Access & Terms of Use
embargoed access
Embargoed until 2024-10-25
Copyright: Xie, Zhouzun
Embargoed until 2024-10-25
Copyright: Xie, Zhouzun
Altmetric
Abstract
The polydisperse solid-liquid system has been practised in many chemical engineering applications. A fundamental understanding of complex multi-phase flow with a wide particle size distribution (PSD) in the system is beneficial for process control and reactor optimisation, yet the currently existing numerical models, including conventional computational fluid dynamics - discrete element method (CFD-DEM), fail to capture the cross-scale inter-phase/particle interactions. Accordingly, multi-resolution models are developed in this thesis for the high-fidelity simulation of polydisperse solid-liquid systems.
1) A smoothed volume distribution model (SVDM) is first developed based on the unresolved CFD-DEM framework, with the capability of simulating the polydisperse solid-liquid system with a coarse-to-fine particle size ratio of up to 20. Via studying the migration of fine particles in suspension flow through a packed bed of coarse particles, the migration mechanism of fine particles is proposed and the inherent fundamental of clusters are elucidated. Via investigating the bed hydrodynamics in a bi-disperse solid-liquid fluidised bed (SLFB), the segregation and mixing mechanisms of particles in solid-liquid systems are illuminated. Via quantifying the solid transportation behaviours during the rapid filtration of dual-media filters, a probabilistic model is derived and verified for predicting clogging performance. This work establishes an effective framework to handle complex polydisperse solid-liquid systems.
2) Two acceleration methods (i.e., coarse-grained method and machine learning method) are studied, with the capability of simulating solid-liquid systems with improved computational efficiency at spatial and temporal scales, respectively. The coarse-grained method is employed to simulate large-scale particulate systems for unveiling the sedimentation mechanism of particles in water. The machine learning method is used to predict mixing and segregation behaviours in a solid-liquid system. This work provides an efficient method to predict granular flow behaviours in solid-liquid systems.
3) Further, a hybrid CFD-DEM model combining the resolved and unresolved CFD-DEM frameworks is originally developed, with the capability of simulating the polydisperse solid-liquid system with unlimited coarse-to-fine particle size ratios, for the first time. A resolved part obtains the fluid details around each coarse particle without extra models using fine grids (i.e., grid size to particle diameter ratio, lm/dp < 1/10), an unresolved part describes the fluid-fine particle interactions with empirical correlations using coarse grids (lm/dp > 3), and a semi-resolved part denotes the medium particle behaviours with a kernel-based approximation using medium grids (1/10 < lm/dp < 3). This work delivers a novel idea for modelling cross-scale solid-liquid flow and has the potential application to any polydisperse solid-liquid systems.
This thesis represents collection of a suite of innovative numerical works of polydisperse particulate flows in solid-liquid systems and provides a range of numerical tools for understanding and optimising polydisperse solid-liquid flow systems.