the simulation results with flow maps based on experiments.

Finally, the fourth stage presents the application of a machine learning classifier to predict blockages at a given flow regime, indicating a relatively new and developing approach in this field. A random forest classifier was applied using both experimental and simulation data. Experimental data were obtained from a lab-scale flow loop with ice slurry in decane, and the simulations are based on the CFD-DEM method. The results of this research include a flow regime map with blockage formation boundaries and its changes with variations in cohesion.

This thesis contributes not only to a better understanding of plugging in multiphase flow with sticky particles but also offers practical insights through a CFD-DEM approach. Additionally, it demonstrates the potential of advanced predictive models based on machine learning.