The severe rise in global surface temperature and its associated damages to the ecosystem are primarily attributed to anthropogenic carbon dioxide emissions. The present CO2 levels in the air today and the continuing emissions necessitate even the most ambitious development scenarios to rely on net negative CO2 emissions to limit the global temperature rise to 1.5 °C until the year 2100. Direct Air Capture (DAC) refers to the direct extraction of carbon dioxide from the ambient air and addresses the need for net negative emissions. As the technology is still in its infancy, understanding and modeling the occurring processes is crucial.
This thesis addresses the challenges in modeling adsorption and desorption processes in solid-sorbent Direct Air Capture, focusing on the complexities introduced by the porous structures of sorbent materials and the effects of humidity. Current models fail to incorporate the full spectrum of mass transfer processes, or do not consider the significant effects of humidity on the process. To bridge this gap, a detailed mass transfer and reaction kinetics model at the particle scale was developed, aiming to provide a more accurate framework for predicting the performance of amine-functionalized sorbents in DAC applications. The model was developed on a particle scale, and subsequently included in contactor models of a packed-bed and a monolith. The study considered all relevant transport mechanisms, finding that pore diffusion and reaction kinetics are critical in governing sorbent uptake dynamics. The contactor model effectively compares the performance of packed-bed and monolith reactors, highlighting a tradeoff between productivity and energy efficiency, with significant potentials for energy savings offered by the monolith.

Medical oxygen is crucial for effective treatment of patients, yet many low- and middle-income countries (LMICs) without a good healthcare infrastructure have been unable to provide sufficient oxygen therapy for years now. The COVID-19 pandemic highlighted the critical need for more oxygen, leading to supply shortages and rising prices even in developed countries. Currently, most medical oxygen is produced by a few large companies using centralized Cryogenic Air Separation Units (CAS). Pressure Swing Adsorption (PSA) offers a promising alternative, enabling local production of medical oxygen without reliance on suppliers. PSA leverages the pressure dependency of adsorption isotherms to selectively adsorb nitrogen, concentrating the oxygen.

This research has investigated the possibilities of designing a system that not only meets the specific oxygen demands of a Dutch hospital but also complies with the Dutch legal frameworks and forms an economically feasible case. A case study at Reinier de Graaf hospital (RdGG) in Delft used real oxygen consumption data. The applicable regulatory frameworks have been obtained by research in literature and with help of a representative at RdGG. For the technical design and optimisation, a numerical model of the PSA-cycle has been developed. This model is based on the Ideal Adsorption Solution Theory (IAST) and makes use of the Linear Driving Force model for adsorption kinetics. The model behaviour has been validated by experiments within a self-built, small-scale PSA-unit in the Process & Energy (P&E) lab. After this, the setup was scaled to hospital size within the model. A schematic PSA-plant has been designed based on the regulatory frameworks and system requirements which came from the numerical modelling. The economic viability has been addressed based on quotes, consumption data and electricity prices which were provided by RdGG.

Dutch regulations require hospitals to have three oxygen sources, each capable of independently providing the full design flow, with PSA units obligated to produce 93% oxygen concentration (± 3%) and meet strict impurity limits. To achieve the required oxygen purity, a vacuum pressure of 0.05 bar and an adsorption pressure of 6 bar with Oxysiv MDX zeolite are optimal. Depressurization should occur close to atmospheric pressure (1.05 bar) for efficiency. The adsorption column is most effective with a diameter-to-length ratio of 0.1 and a dimensionless time, tc, of 9.56 seconds for maximum oxygen production. The dimensionless time represents the time it takes for the gas to pass through the length of the column based on its inlet velocity.
The PSA unit at RdGG must handle a normal demand of 10 Nm³/h and a peak demand of 80 Nm³/h for three consecutive days. The second source can handle peak demand for three days, while the backup source can manage only 9 hours (tvital). If the PSA-storage has been depleted and the second source fails, tvital is the maximum refill time of the PSA-storage. This requires a production rate of 0.40 mol/s and two packages of storage cylinders (1.6 m³) which store the oxygen at 200 bar. This setup necessitates two 336L columns (3.5m height, 0.35m diameter, 0.36 m/s inlet velocity) and an air inlet flow of 126 m³/hr, supported by a vacuum pump and a compressor. The cycle steps for the adsorption, depressurization and vacuum (tAD = 79 s, tDP = 67 s, tVC = 112.1 s) total 258.7 s, allowing two columns to run in parallel. A 3 m³ air pressure vessel at 10 bar maintains PSA operation for three minutes during compressor disturbances. This setup achieves 93.61% oxygen purity with 41% recovery for 93% oxygen concentration, increasing to 58% recovery for 90% oxygen concentration. Each column requires 127 kg of zeolite, producing 5.75 mol O₂ kg⁻¹h⁻¹ (0.66 m³ O₂ kg⁻¹h⁻¹ at adsorption pressure), with an energy consumption of 29.38 kJ per mole of oxygen, totalling 255 kWh/tO₂.

The total investment for a PSA plant is estimated at 215,190 EUR, with fixed equipment costs of 176,800 EUR. Using straight-line depreciation over 20 years results in annual depreciation of 9,890 EUR. Annual costs also include energy and maintenance.
For an oxygen demand of 234 tons in 2023, the annual cost would be 25,590 EUR using PSA, significantly less than the current 60,840 EUR with the Liquefied Oxygen (LOX) installation. The levelized cost of oxygen over a period of 20 years for PSA is 128.61 EUR/ton, which is 2.1 times lower than the LCOO of LOX which is 270.52 EUR/ton.

The implementation of PSA plants for localized oxygen production in the Netherlands is both technically and economically feasible. A case study at Reinier de Graaf hospital shows significant cost savings compared to current liquid oxygen setups, with potential annual savings of 35,250 EUR. Required equipment is available and the setup size allows it to be installed inside or next to the hospital. Regulatory frameworks are already present and define clear requirements for using a PSA system as primary source. When implementing the right flow and pressure mechanisms the PSA plant can be connected to the current second and backup source, minimizing system adjustments for a smooth integration.

Future research should focus on several key areas to enhance the feasibility of using PSA in Dutch hospitals. Regulatory frameworks need approval or revision by experts in medical oxygen production. Analysing zeolite degradation and validating its performance are crucial for simulating the PSA unit's lifecycle. Improved modelling techniques, considering temperature variations, research on the validity of current assumptions and preventing backflow in the experimental setup will yield more accurate results. Additionally, industry practices should be reviewed to refine the proposed setup and obtaining precise cost estimates from quotes or hospital bills will improve the economic analysis.

The objective of this research is to reduce fines inside a novel comminution machine. Fines are detrimental to the energy efficiency of the comminution process and induce additional wear. An experimental setup was developed and built, the slice rig (SR), to obtain air flow measurements and mass rates which are used to determine the energy efficiency of the fines extraction.

Magnetic Particle Tracking (MPT) is a relatively new non-invasive measurement technique which is often used to study dense granular flow. Its basic principle relies on tracking the movement of a single magnetic tracer by means of measuring the magnetic field strength at a suitable distance from the tracer. By assumption of a magnetic dipole and the use of minimization techniques, both location and orientation of the tracer can be determined. MPT is therefore uniquely suited for the study of non-spherical particles. The performance of the localization is largely dependent on the signal-to-noise ratio and very often relies on nonlinear optimization techniques, as the definition of the magnetic field generated by a dipole is highly nonlinear and has five degrees of freedom. In this paper, we present a semi-algebraic solution by decoupling the estimation of the position and orientation in separate algebraic solutions. The two estimates are mutually dependent, necessitating an iterative approach between the two. The main benefits of this new approach is in the speed and robustness of the algorithm, which are much higher than for the classical constrained nonlinear optimization techniques.

@en

Hydrogen has emerged as a promising candidate for energy storage, offering an alternative to fossil fuels as a primary fuel source. It can be safely stored in a liquid organic hydrogen carrier (LOHC) and recovered upon demand through the reversible hydrogenation/dehydrogenation of the LOHC. Voyex, a technology-driven startup, is developing a novel LOHC that can be loaded onto ships and dehydrogenated onboard, providing hydrogen for ship propulsion as an alternative to fossil fuels. The successful implementation of this new technology requires a dehydrogenation reactor that may provide the conditions necessary for LOHC dehydrogenation to meet the ship's power demand.

This work develops a computational fluid dynamics (CFD) model to simulate the LOHC dehydrogenation reaction within a cocurrent upward-flow fixed-bed reactor. While lab-scale studies have examined the effect of operating conditions on the dehydrogenation of specific LOHCs, only a few works have studied LOHC dehydrogenation reactors and the inner flow details have yet to be understood. A CFD model will allow for a better understating of the flow, temperature, and species distribution within the reactor, and help gain more insight into the design and operating parameters on the reactor performance as measured by conversion and hydrogen yield.

This thesis delves into the distinct phenomena at play within an LOHC dehydrogenation reactor including hydrodynamics, heat transfer, species transport, and reaction kinetics, discussing ways to incorporate them into a numerical model. These effects are brought together in an Eulerian multiphase CFD model, as this approach is found to provide a good balance between modelling accuracy and computational demand.

The result of this research is a computationally inexpensive Eulerian multiphase CFD model capable of adapting to various LOHC systems and reactor configurations. Through multiple sensitivity analyses, the impact of design variables including reactor dimensions, temperature, flow rate, and catalyst pellet diameter on reactor performance is explored. These analyses yield valuable insights and design parameters to enhance reactor efficiency. Additionally, these analyses lead to several design improvement proposals such as raising the reactor inlet temperature, using a low inlet LOHC mass flux, incorporating gas-liquid separation methods, adding internal heating mechanisms, and employing internal baffles to disrupt flow and enhance heat transfer. The findings underscore potential for reactor design improvements, thereby proving the CFD model as a research and design tool for packed-bed multiphase reactors.

Particle-laden pipe flows are ubiquitous in industrial applications. Examples are industries like dredging, slurry transport, and the transport of reactants and products in chemical industries. One of the most important factors in these transport processes is the friction coefficient which relates directly to the pumping power which is a significant parameter for industries when viewed from an economic standpoint. The migration of particles in dense suspensions can significantly impact the friction coefficient in pipe flow. The clustering of particles in the core can lead to a reduction in the friction factor compared to a well-mixed particle suspension. This reduction is attributed to the decreased effective viscosity near the pipe wall. While the development lengths of single-phase flows are well known, limited knowledge exists regarding the development of velocity and concentration profiles in suspension flows.

The goal of this thesis is to study the development of neutrally buoyant suspension pipe flows. The experimental setup was validated using pressure drop measurements and the entrance length for single-phase pipe flow was obtained. The experiments for suspension flows involved varying the suspension Reynolds number and volume fractions keeping the particle size constant. Ultrasound imaging technique is used to circumvent the opacity of the suspension to study the development of suspension pipe flow. The inlet conditions for concentration were characterized, obtaining a uniform distribution at the inlet. Measurements are conducted at various locations downstream of the pipe and the velocity of the dispersed phase is obtained using ultrasound imaging velocimetry. Additionally, insights into the development of the concentration profile are obtained by checking the convergence towards a fully developed intensity profile, despite the fact that the image intensity doesn’t directly correlate to concentration profiles. The velocity profiles and intensity profiles were analyzed to understand the effect of radial migration on both concentration and velocity profiles. The entrance lengths for concentration and velocity were obtained for volume fractions ranging from 0.17 to 0.25 and suspension Reynolds numbers ranging from 500-2000. The results obtained revealed that the entrance length for concentration was greater than the entrance length for velocity. Scaling of the concentration entrance length with suspension Reynolds number and volume fractions were determined, and suspension Reynolds number scaled with an exponent of -1.62 and volume fraction scaled with an exponent of -2.1. This implies that the entrance length decreases with an increase in suspension Reynolds number and volume fraction. However, no definite trend was observed for the velocity entrance length.

Lubrication forces play a major role in the behaviour of fluid–solid systems, where they affect the collisions between particles. Current implementations of lubrication forces in unresolved simulations often suffer from shortcomings, such as neglecting parts of the physics or relying on arbitrarily defined parameters. In this short communication, we propose a novel implementation, rigorously defined based on physical and numerical factors. Both particle roughness and deformation are considered, and the model accuracy is demonstrated through comparison with experimental results.

@en

The hydrodynamics of the Expanded Bed Adsorption process is studied through simulations combining Computational Fluid Dynamics and the Discrete Element Method. A representative base case is defined, based on process design parameters commonly encountered in literature. Then, 19 other cases are defined, each representing a singular adjustment to the column design, material properties, or operating conditions. The parameters that are varied are the expansion factor, liquid viscosity, bed aspect ratio, mean particle density, width of the particle density distribution, width of the particle size distribution, column taper angle, and column alignment angle. The impact of each adjustment on the bed behaviour is discussed, using the local particle size distribution and solids dispersion coefficient as main indicators of bed stability. Optimal performance was found for an expansion factor of two to three, and the combination of particle size distribution and particle density distribution was found to greatly improve bed stability. The mixing process of the liquid and solid phases is concluded to be of highly complex nature, and cannot simply be predicted from the liquid flow velocity.

@en

The blast furnace hearth plays an important role in the operational stability and lifetime of the reactor. The quasi-stagnant bed of coke particles termed the deadman undergoes complex interaction with the flowing hot metal, and remains largely ill-understood. In this work, a cold model blast furnace hearth is presented, and studied using both numerical and experimental techniques. Magnetic Particle Tracking (MPT) is used to investigate the individual particle behaviour within the cylindrical, opaque bed. At high liquid holdup, the particle bed was found to alternate between floating and sitting states, following the liquid level during the tapping and filling cycle. This bed motion was found to induce a migration of particles, thereby slowly renewing the deadman. The rate of horizontal migration increases with the vertical bed amplitude, and the renewal of particles is concentrated around the opening of the tap hole. No direct influence of the coke-free space on the tapping rate was found in these experiments. Instead, the disturbance of the packing in front of the tap hole was observed to lead to a higher tapping rate. Additionally, a coupled numerical framework is presented, in which Computational Fluid Dynamics (CFD), the Volume of Fluid (VOF) method and the Discrete Element Method (DEM) are combined. A simulation set-up is presented which closely replicates the experimental conditions. The position and movement of the floating bed are found to be well-predicted by the VOF/CFD-DEM model. Particle trajectories are presented, and migration of particles within the deadman is observed. Alongside the particle motion, the liquid flow pattern during draining of the vessel is visualised. It is concluded that a coke-free space underneath the deadman significantly impacts the shape of the liquid flow pattern, which affects the erosion processes within the blast furnace hearth.

@en

Liquid-solid systems are frequently encountered in industrial processes and it is broadly recognised that numerical simulations are a useful tool for gaining insight in these processes. In this study, the unresolved CFD-DEM approach is extended with a complete momentum coupling for liquid-solid flows. Established correlations are used for the drag and lift forces, while new implementations are introduced for the unsteady interaction forces. A virtual mass force model based on the work of Felderhof (Felderhof 1991) is introduced, which accounts for the local particle volume fraction and the liquid-solid density ratio. The Basset history force, which is usually neglected due to computational difficulties related to its implementation, is evaluated according to the approach proposed by Parmar et al. A liquid fluidised bed is used as a demonstration case for the extended model. In this work, it is shown that with appropriate stabilisation measures, the Basset history force is approximated accurately (within 5%), while computational efficiency is maintained ( < 30% increase in computational time). Furthermore, the relevance of the complete momentum coupling is demonstrated by analysis of the solids mixing in the liquid fluidised bed. It is shown that when accounting for the complete interaction force, solids mixing is up to 20% slower compared to simulations with the drag-only approach.

@en

The solubility and diffusivity of CO₂ in a series of 1-alkyl-3methylimidazolium tricyanomethanide ionic liquids ([C_nmim][TCM] with n = 2, 4, 6, 7, 8; ILs) was studied using a magnetic suspension balance at temperatures ranging from 298 to 353 K and pressures up to 2 MPa. The effects of temperature, pressure, and alkyl chain length on CO₂ solubility and diffusivity were examined. The electrolyte PC-SAFT (ePC-SAFT) equation of state was used to describe the solubility of CO₂ in the ILs. The Henry's law constant and the excess properties of solvation (Gibbs free energy, enthalpy, and entropy) were calculated. A series of equations derived from Fick's second law were evaluated, and a Fourier expansion of Fick's second law of diffusion was found to be the most suitable model for deriving diffusivities from gravimetric data. The diffusivities range from 10^-10 to 10^-9 m²·s^-1 in the temperature and pressure ranges applied. The activation energies for CO₂ diffusion (12-16 kJ·mol^-1) were found to be in the range of traditional solvents. (Graph Presented).

@en