In the United States, nearly half a million bone graft operations are performed annually to repair defects arising from birth defects, trauma, and disease, making bone the second most transplanted tissue. Autogenous bone is the current gold standard for bone grafts; however it is in limited supply and results in a second injury at the donor site. A promising alternative is a tissue engineered bone graft composed of a biomaterial scaffold, pharmaceutics, and osteoprogenitor cells. One source of osteoprogenitor cells is bone marrow stroma, which can be obtained from the patient - minimizing the risk of an immune response - directed in vitro to proliferate, and differentiate into a bone-like tissue. To date, tissue engineered bone grafts have not been clinically effective; thus, strategies must be developed to improve efficacy. I hypothesize that to facilitate tissue healing in a manner similar to autogenous bone tissue engineering bone must possess a mineralized collagen matrix to support tissue integration, and angiogenic factors to stimulate vascular infiltration, and osteogenic factors to direct normal bone remodeling. I propose that these factors can be synthesized by osteoprogenitor cells in vitro when cultured under the appropriate conditions. Previous work has demonstrated that perfusion culture of osteoprogenitor cells within 3D scaffolds stimulates phenotypic markers of osteoblastic differentiation, but those studies did not determine whether the effects were a consequence of shear stress or increased nutrient availability. Consequently, this work has involved studies in a planar geometry, where nutrient effects are negligible. Three studies that characterize the effect of fluid flow on osteoblastic differentiation of osteoprogenitor cells are presented here. The objective of the first study was to determine the effect of shear stress magnitude on cell density and osteocalcin deposition. In this study, radial flow chambers were used to generate a spatially dependent range of shear stresses (0.36 to 2.7 dynes/cm2) across single substrates, and immunofluorescent techniques were used to assay cell phenotype as a function of shear stress. The objective of the second study was to determine the effect of the duration of fluid flow on cell density and phenotypic markers of differentiation. Here, parallel plate flow chambers were used to generate a single shear stress at the cell surface, and entire cell layers were assayed for expression of osteoblastic genes. The objective of the third study was to compare continuous and intermittent fluid flow strategies. In this study, a microprocessor-controlled actuator was added to the flow loop to periodically halt flow, and markers of mechanosensation and osteoblastic differentiation were measured. These studies demonstrated that shear stresses of 0.36 to 2.7 dynes/cm2 stimulate late phenotypic markers of osteoblastic differentiation but not cell proliferation. In addition, this osteogenic effect is sensitive to duration of fluid flow but insensitive to the magnitude of shear stress. Further, intermittent fluid flow enhances cell retention, biochemical markers of mechanotransduction, and synthesis of the angiogenic factor vascular endothelial growth factor (VEGF). Thus, these studies suggest that intermittent fluid flow may be an attractive component of a biodynamic bioreactor for in vitro manufacture of clinically effective tissue engineered bone grafts. Future studies will further investigate intermittent fluid flow strategies and three-dimensional studies with scaffolds suitable for bone tissue engineering. / Ph. D.

Glioblastoma is the most common and deadly primary brain tumor with an average survival of 15 months following diagnosis. Characterized as highly infiltrative with diffuse tumor margins, complete resection and annihilation of tumor cells is impossible following current standard of care therapies. Thus, tumor recurrence is inevitable. Interstitial fluid surrounds all of the cells in the body and has been linked to elevated invasion in glioma, which highlights the importance of this understudied fluid compartment in the brain. The primary objective of this dissertation was to identify specific interstitial fluid transport behaviors associated with elevated invasion surrounding glioma tumors. We first describe our methods to measure interstitial fluid flow in the brain using dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), a clinically used, non-invasive imaging modality. We highlight the versatility of the technique and the possibilities that could arise from widespread adoption into existing perfusion-based imaging protocols. Using this method, we examined transport associated with invasion in a murine GL261 cell line. We found that elevated interstitial fluid velocity magnitudes, decreased diffusion coefficients and regions with accumulating flow were significantly associated with invasion. We tested the validity of our invasive trends by extending our analysis to multiple, clinically-relevant tumor locations in the brain. Interestingly, we found invasion did not follow the same trends across brain regions indicating location-specific structures may drive both interstitial flow and corresponding invasion heterogeneities. Lastly, we aimed to manipulate flow by engaging with the meningeal lymphatics, an established pathway for interstitial fluid drainage. Over-expression of VEGF-C in the tumor microenvironment neither enhanced drainage nor altered invasion in comparison to our control, indicating other tumor-secreted growth factors, such as VEGF-A, may play a larger role in mediating flow and invasion. Taken together, by identifying specific transport factors associated with invasion, we may be better equipped to target and treat infiltrative tumor margins, ultimately extending survival in patients diagnosed with this devastating disease. / Doctor of Philosophy / Glioblastoma is the most common and deadly primary brain tumor with an average survival of 15 months following diagnosis. Characterized as highly infiltrative with diffuse tumor margins, complete resection and annihilation of tumor cells is impossible following current standard of care therapies. Thus, tumor recurrence is inevitable. Interstitial fluid surrounds all of the cells in the body and has been linked to elevated invasion in glioma, which highlights the importance of this understudied fluid compartment in the brain. The primary objective of this dissertation was to identify specific interstitial fluid transport behaviors associated with elevated invasion surrounding glioma tumors. We first describe our methods to measure interstitial fluid flow in the brain using dynamic contrast enhanced magnetic resonance imaging (DCE-MRI), a clinically used, non-invasive imaging modality. We highlight the versatility of the technique and the possibilities that could arise from widespread adoption into existing imaging projects. Using this method, we examined transport associated with cancer cell invasion in a mouse tumor cell line. We found that interstitial fluid speeds were elevated while diffusion was decreased in regions of invasion. Further, regions that had interstitial fluid flow congregation were significantly associated with invasion. We tested the validity of these invasive trends by extending our analysis to multiple, clinically-relevant tumor locations in the brain. Interestingly, we found invasion did not follow the same trends across brain regions, indicating location-specific structures may drive both interstitial flow and invasion differences. Lastly, we aimed to manipulate flow by engaging with the meningeal lymphatics, an established pathway for interstitial fluid drainage. Following administration of a meningeal lymphatic-relevant protein, we saw no changes in flow or invasion in comparison to our untreated control, indicating other tumor-secreted proteins may play a larger role in these responses. Taken together, by identifying specific transport factors associated with invasion, we may be better equipped to target and treat infiltrative tumor margins, ultimately extending survival in patients diagnosed with this devastating disease.

The lymphatic system plays essential roles in regulating fluid balance and immunosurveillance. Across the body, local lymphatic vessels collect waste in the form of lymph and deliver it to nearby lymph nodes (LNs) to be filtered and screened for pathogens. With broad implications in adaptive immunity, cancer metastasis, and cancer treatment, developing novel in vitro models will provide new platforms to explore lymphatic function in health and disease. This dissertation sought to develop tissue-specific engineered models of the LN stroma and the meningeal lymphatics to examine the transport of cells and fluid. Within the LN, fibroblastic reticular cells (FRCs) maintain a network of extracellular matrix conduits that guide varying rates of interstitial fluid flow (IFF) based on inflammatory state. Eventually, that flow exits the LN through the afferent lymphatics, consisting of lymphatic endothelial cells (LECs). We first developed a spatially organized model of the LN stroma consisting of a monolayer of LECs on the underside of a tissue culture insert and an FRC-laden hydrogel within. We demonstrate that high magnitude IFF (3.0 µm/s) had positive impacts on FRCs but disrupted the integrity of the LEC barrier, and these effects were accompanied by increased secretion of a variety of inflammatory chemokines. We also show that IFF of any magnitude decreased T cell egress from the model. Next, we sought to apply the LN stroma model toward understanding metastasis. LN metastasis is the most important prognostic factor in breast cancer, with size of metastasis informing treatment plan. Metastasis greatly alters the structure of the LN, which in turn alters transport. However, the impact of altered transport on cancer progression is not well understood. We added different numbers of breast cancer cells to our LN stroma model to examine tumor burden. We found that tumor cells invaded the LEC barrier at similar numbers regardless of initial burden. Additionally, at the highest tumor burden, diffusivity in the stroma was significantly decreased. Most excitingly, flow velocity was positively correlated with FRC spread in the hydrogel, demonstrating the contributions of FRCs to transport. Finally, we looked to the central nervous system (CNS). The meningeal lymphatics are responsible for draining cerebrospinal fluid to the cervical lymph nodes for CNS immunosurveillance. We developed a simple model of a meningeal lymphatic vessel lumen consisting of a monolayer of LECs on the underside of a tissue culture insert and a monolayer of meningeal fibroblasts within. This is, to our knowledge, the very first in vitro model of the meningeal lymphatics. We demonstrate that our model has barrier function and is capable of immune cell transmigration and egress. We examined how systemic chemotherapy for breast cancer could cause off-target disruption of the meningeal lymphatics and found that docetaxel was significantly deleterious. We further began to explore leukemia cell behavior in our LN stroma and meningeal lymphatics model. Throughout this dissertation, we emphasize the importance of incorporating fluid and cell transport into engineered models of immunity. These models represent a step toward building up the complexity of in vitro lymphatic models to improve pre-clinical screening and understand pathophysiology. / Doctor of Philosophy / The lymphatic system plays essential roles in regulating fluid balance and immune system surveillance. Across the body, local lymphatic vessels collect waste in the form of lymph and deliver it to nearby lymph nodes (LNs) to be filtered and screened for pathogens like viruses or bacteria. With broad implications in immunity, cancer metastasis, and cancer treatment, developing novel models in the lab using human cells and 3-dimensional biomaterials will provide new platforms to explore lymphatic function in health and disease. This dissertation sought to develop engineered models that were specific to the lymph node stroma and the meningeal lymphatics to examine the transport of cells and fluid. Within the LN, fibroblastic reticular cells (FRCs) maintain a network of channels that guide varying rates of interstitial fluid flow (IFF) based on how inflamed the LN is. Eventually, that flow exits the LN through the afferent lymphatics, consisting of lymphatic endothelial cells (LECs). We first developed a spatially organized model of the LN stroma consisting of LECs on the underside of a porous membrane and an FRC-laden hydrogel above the membrane and demonstrated that high magnitude IFF altered morphology, immune cell behavior, and inflammatory protein secretion in the model. Next, we sought to apply the LN stroma model toward understanding cancer metastasis. LN metastasis is the most important prognostic factor in breast cancer, with size of metastasis informing treatment plan. Metastasis greatly alters the structure of the LN, which in turn alters the transport of lymph and immune cells. However, the impact of altered transport on cancer progression is not well understood. We added different numbers of breast cancer cells to our LN stroma model to examine tumor burden and found that tumor cells invaded the LECs at similar rates regardless of initial density, but that diffusion, a transport parameter, was significantly changed by high tumor cell density. Finally, we looked to the central nervous system (CNS). The meningeal lymphatics are responsible for draining cerebrospinal fluid to the cervical lymph nodes to screen for pathogens in the CNS. We developed a simple model of a meningeal lymphatic vessel lumen consisting of LECs and meningeal fibroblasts on either side of a porous membrane. This is, to our knowledge, the very first in vitro model of the meningeal lymphatics. We examined how systemic chemotherapy for breast cancer could cause off-target disruption of the meningeal lymphatics and found that docetaxel was significantly damaging to the model. Throughout this dissertation, we emphasize the importance of incorporating fluid and cell transport into engineered models of lymphatics. These models represent a step toward building up complexity to improve the toolset for pre-clinical screening and studying disease progression.

Doubling transistor count for every two years in a computer chip, transmitter and receiver (T/R) module of a phased-array antenna that demands higher power with smaller dimensions are all results of miniaturization in electronics packaging. These technologies nowadays depend on improvement of reliable high performance heat sink to perform in narrower volumes. Employing microchannels or open cell metal foam heat sinks are two recently developing promising methods of cooling high heat fluxes. Although recent studies especially on microchannels can give a rough estimate on performances of these two methods, since using metal foams as heat sinks is still needed further studies, a direct experimental comparison of heat exchanger performances of these two techniques is still needed especially for thermal design engineers to decide the method of cooling. For this study, microchannels with channel widths of 300 &micro / m, 420 &micro / m, 500 &micro / m and 900 &micro / m were produced. Also, 92% porous 10, 20 and 40 ppi 6101-T6 open cell aluminum metal foams with compression factors 1,2, and 3 that have the same finned volume of microchannels with exactly same dimensions were used to manufacture heat sinks with method of vacuum brazing. They all have tested under same conditions with volumetric flow rate ranging from 0,167 l/min to 1,33 l/min and 60 W of heat power. Channel height was 4 mm for all heat sinks and distilled water used as cooling fluid. After experiments, pressure drops and thermal resistances were compared with tabulated and graphical forms. Also, the use of metal foam and microchannel heat sinks were highlighted with their advantages and disadvantages for future projects.

Past studies (primarily on steady state problems) that have compared the penalty and the velocity-pressure finite element formulations on a variety of problems have concluded that both methods yield solutions of comparable accuracy, and that the choice of one method over the other is dictated by which of the two is more efficient. In this work, we show that the penalty finite element method yields inaccurate solutions at large times on a class of transient problems, while the velocity-pressure formulation yields solutions that are in good agreement with the analytical solution. Numerical studies are conducted on various problems to compare these two formulations on the basis of rates of convergence, total number of equations to be solved and accuracy of results. We found that both formulations give almost the same rates of convergence in all problems, however the penalty formulation involves lesser number of equations than the velocity-pressure formulation due to implicit treatment of pressure field, and hence is more efficient. In some of the problems we have also compared a finite volume method with the penalty and velocity-pressure formulations on the basis of accuracy and computational cost.

In vitro mechanotransduction studies, uncovering the basic science of the response of cells to mechanical forces, are essential for progress in tissue engineering and its clinical application. Many varying investigations have described a multitude of cell responses, however as the precise nature and magnitude of the stresses applied are infrequently reported and rarely validated, the experiments are often not comparable, limiting research progress. This thesis provides physical and biological validation of a widely available fluid stimulation device, a see-saw rocker, as an In vitro model for cyclic fluid shear stress mechanotransduction. This allows linkage between precisely characterised stimuli and cell monolayer response in a convenient six-well plate format. Computational fluid dynamic models of one well were analysed extensively to generate convergent, stable and consistent predictions of the cyclic fluid velocity vectors at a rocking frequency of 0.5 Hz, accounting for the free surface. Validation was provided by comparison with flow velocities measured experimentally using particle image velocimetry. Qualitative flow behaviour was matched and quantitative analysis showed good agreement at representative locations and time points. A maximum shear stress of 0.22Pa was estimated near the well edge, and time-average shear stress ranged between 0.029 and 0.068Pa, within the envelope of previous musculoskeletal In vitro fluid flow investigations. The CFD model was extended to explore changes in culture medium viscosity, rocking frequency and the robustness to position on the rocking platform. Shear stress magnitude was shown to increase almost linearly with an increase in the viscosity of culture medium. Compared with 0.5 Hz, models at 0.083 and 1:167 Hz, the operational limits of the see-saw rocker, indicated a change in shear stress patterns at the cell layer, and a reduction and increase in mean shear stress respectively. At the platform edge at 0.5 Hz, a 1.67-fold increase in time-average shear stress was identified. Extensive biological validations using human tenocytes underlined the versatility of the simple In vitro device. The application of fluid-induced shear stress at 0.5 Hz under varying regimes up to 0.714Pa caused a significant increase in secreted collagen (p < 0.05) compared to static controls. Tenocytes stimulated at a shear stress magnitude of 1.023Pa secreted significantly less collagen compared to static controls. The potential for a local maximum in the relationship between collagen secretion rate and shear stress was identified, indicating a change from anabolic to catabolic behaviour. Collagen biochemical assay results were echoed with antibody stains for proteins, where a co-localisation of connexin-32 with collagen type-I was also identified. A custom algorithm showed that four hours of fluid-induced shear stress of 0:033Pa intermittently applied to tenocytes encouraged alignment and elongation over an eight day period in comparison to static controls. Primary cilia were identified in human tenocyte cultures and bovine flexor tendon tissue; however primary cilium abrogation In vitro using chloral hydrate proved detrimental to cell viability. Collaborative investigations identified that ERK signalling and c-Fos transcription factor expression peaked after the application of 0.012Pa at 0.083 Hz for 20 minutes and anabolic collagen gene expression relative quantities increased after 48 hours of rocking at 0.083 Hz. In conclusion, validated shear stresses within a six-well plate, induced by cyclic flow from a see-saw rocker, provides an exceptional model for the In vitro study of dynamic fluid shear stress mechanobiology. Biological investigations have been linked to precise applied shear stress, creating a foundation for understanding the complex relationship between tenocytes and fluid-induced shear stress In vitro. Using this model, research is repeatable, comparable and accurately attributed to shear stress, accelerating the scientific advancement of musculoskeletal mechanobiology.

Thesis (MScEng (Process Engineering))--Stellenbosch University, 2008. / A full scale, three dimensional, transient, mathematical model was developed to simulate fluid flow and mixing in gas stirred ladles. The volume of fluid (VOF) and discrete phase (DPM) models were used in combination to account for multiphase aspects, and a slightly modified version of the standard - model was employed for turbulence modelling. The model was validated to compare well against published physical modelling results. Model results were interpreted from the fundamental grounds of kinetic energy transport within the ladle. This approach led to the specification of three key measures of mixing efficiency: The rate and efficiency of kinetic energy transfer from the buoyant gas to the bulk steel as well as the total kinetic energy holding capacity of the ladle. These measures describe the quantity of mixing in any specific ladle setup, whereas the traditional measure of mixing time reflects mixing quality, i.e. the degree of kinetic energy distribution through the entire ladle. The model was implemented in designed experiments to assess various operating and design variables pertaining to mixing quantity and quality. Considerable time was invested in finding the correct balance between numerical accuracy and computational time so that the model could be used to generate the required data within the given time frame. Experiments on the operating variables drew an important distinction between factors influencing the shape and the strength of gas induced flow patterns. Flow pattern strengthening variables, such as gas purge rate, significantly affected the quantity of mixing, but had a limited effect on mixing quality. Variables such as mass loading that influence the shape of the flow patterns had much larger potential to influence both the quantity and quality of mixing. Minimization of turbulence losses in the region of the plume eye was identified as the primary outcome of ladle design. It was shown that a taller vessel allowed more distance over which the plume could disperse, thereby reducing velocity gradients and subsequent turbulence generation at the free surface. Multiple tuyere systems yielded similar improvements by dividing the gas flow into several weakened plumes. Surface wave formation was investigated as an added mixing mechanism and demonstrated to be impractical for application in full scale gas stirred ladles. The conditions for resonance between the surface wave and the bubble plume were met only in vessels with a very low aspect ratio. Performance improvements offered by swirl in these ladles could easily be replicated in more practical ways. This study demonstrated the potential of mathematical modelling as a tool for in-depth investigation into fluid flow and mixing in the hostile environment of a full scale gas stirred ladle. Scaled-down cold models are the only alternative and can offer no more than a reasonably reliable predictive framework. The ease of flow data extraction from the numerical model also proved invaluable in facilitating a fundamental understanding of the effects of various important independent variables on ladle hydrodynamics. At this stage of development, however, the model is recommended for use on a comparative basis only. Two important developments are required for complete quantitative agreement: The inclusion of turbulence modulation by the bubbles and the increased turbulence kinetic energy dissipation rate in the vicinity of the free surface. A general strategy was developed to account for these effects and it compared favourably with published cold model results. Further research is required to generalize this approach for application in full scale gas stirred ladles.

Thesis (MScEng (Process Engineering))--University of Stellenbosch, 2010. / ENGLISH ABSTRACT: Mass transfer efficiency in distillation, absorption and stripping depends on both thermodynamic efficiency and hydrodynamic behaviour. Thermodynamic efficiency is dependent on the system kinetics while hydrodynamics is the study of fluid flow behaviour. The focus of this thesis is the hydrodynamic behaviour in tray columns, which affects entrainment. In order to isolate hydrodynamic behaviour from the thermodynamic behaviour that occurs inside sieve tray columns, investigations are conducted under conditions of zero mass transfer. When the gas velocity is sufficiently high to transport liquid droplets to the tray above, entrainment occurs. The onset of entrainment is one of the operating limits that determines the design of the column and thus impacts on the capital cost. By improving the understanding of the parameters that affect entrainment, the design of the tray and column can be improved which will ultimately increase the operability and capacity while reducing capital costs. Existing correlations predicting entrainment in sieve tray columns are based on data generated mainly from an air/water system. Previous publications recommend that more testing should be performed over larger ranges of gas and liquid physical properties. An experimental setup was therefore designed and constructed to test the influence of the following parameters on entrainment: 1. gas and liquid physical properties 2. gas and liquid flow rates 3. tray spacing The experimental setup can also measure weeping rates for a continuation of this project. The hydrodynamic performance of a sieve tray was tested with air and water over a wide range of gas and liquid flow rates and at different downcomer escape areas. It was found that the downcomer escape area should be sized so that the liquid escaping the downcomer always exceeds a velocity of approximately 0.23 m/s in order to create a sufficient liquid seal in the downcomer. For liquid velocities between 0.23 and 0.6 m/s the area of escape did not have an effect on the percentage of liquid entrained. It was also established that entrainment increases with increasing gas velocity. The rate at which entrainment increases as the gas velocity increase depends on the liquid flow rate. As soon as the liquid flow rate exceeded 74 m3/(h.m) a significant increase in entrainment was noted and the gas velocity had to be reduced to maintain a constant entrainment rate. This is because the increased liquid load requires a longer flow path length for the froth to fully develop. The undeveloped froth, caused by the short (455 mm) flow path, then creates a non-uniform froth that is pushed up against the column wall above the downcomer. Consequently, the froth layer is closer to the tray above resulting in most of the droplets ejected from the froth reaching the tray above and increasing entrainment. By reducing the gas velocity, the froth height and ejecting droplet velocity is reduced, resulting in a decrease in entrainment. The results from the experiments followed similar trends to most of the entrainment prediction correlations found in literature, except for the change noted in liquid flow rates above 74 m3/(h.m). There was, however, a significant difference between the experimental results and the correlations developed by Hunt et al. (1955) and Kister and Haas (1988). Although the gas velocities used during the air/water experiments were beyond the suggested range of application developed by Bennett et al. (1995) their air/water correlation followed the results very well. The entrainment prediction correlation developed by Bennett et al. (1995) for non-air/water systems was compared with the experimental air/water results to test for system uniformity. A significant difference was noted between their non-air/water prediction correlation and the air/water results, which motivates the need for a general entrainment prediction correlation over a wider range of gas and liquid physical properties. Based on the shortcomings found in the literature and the observations made during the experiments it is suggested that the influence of liquid flow path length should be investigated so that the effect on entrainment can be quantified. No single correlation was found in the literature, which accurately predicts entrainment for a large range of liquid loads (17 – 112 m3/(h.m)), high superficial gas velocities (3 – 4.6 m/s) and different gas and liquid physical properties. It is therefore recommended that more work be done, as an extension of this project, to investigate the influence of gas and liquid physical properties on entrainment (under zero mass transfer conditions) for a large range of liquid (5 – 74 m3/(h.m)) and gas (2 – 4.6 m/s) flow rates. In order to understand the effect of droplet drag on entrainment, tray spacing should be varied and increased to the extent where droplet ejection velocity is no longer the mechanism for entrainment and droplet drag is responsible for droplet transport to the tray above. Since it is difficult and in most cases impossible to measure exact gas and liquid loads in commercial columns, another method is required to measure or determine entrainment. Since liquid hold-up was found to be directly related to the entrainment rate (Hunt et al. (1955), Payne and Prince (1977) and Van Sinderen et al. (2003) to name but a few), it is suggested that a correlation should be developed between the dynamic pressure drop (liquid hold-up) and entrainment. This will contribute significantly to commercial column operation from a hydrodynamic point of view.

No description available.

Nanofiltration (NF) can be used in water and wastewater treatment as well as water recycling applications, removing micropollutants such as hormones. Due to their potential health risk it is vital to understand their removal mechanisms by NF membranes aiming at improving and developing more effective and efficient treatment processes. Although NF should be effective and efficient in removing small molecular sized compounds such as hormones, the occurrence of adsorption onto polymeric membranes results in performances difficult to predict and with reduced effectiveness and efficiency. This study aims firstly at defining, understanding and quantifying the relevant filtration operation parameters and, secondly, in identifying the physical mechanisms of momentum and mass transfer controlling the adsorption and transport of hormones onto polymeric NF membranes in cross-flow mode. The hormones estrone (E1) and 17-b-estradiol (E2) were chosen as they have very high endocrine disrupting potency. The NF membranes used and tested were the NF 270, NF 90, BW30, TFC-SR2 and TFC-SR3 since they have a wide span of pore sizes. The first step is to experimentally acquire the knowledge of how fluid flow hydrodynamics and mass transfer close to the membrane affect hormone adsorption. The focus will be particularly on the effect of operating pressure, circulating Reynolds numbers (based on channel height, Reh) and hormone feed concentration. These hydrodynamic parameters play an important role in concentration polarisation development at the membrane surface. A Reh increase from 400 to 1400 for the NF 270 membrane caused the total mass adsorbed of E1 and E2 to decrease from 1.5 to 1.3 ng.cm-2 and 0.7 to 0.5 ng.cm-2, respectively. In contrast, a pressure increase from 5 to 15 bar yielded an increase in the adsorbed mass of E1 and E2 from 1.0 to 1.8 ng.cm-2 and 0.5 to 0.7 ng.cm-2, respectively. Moreover, increasing hormone feed concentration caused an increase in the mass adsorbed for both hormones. These observations led to the conclusion that adsorption is governed by the initial concentration at the membrane surface which, in turn, depends on the hormone feed concentration, operating Reh and pressure. Membrane retention, however, depends on the initial polarisation modulus, defined as the ratio between the initial concentration at the membrane surface and the initial feed concentration. The same trends were obtained for the TFC-SR2 membrane. However, this membrane has a much lower permeability compared to the NF 270 one (7.2 vs 17 L.h-1.m-2.bar-1, respectively) and concentration polarisation is less severe. The experimental variations in mass adsorbed and retention as a function of the operating filtration parameters (Reh and pressure) were therefore lower. Based on these experimental results, a sorption model was developed. This model predicts well both feed and permeate transient concentrations for both hormones and membranes (NF 270 and TFC-SR2) in the common range of operating pressures and Reh of spiral-wound membrane modules. The model was further applied for E2 in the presence of background electrolyte, yielding good predictions. These findings are an important advancement in determining which membrane would be more suitable to effectively remove hormones with a substantial reduction of experimental work. The above-mentioned developed model does not give insight into the phenomena occurring inside the membrane since it focuses on the feed conditions. However, membrane characteristics, such as material and pore radius were found to have an impact in adsorption and retention of hormones. It was found experimentally that polyamide, from which the active layer of the NF membranes is made, adsorbs three times more mass of hormone than any other polymers constituting the membranes. Since this active layer is the membrane selective barrier of the membrane that is in contact with the largest hormone concentration (due to concentration polarization in the feed solution) it is concluded that the active layer adsorbs most of the hormones. Further experimental work carried out in this thesis showed that increasing the pore radius from 0.32 nm to 0.52 nm increased the E2 mass adsorbed from 0.17 ng.cm-2 to 1.1 ng.cm-2 and decreased the retention from 88% to 34%. These results show that the wider the pore, the larger the quantity of hormone that penetrates (i.e. partitions) inside the membrane and, therefore, the more the membrane adsorbs the hormone. For membranes of similar pore radius, the membrane with larger internal surface area was found to adsorb more. All the previous results led to the establishment of a new model for the hormone transport inside the membrane pore taking convection, diffusion and adsorption into account. Since the differential equation describing the transport with adsorption inside the pore has no analytical solution, a numerical model based on the finite-difference approach was applied. With such a model, its validation against experiments and parametric studies it was possible to understand the transport mechanisms of adsorbing hormones through NF membranes. The results show that for low pressures the hormone transport is diffusion dominated. In contrast, for higher pressures (above 11 bar) the transport is convection dominated, showing that a purely diffusion transport model does not describe well the actual transport phenomena of hormones in NF membranes. Furthermore, it was found that two similar molecules can behave very differently in terms of adsorption on the membrane. E1, which adsorbs 20% more than E2 in static mode, being slightly smaller than E2, partitions more inside the membrane pore and adsorbs double under filtration conditions. This study contributes to illuminating the adsorption mechanisms of hormones onto NF membranes by understanding what parameters control adsorption such as hydrodynamics, materials, structure, etc. This not only identifies a potential problem in large scale applications, but it also provides an understanding of the mechanisms involved in the removal of these hormones and a tool that can be used to design future membranes for the improvement of micropollutant removal.