An engineered biological filtration (biofiltration) process treating a nutrient-enriched, low-alkalinity, organic-laden surface water downstream of conventional coagulation-clarification and upstream of an ultrafiltration (UF) membrane process was assessed for its treatment effectiveness. The impact of biofiltration pretreatment on UF membrane performance was evaluated holistically by investigating the native source water chemistry and extending the analysis into the drinking water distribution system. The biofiltration process was also compared in treatment performance to two alternative pretreatment technologies, including magnetic ion exchange (MIEX®) and granular activated carbon (GAC) adsorption. The MIEX®, GAC adsorption, and biologically active carbon (BAC) filtration pretreatments were integrated with conventional pretreatment then compared at the pilot-scale. Comparisons were based on collecting data regarding operational requirements, dissolved organic carbon (DOC) reduction, regulated disinfection byproduct (DBP) formation, and improvement on the downstream UF membrane operating performance. UF performance, as measured by the temperature corrected specific flux or mass transfer coefficient (MTC), was determined by calculating the percent MTC improvement relative to the existing conventional-UF process that served as the control. The pretreatment alternatives were further evaluated based on cost and non-cost considerations. Compared to the MIEX® and GAC pretreatment alternatives, which achieved effective DOC removal (40 and 40 percent, respectively) and MTC improvement (14 and 30 percent, respectively), the BAC pretreatment achieved the lowest overall DOC removal (5 percent) and MTC improvement (4.5 percent). While MIEX® relies on anion exchange and GAC relies on adsorption to target DOC removal, biofiltration uses microorganisms attached on the filter media to remove biodegradable DOC. Two mathematical models that establish an empirical relationship between the MTC improvement and the dimensionless alkalinity to substrate (ALK/DOC) ratio were developed. By combining the biofiltration results from the present research with findings of previous studies, an empirical relationship between the MTC improvement versus the ALK/DOC ratio was modeled using non-linear regression in Minitab®. For surface water sources, UF MTC improvement can be simulated as a quadratic or Gaussian distribution function of the gram C/gram C dimensionless ALK/DOC ratio. According to the newly developed empirical models, biofiltration performance is optimized when the alkalinity to substrate ratio is between 10 and 14. For the first time a model has thus been developed that allows for a predictive means to optimize the operation of biofiltration as a pretreatment prior to UF membrane processes treating surface water.

Membrane technology has been used in water purification for decades. However, membrane fouling remains a limiting factor. One way to control fouling is through surface modification. Several studies report that increasing surface hydrophilicity can reduce membrane fouling. Surface modification via physical coating (i.e., thin-film composite membrane) was explored in this research to prevent membrane fouling. Before making thin-film composite membranes, it was important to study structure/property relations in a series of potential coating materials. This research aims to contribute to a better fundamental understanding of the structure/property relations which govern water transport, rejection of model foulants (i.e., emulsified oil droplet or protein), and fouling characteristics in hydrogels based on poly(ethylene glycol) diacrylate (PEGDA) and N-vinyl-2-pyrrolidone (NVP). Crosslinked poly(ethylene glycol) (PEG) free-standing films were prepared by UV-induced photopolymerization of PEGDA crosslinker in the presence of varying amounts of water or monofunctional poly(ethylene glycol) acrylate (PEGA). The crosslinked PEGDA films exhibited polymerization induced phase separation (PIPS) when the water content of the prepolymerization mixture was greater than 60 wt%. Visible light absorbance measurements, water uptake, water permeability, and salt kinetic desorption experiments were used to characterize the structure of these phase-separated, crosslinked hydrogels. The films with PIPS exhibited a porous morphology in cryogenic scanning electron microscope (CryoSEM) studies. Dead-end filtration experiments using deionized water and bovine serum albumin (BSA) solutions were performed to explore the fundamental transport and fouling properties of these materials. The total flux of pure water through the films after prior exposure to BSA solution was nearly equal to that of the as-prepared material, indicating that these PEGDA films resist fouling by BSA under the conditions studied. Crosslinked NVP free-standing films were prepared by UV-induced photopolymerization in the presence of water, with NVP as the monomer and N,N’-methylenebisacrylamide (MBAA) as the crosslinker. A series of crosslinked films were polymerized at various prepolymerization water contents, NVP/MBAA ratios and at various levels of UV light intensity in the polymerization. Like PEGDA, the NVP films also underwent phase-separation during polymerization. The influence of monomer/ crosslinker ratio, prepolymerization water content, and UV intensities on membrane morphology and water transport was characterized with CryoSEM, bio-atomic force microscope (Bio-AFM) and dead-end filtration. Molecular weight cutoff (MWCO) measurements were used to characterize the sieving property of crosslinked NVP films polymerized at different UV intensities. UV intensity was found to have an impact on the interconnectivity of crosslinked membranes. Finally, tests of fouling resistance to protein solution (bovine serum albumin) and oily water emulsion were performed. The NVP crosslinked films had good protein and oily water fouling resistance. Overall, both crosslinked PEGDA and NVP films exhibit fouling resistance to oily water emulsions or protein solution. NVP films had more porous structure and higher water permeability than did PEGDA films, while the more compact structure of PEGDA films led to better rejection of model foulants (e.g., protein) than in NVP films. Based on different applications (e.g., oil/water separation, protein filtration), different coating materials must be chosen according to the membrane morphology, transport property, and rejection of model foulants to achieve the highest water flux and foulant rejection in membranes used for water purification. / text

The exploitation of lignocellulosic materials with the aim of producing high value-added products will potentially counteract concerns related to the depletion of fossil resources or exponential population growth. Bioethanol produced from lignocellulosic agriculture residue exhibits promising alternative to the petroleum-based fossil fuel which reduces net emission of greenhouse gases (GHG). But, due to certain technological barriers, the large scale production of lignocellulosic bioethanol has not been successfully commercialized. In this thesis, membrane filtration as an energy efficient separation process with low environmental impact was chosen with a possibility of improvement. Interconnected multi-staged microfiltration submerged membrane bioreactors (MBRs) set-up has been applied in order to separate suspended solids, obtain high concentration of yeast inside the bioreactor, and recover particle-free ethanol stream in a continuous high productivity process. The MBRs were effectively optimized comparing to different constant permeate fluxes of 21.9 LMH, 36.4 LMH, and 51 LMH. Moreover, membrane bioreactor performed effectively at low flux 21.9 LMH up to 262 h comparing to other applied fluxes. During continuous hydrolysis, membrane showed the capability of lignin recovery nearly 70% of medium SS content in all applied flux. Although the conversion rate of total sugars by concentrated cells were similar, yeast cells proved the capability of inhibitor tolerance, and to co-utilize 100% of glucose and up to 89% of xylose, resulted in bioethanol volumetric productivity of 0.78 g ethanol/l per hour 1.3 g ethanol/l per hour and 1.8 g ethanol/l per hour for 21.9 LMH, 36.4 LMH, and 51 LMH respectively. Moreover, the effect of different factors such as filtration flux, medium quality and backwashing on fouling and cake-layer formation in submerged MBRs during continuous filtration was thoroughly studied.

Les eaux usées peuvent possiblement être traitées par un système membrane intégré et combinant les procédés de bioréacteur à membrane (BAM) et d’osmose inverse (OI) pour une élimination efficace des micropolluants en vue de la réutilisation des eaux. Cependant, le rejet des concentrats d’OI dans l’environnement pourraient représenter un danger en raison de la toxicité de certains de leurs composés (micropolluants, sels, matières organiques). Une des solutions possibles peut être de recycler le concentrat d’OI vers le BAM. Néanmoins, une étude approfondie s’impose pour une telle configuration car le recyclage mettrait en jeu la recirculation de matière organique non biodégradable, ou de fortes concentrations en sels ou micropolluants, qui pourraient finalement engendrer, directement ou indirectement, un colmatage de la membrane ainsi qu’une modification de l’activité bactérienne dans le BAM. Les effets du recyclage de concentrat d’OI sur les performances de BAM ont été étudiés de deux différentes manières, en distinguant les effets à court-terme (ou court temps de contact) et les effets à long-terme (ou long temps de contact). Les résultats montrent qu’après un temps de contact de 3 heures entre le concentrat et les boues, les concentrations en protéines et polysaccharides dans le surnageant restent inchangées par rapport au début de l’opération. Une analyse HPLC-SEC a permis d’étudier les effets du concentrat d’OI sur la production de matières microbiennes solubles de types protéique. Un pic de concentration en substances protéiques ayant une masse moléculaire de 10 à 100 kDa a été observé dans le surnageant juste après l’addition du concentrat d’OI. Le pouvoir colmatant des boues n’a lui pas été modifié après l’injection du concentrat d’OI. Cette observation ouvre sur la possibilité de développer une opération d’OI comme traitement tertiaire en aval du BAM. La combinaison BAM-OI pourrait donc être une solution envisageable pour traiter le concentrat d’OI. Pour les longs temps de contact, les résultats ont montré que l’impact de l’effluent toxique (concentrat d’OI) sur les boues dépendait du rendement de l’opération d’OI et des caractéristiques du concentrat. Les mêmes tendances ont été observées quelle que soit la composition du concentrat en sels et en matière organique, puisqu’une augmentation de la concentration en protéine a été mise en évidence. L’effet du recyclage du concentrat d’OI a aussi été étudié à différents débits et avec différentes caractéristiques. Les effets sur les performances globales du BAM ainsi que sur son colmatage ont plus particulièrement été investigués. Le taux d’abattement en termes de Demande Chimique en Oxygène (DCO) est, dans tous les cas, supérieur à 93 %, quel que soit le débit de recyclage. Des résultats similaires ont été obtenus en termes de Carbone Organique Dissous. De plus, l’efficacité de la nitrification n’a pas été affectée en présence de concentrat d’OI dans le BAM. L’analyse HPLC-SEC a révélé un pic important de concentration en composés protéiques dans le surnageant, avec des masses moléculaires comprises entre 10 et 100 kDa et entre 100 et 1000 kDa. Par conséquent, une augmentation significative du pouvoir colmatant des boues a été observée et attribuée à la présence de protéines. Par ailleurs, le recyclage du concentrait d’OI n’a pas eu d’effet sur l’élimination de la carbamazépine et du diclofenac dans le BAM. Au contraire, l’élimination du ketoprofene a légèrement baissé, en passant de 94 à 72 %. Enfin, l’effet du recyclage de concentrat d’OI sur la biodégradation a été révélé comme insignifiant, ce qui indique que le recyclage du concentrat d’OI pourrait être une bonne alternative pour réduire les concentrats d’OI et limiter leur rejet dans l’environnement. / Wastewater effluents can be treated by an integrated membrane system combining membrane bioreactors (MBR) and reverse osmosis (RO) for effective removal of micropollutants in the field of high-quality water reuse. However, discharging the RO concentrate waste stream directly into the natural environment could lead to serious problems due to the toxic components contained in the concentrates (micropollutants, salts, organic matter). A possible solution could be the recirculation of RO concentrate waste to the MBR. However, such an operation should be studied in detail since the recirculation of non-biodegradable organic matter or high concentrations of salts and micropollutants could directly or indirectly contribute to MBR membrane fouling and modification of the biodegradation activity. The effects of RO concentrate recirculation on the MBR performances were investigated in two different ways of contact, i.e. short term peak contact and long-term continuous contact at various operating conditions. The results demonstrated that after 3 hours of contact time between the sludge and concentrate, the same values of both protein and polysaccharide concentrations were found in the supernatant, compared to that at the beginning of the reactor. HPLC-SEC analysis was employed to study the effects of RO concentrate on the production of protein-like SMPs. A significant peak of protein-like substances with a molecular size of 10-100 kDa was observed immediately in the supernatant after the addition of RO concentrate. Besides, no significant change was found of the sludge fouling propensity after the injection of RO concentrate into the activated sludge. This finding proposes the opportunities to develop RO process as a tertiary treatment of the membrane bioreactor (MBR), hence, the integrated MBR - RO concept with the RO concentrate recirculation to the MBR might be a solution to treat the concentrate waste stream produced by RO. During the long-term continuous contact, the results demonstrated that the impact of the toxic flow on activated sludge depends on the recovery of the RO step and the characteristics of the concentrate but the same trends were observed whatever the organic matter and salt contents of the concentrates: the concentration of proteins increased. The effects of the reverse osmosis concentrate recirculation, at different flow rates and with different characteristics, to the MBR were investigated. Their impacts on MBR global performances, especially the MBR fouling were evaluated. The removal efficiencies of chemical oxygen demand (COD) at the different flow rates of concentrate were greater than 93%. Similar results for the dissolved organic carbon removal efficiency were found in the MBR. Additionally, the presence of RO concentrate in the MBR did not inhibit the nitrification process. HPLC-SEC analysis employed to study the effects of RO concentrate on the production of protein-like SMPs demonstrated a significant peak of protein-like substances corresponding to 10-100 kDa and 100-1000 kDa molecules in the supernatant. Thus a significant increase of sludge fouling propensity was observed, which could be attributed to the increased quantity of protein-like substances. Furthermore, the recirculation of RO concentrate to the MBR did not significantly affect the removal of carbamazepine and diclofenac in the MBR. Meanwhile, the removal rate of ketoprofen was impacted slightly by the RO concentrate recycling to the MBR (from 94 to 72%). Finally, the effect of the concentrate on sludge activity was studied and no significant effect was observed on biodegradation, indicating that the return of the concentrate to the MBR could be a good alternative for the reduction of concentrate quantities before disposal to the environment.

This dissertation investigates the influence of surface heterogeneities on colloid deposition. First, deposition of colloidal particles on a nanofiltration membrane during cross flow membrane filtration was studied under different operating pressures and solution chemistries. An atomic force microscope (AFM) was then used to observe the deposit morphology formed on the membrane. At the initial stages of fouling, more particles preferentially accumulate near the peaks than in the valleys of the rough nanofiltration membrane surface. This study demonstrates that it is difficult to isolate, correlate and assess the effects that physical (roughness) heterogeneity and chemical heterogeneity has on colloid deposition based on experiments involving surfaces where the physical and chemical heterogeneities are uncorrelated or randomly distributed. In the second phase of the study, the deposition of model colloidal particles onto patterned charge-heterogeneous surfaces was studied both experimentally and theoretically. Controlled charge heterogeneity was created experimentally employing self assembled monolayers of alkanethiols patterned onto gold substrates using a soft lithographic technique. Model colloidal particles and fluorescent nanoparticles were sequentially deposited onto the patterned substrate under no flow (quiescent) conditions, and the deposited structures and the micro-patterns were imaged in situ using a combination of phase contrast and fluorescence microscopy. This study indicates that particles tend to preferentially deposit at the edges of the chemically favourable stripes. The theoretical investigation involved the formulation of a mathematical model based on Random Sequential Adsorption (RSA). This study showed that a simple binary probability distribution assumed in the model is able to predict the experimental deposit morphology adequately, particularly the periodicity of the underlying patterns on the substrate. Furthermore, the effect of charge heterogeneity on the electrostatic double layer interaction between a particle and a charge heterogeneous planar surface was studied numerically employing a 3D finite element model. In this system, significant lateral forces at close separation distances were observed, and found to be appreciably higher when the particle is near the edge of a heterogeneous region of the substrate. From the above studies, it can be concluded that by altering/controlling the chemical heterogeneity of the substrate, it is possible to achieve significant control on the resulting deposit morphology.

No description available.

Malgré une longue histoire de développement, l'hémodialyse (rein artificiel) possède encore quelques limitations, telles que la perte des propriétés initiales de la membrane en cours de traitement à cause du colmatage et la mauvaise élimination des toxines urémiques de taille moyenne. La présente étude fait partie d'un projet européen nommé BioArt dont le but est d'apporter des solutions à ces limites. Dans cet objectif, l'un des partenaires du projet a proposé le développement d'un nouveau concept de membrane double couche au sein de laquelle sont incorporées des particules adsorbantes. Une caractérisation complète de cette nouvelle membrane était alors nécessaire, plus précisément l'impact de la matrice mixte sur l'élimination des toxines urémiques de divers groupes devait être évalué, ainsi que la propension du matériau membranaire à se colmater. Les études des phénomènes de colmatage sont classiquement menées à l'échelle macroscopique (faisceau de fibres creuses) sans analyse à l'échelle d'une fibre isolée. Le but premier de la présente thèse a alors été de proposer un dispositif permettant une étude du colmatage membranaire induit par la protéine à l'échelle microscopique. Un dispositif microfluidique transparent dans lequel la membrane polymère est insérée a été élaboré et mis en œuvre pour la filtration des protéines modèles : l'albumine de sérum bovin (BSA) et l'a-lactalbumine. Grâce au couplage avec la microscopie de fluorescence, différents modes d'adsorption des protéines sur la surface de la membrane ont été observés et liés aux variations des conditions hydrodynamiques à l'intérieur de la puce. Il a été constaté, sous certaines conditions, une différence dans l'accumulation de protéines entre l'entrée, le centre et la sortie du canal tandis que dans d'autres conditions cet effet s'annule. En outre, un phénomène inattendu, l'agrégation de l'a-lactalbumine, a été observé au cours de la filtration. La localisation dans le canal et la forme des agrégats dépendent également des conditions hydrodynamiques et de la pression transmembranaire appliquée. Dans le but d'optimiser la conception de la membrane vis à vis de son aptitude à éliminer des molécules de taille moyenne de la circulation sanguine, un modèle mathématique a été proposé. L'objectif du modèle était, en prenant en compte la présence de particules adsorbantes à l'intérieur de la membrane double couche, de rendre compte de la combinaison des trois mécanismes d'élimination du soluté : la convection, la diffusion et l'adsorption. Le modèle permet de prédire l'influence de divers paramètres tels que la diffusivité de la molécule, l'épaisseur de la membrane, la présence de la convection, la charge en particules adsorbantes, sur l'intensification des flux à travers la membrane. Le modèle semble être un outil utile pouvant être appliqué à l'optimisation de membranes pour l'élimination des toxines. / Despite a long history of development, the hemodialysis procedure (artificial kidney) still possesses some limitations, such as loss of the initial properties of the membrane due to fouling and poor removal of the middle sized uremic toxins. The present study is part of an European project named BioArt the aim of which was to overcome these limitations. In that objective, one of the partners of BioArt project reported on the development of the novel promising concept of double layer membrane with embedded adsorptive particles. A thorough characterization of the new membrane was then necessary, more precisely the extent to which mixed matrix layer can improve the removal of the uremic toxins from various groups needed to be evaluated, as well as the propensity of the membrane material to become fouled. The studies of the fouling phenomena are conventionally performed at the macro scale (bundle of hollow fibers) without insights of what is happening at the scale of an isolated fiber. Therefore, the primary aim of the present Thesis was to transfer the research of the protein-induced membrane fouling from the macro to the micro scale. A novel transparent microfluidics device with the polymeric membrane inside has been developed and applied for the filtration of model proteins: bovine serum albumin (BSA) and a-lactalbumin. Thanks to the coupling of the microchip with the fluorescent microscopy, different patterns of protein deposition on the membrane surface were observed and related to the variations in the hydrodynamic conditions inside the microchip. It was found that at certain conditions one may observe the difference in protein accumulation in the inlet, the middle, and the outlet of the channel while at other conditions this effect vanishes. Additionally, the unexpected phenomena of a-lactalbumin aggregation was observed over the course of filtration. The location and shape of the aggregates were also dependent on the hydrodynamic conditions and the applied transmembrane pressure. Aiming to address the problem of membrane design optimization for the enhancement of the middle molecules elimination from the bloodstream, a mathematical model, which accounts for the presence of adsorptive particles inside the complex double-layer membrane, has been proposed. The objective of the model was to understand the interplay of three solute removal mechanisms: convection, diffusion, and adsorption. The model allows predicting the influence of various parameters such as molecule diffusivity, membrane thickness, the presence of convection, content of adsorptive particles on the flux intensification across the membrane. The developed model seems to be a useful tool, which may be applied to design optimized membranes for the removal of toxins.

La microfiltration est un procédé baro-membranaire largement utilisé pour stabiliser et clarifier les jus de fruits ou pour concentrer leur fraction pulpeuse. Toutefois le colmatage membranaire limite la productivité économique du procédé. Même si de nombreuses études ont été menées pour comprendre et limiter le colmatage, aucun véritable outil n'est aujourd'hui disponible pour prévoir la filtrabilité d'un jus de fruits. Le développement de ce type d'outils prévisionnels est pourtant essentiel pour faciliter l'optimisation des conditions de filtration et pour une meilleure maitrise du colmatage membranaire. Dans ce contexte, l'étude a été axée sur l'identification de nouveaux critères de filtrabilité des jus de fruits au travers de la compréhension et de la caractérisation des mécanismes de colmatage qui interviennent.En choisissant le jus d'orange comme modèle, une approche statistique a permis, dans un premier temps, une sélection rapide et pertinente des caractéristiques physico-chimiques des jus de fruits directement corrélées à leur filtrabilité, à savoir la matière sèche totale, l'extrait sec soluble, le pH, la conductivité et la granulométrie. Ensuite, des stratégies expérimentales complémentaires basées sur des filtrations à l'échelle du laboratoire ont été proposées pour identifier les différentes fractions colmatantes en fonction de leur taille ainsi que les mécanismes de colmatage associés. Ces stratégies novatrices ont permis d'étudier le pouvoir colmatant du jus d'orange en prenant en compte toute sa complexité et de proposer par la suite des cartographies de composés colmatants. Ces cartographies s'avèrent particulièrement intéressantes pour anticiper le comportement colmatant des jus de fruits sans avoir recours à des filtrations longues et coûteuses à grande échelle. Les résultats ont montré que le phénomène le plus important qui régit le colmatage membranaire global au cours de la filtration du jus d'orange semble être le colmatage externe par dépôt. Les objectifs de la dernière partie du travail, ont été de considérer plus fortement le caractère complexe et hétérogène des jus de fruits, pouvant les conduire à des changements brutaux de comportement lors de leur filtration. Pour cela des mesures rhéologiques en mode dynamique couplées à une stratégie d'isolement spécifique des différentes fractions colmatantes du jus d'orange ont été menées pour déterminer le comportement viscoélastique de la fraction insoluble du jus d'orange au cours de sa concentration. Les observations ont montré que les énergies des interactions entre les particules insolubles sont des fonctions croissantes de la concentration de la fraction insoluble et dépendent de la taille des particules. De plus, la transition vers un comportement « solide » du jus d'orange semble être favorisée par la présence des fractions particulaire et supra-colloïdale. Ces observations sont cohérentes avec le mécanisme de colmatage prédominant dû à l'accumulation de ces fractions à la surface de la membrane. La confrontation de nouvelles études de filtrabilité à l'échelle du laboratoire avec la performance de la filtration à grande échelle permettra de généraliser les stratégies proposées pour prédire la performance de la filtration de nouveaux jus de fruits. / Microfiltration is a pressure driven process successfully used to clarify and stabilize fruit juices or to concentrate their pulpy fraction. However, membrane fouling remains the critical factor governing the overall economic productivity of this process. Even if, many studies have been made to understand and limit membrane fouling, to date, there stills a lack of predicting tools to evaluate the filterability of fruit juices. However, it is interesting to develop practical and efficient tools for the prediction of fruit juices filterability in order to adapt and optimize filtration conditions to the fouling behavior of the filtered juice. In this respect, this study focused on the identification of new criteria of fruit juices filterability through the comprehension and the characterization of the potential fouling mechanisms.First, orange juice was chosen as test matrix and a statistical approach was used to select simple and relevant physico-chemical characteristics of this juice in relation with its filterability, namely dry matter, TSS, pH, conductivity, and particles size. Later, additional experimental strategies based on filtration tests at lab-scale were developed to identify the relevant size-classes of foulant compounds and related fouling mechanisms. These innovative strategies allowed studying the fouling potential of orange juice with the consideration of its complexity and proposing cartographies of foulant compounds. These cartographies are particularly interesting to anticipate the fouling behavior of fruit juices without any costly filtration operation at large scale. Moreover, results showed that external fouling seems to be the predominant fouling mechanism governing the overall membrane fouling during orange juice filtration. Finally, the aim of the last part of the work was to highlight the role of the complex and heterogeneous nature of fruit juices in their unexpected behavior during their filtration. For this purpose, the viscoelastic behavior of orange juice suspended solids during their concentration was obtained from rheological measurements in dynamic mode coupled with specific isolation of orange juice foulant fractions. Results showed that the energy of the interactions between juice particles increased as the suspended solids concentration increased and depends on the particles size. Furthermore, the solid-like behavior of the juice was enhanced by the presence of supra-colloids and large particles. These observations are in accordance with the predominant fouling mechanism that is due to the accumulation of these fruit fractions on the membrane surface. The confrontation of new filterability studies at laboratory scale with the performance of large-scale filtration will enable to generalize the proposed strategies to predict the performance of filtration of new fruit juices.

Membrane Distillation (MD) is an emerging sustainable desalination technique. While MD has many advantages and can be powered by solar thermal energy, its main drawback is the low water production rate. However, the MD process has not been fully optimized in terms of its manipulated and controlled variables. This is largely due to the lack of adequate dynamic models to study and simulate the process. In addition, MD is prone to membrane fouling, which is a fault that degrades the performance of the MD process. This work has three contributions to address these challenges. First, we derive a mathematical model of Direct Contact Membrane Distillation (DCMD), which is the building block for the next parts. Then, the proposed model is extended to account for membrane fouling and an observer-based fouling detection method is developed. Finally, various control strategies are implemented to optimize the performance of the DCMD solar-powered process. In part one, a reduced-order dynamic model of DCMD is developed based on lumped capacitance method and electrical analogy to thermal systems. The result is an electrical equivalent thermal network to the DCMD process, which is modeled by a system of nonlinear differential algebraic equations (DAEs). This model predicts the water-vapor flux and the temperature distribution along the module length. Experimental data is collected to validate the steady-state and dynamic responses of the proposed model, with great agreement demonstrated in both. The second part proposes an extension of the model to account for membrane fouling. An adaptive observer for DAE systems is developed and convergence proof is presented. A method for membrane fouling detection is then proposed based on adaptive observers. Simulation results demonstrate the performance of the membrane fouling detection method. Finally, an optimization problem is formulated to maximize the process efficiency of a solar-powered DCMD. The adapted method is known as Extremum Seeking (ES). A Newton-based ES is designed and the proposed model is used to accurately forecast the distilled water flux. Although good results are obtained with this method, a practical modification to the ES scheme is proposed to enhance the practical stability.

Membrane bioreactors (MBRs) have proven to be an extremely effective wastewater treatment process combining ultrafiltration with biological processes to produce high-quality effluent. However, one of the major drawbacks to this technology is membrane fouling – an inevitable process that reduces permeate production and increases operating costs. The prediction of membrane fouling in MBRs is important because it can provide decision support to wastewater treatment plant (WWTP) operators. Currently, mechanistic models are often used to estimate transmembrane pressure (TMP), which is an indicator of membrane fouling, but their performance is not always satisfactory. In this research, existing mechanistic and data-driven models used for membrane fouling are investigated. Data-driven machine learning techniques consisting of random forest (RF), artificial neural network (ANN), and long-short term memory network (LSTM) are used to build models to predict transmembrane pressure (TMP) at various stages of the MBR production cycle. The models are built with 4 years of high-resolution data from a confidential full-scale municipal WWTP. The model performances are examined using statistical measures such as coefficient of determination (R2), root mean squared error, mean absolute percentage error, and mean squared error. The results show that all models provide reliable predictions while the RF models have the best predictive accuracy when compared to the ANN and LSTM models. The corresponding R2 values for RF when predicting before, during, and after back pulse TMP are 0.996, 0.927, and 0.996, respectively. Model uncertainty (including hyperparameter and algorithm uncertainty) is quantified to determine the impact of hyperparameter tuning and the variance of extreme predictions caused by algorithm choice. The ANN models are most impacted by hyperparameter tuning and have the highest variability when predicting extreme values within each model’s respective hyperparameter range. The proposed models can be useful tools in providing decision support to WWTP operators employing fouling mitigation strategies, which can potentially lead to better operation of WWTPs and reduced costs. / Thesis / Master of Applied Science (MASc)