Fluidized beds for large-scale biomass gasification are among the most efficient and widely used reactor types for converting solid biomass into syngas or for biofuel production via Fischer-Tropsch or methanol synthesis. Several fluidized bed designs have focused primarily on conventional geometries, such as cylindrical and conical beds, typically with a distributor plate at the bottom. Many studies have been devoted to correlating the interplay of hydrodynamics and heat & mass transfer in fluidized beds, both from an experimental and computational point of view. Several direct correlations for full-scale reactors already exist. These are typically
only applicable to conventional reactor geometries. There is, though, a dearth of information on how other methods of fluidization, such as using nozzles/orifices at different locations within the bed, influence fluidization and the interplay between the gas and the solid phases. It is therefore needed to provide a better understanding of the physics in industrial-scale fluidized bed reactors by creating lumped models that sit between single correlations for full fluidized beds and detailed CFD simulations, and that can provide predictions in a matter of minutes rather than days or weeks.

The main aim of this research is to gain insight into the hydrodynamics of a novel conical, jet-spouted, bubbling fluidized bed using different bed materials, namely corn grits (polenta) and porous alumina (widely used in industries as a catalyst support material). The novel idea described in this thesis is to develop a well-validated model that can provide deeper insights into the hydrodynamics of such less-studied bed configurations, thus helping to improve their scaling and design. Furthermore, this can be used to improve lumped modelling work in future studies.

While lumped models are used extensively in the literature, they are also a source of confusion, especially for people entering the field who are looking for relatively fast models, because the term ‘lumped model’ has different meanings. Chapter 1 of this thesis, therefore, discusses the various lumped models available in the literature for fluidized bed reactors and critically analyzes the developments necessary to apply them to high-temperature processes.

In recent decades, Computational Fluid Dynamics (CFD) has become a popular tool for studying fluidized beds. The two-fluid model is one of the most popular modelling approaches, but it is unclear how reliable this type of modelling is for more challenging non-standard geometries. Hence, in Chapter 2, we delve into a numerical study of a novel, conical, multi-jet fluidized bed, with three different solvers used to predict the hydrodynamics. All three solvers use the same two-fluid model, the same parameters, and the same computational mesh, yet lead to different results, especially in the lower bed regions, highlighting the effects of differences in numerical discretization approaches. These differences indicate how these approaches preserve momentum injection, capture centerline spouting, and stabilize dense-phase transport. A good agreement of a solver with experiments on a simple, standard (cylindrical/conical with a distributor plate for fluidization) geometry does not guarantee agreement when the same solver is applied to a more complex geometry. This reinstates the well-known importance of experimental validation.

Experimental validation of the dynamics in such complex industrial geometries is often a challenge due to operational and computational limitations. Therefore, scaling down such beds becomes crucial. Hence, in Chapter 3, a scaled-down reactor is made from the scaling analysis of the larger reactor described in Chapter 2. In this chapter, three-dimensional CFD simulations of the reactor bed under various operating conditions using ANSYS Fluent are performed and validated using novel experimental techniques previously unexplored for bubbling fluidized beds, such as wire mesh sensors (WMS) and gamma-ray computed tomography (GCT). Simulations and experiments are performed for different operating conditions. The comparisons show good agreement for different physical parameters for different flow rates. However, some simulation cases show significant deviations from experiments, even for numerically better-resolved computational grids. While the results of a novel design of experiments for maldistribution also show good agreement with experiments performed, it is still too early to decide how reliable the measurements are because of experimental uncertainties and numerical inaccuracies.

To provide a good characterization of the hydrodynamics inside novel configurations and explore the uncertainties involved in them, two experimental methods in such novel configurations, WMS and GCT, are compared to each other in Chapter 4. These are used to obtain the time-resolved void fraction distribution. Used for both static and dynamic experiments, representative of different fluidization regimes, two bed-filling heights are compared. The results demonstrate the potential use of WMS in gas-solid flows and its capability to determine phase distribution.

Once a well-validated numerical model is obtained from novel experimental methods, validated with each other, the numerical and experimental data can be provided as input to parametrize or train a lumped model. Chapter 5 explores this for the novel fluidized bed configuration described in Chapter 2, using pressure sensor data obtained from experiments. This is done using an Artificial Neural Network (ANN), and subsequently for a Gaussian Process Regression (GPR). The idea is to show how Machine Learning (ML) models can be trained using data and how this could be applied to make predictions much faster than with experiments or CFD
simulations.

Lastly, Chapter 6 discusses a valorization study for plastic waste gasification conducted at KNUST, Ghana. Plastic waste has been handled in Ghana largely through the involvement of local governments and private-sector authorities. Many policies have been enacted in West Africa for plastic waste management. However, despite many policies being passed in the last three decades, both at the national and urban levels, there is still a strong focus on the inability to develop rural areas. Several processes, from collection to recycling, are discussed to examine the plastic waste collected and which processes use the most energy. It was observed that 15% of plastic waste is removed from landfills for further processing (pelletization, shredding, etc.). However, this remains a key hotspot in a technology chain in which plastics could be further utilized to produce high-value end products, such
as liquid or gaseous fuels. Furthermore, a quantitative assessment of local waste management infrastructure reveals that mechanical recycling, although operational, entails a 15% material loss rate during pelletization and relies heavily on a thermally
dominated electricity grid. These factors introduce a latent carbon footprint into the recycling value chain. Future strategies must therefore prioritize technologies capable of valorizing the non-recyclable fractions quantified in this study via thermochemical conversion, for example, via fluidized-bed gasification. ... Read more