Dans un procédé de filtration, un fluide traverse une membrane (barrière sélective). Une force motrice s’applique entre les deux côtés de la membrane qui peut être un gradient de pression, température ou un potentiel électrique/chimique. Dans les procédés de filtration par un gradient de pression, certains composés du milieu fluide, traversent la membrane alors que d’autres sont retenues sur la surface membranaire. Ces procédés sont très utiles dans différents domaines de l’industrie, notamment en ce qui concerne le traitement des eaux et des effluents, biotechnologie, agroalimentaire et pharmacie. En plus les procédés de filtration offrent des installations plus compactes avec une optimisation des coûts opérationnels comparant avec des procédés traditionnels de séparation notamment distillation et cristallisation. Par ailleurs, ces procédés se réalisent en absence des additifs chimique et changement de la phase. Dans cette étude, on se focalise sur les procédés de microfiltration. L’inconvénient principal de ces procédés est l’accumulation continue de particules/molécules sur la surface de la membrane. Ceci affecte la sélectivité de la membrane, modifie la qualité et la quantité de liquide passant à travers la membrane et conduit à une augmentation des coûts et de l’énergie. Le Colmatage (encrassement) membranaire se produit dans tous les types de procédés membranaires et par conséquent est connu le principal obstacle à l’utilisation répandue de ces procédés. Différentes techniques sont utiles pour surmonter les effets de l’encrassement de la performance de la membrane: le traitement physico-chimique des membranes utilisées, la modification des conditions opératoires (flux tangentiel de la solution d’alimentation sur la surface de la membrane est souvent appliqué pour réduire au minimum l’accumulation de particules), l’utilisation de membranes moins sensibles au colmatage, etc. Tout dépendant de la nature des solutions traitées, les particules déposées sont très variables. Les micro-organismes, des matières organiques naturelles notamment les protéines, les polysaccharides, les substances humides, les oxydes inorganiques et les sels contribuent au colmatage des membranes. Dans les dernières années, un grand nombre d’études expérimentales ont été investis pour comprendre les mécanismes de colmatage. Il a été souligné que les propriétés physico-chimiques de la membrane, la chimie des solutions et les conditions opératoires sont les trois principaux facteurs influant sur les mécanismes de colmatage. En parallèle, les modèles théoriques ont été proposés pour confirmer / décrire les observations expérimentales. La modélisation du colmatage membranaire est un outil essentiel pour évaluer les mécanismes qui le causent. Il permet également prédire la performance du système de filtration et par conséquent trouver des stratégies adaptées pour empêcher la modification de la performance membranaire pendant le procédé de filtration. En général, les modèles de classifient en deux grandes catégories: les modèles de transport de masse qui se concentrent sur le transport de solutés dans le procédé de filtration, et les modèles de colmatage basés sur le blocage des particules/molécules sur la surface ou à l’intérieur de la membrane. Dans la plupart des cas, les modèles dépendent fortement des paramètres empiriques ou semi-empiriques et restent phénoménologique. 1. Avoir une meilleure compréhension des mécanismes du colmatage membranaire lors de la filtration d’un milieu liquide contenant les micro-organismes en suspension. Il est important de souligner que des eaux industrielles et des eaux usées dans plusieurs domaines appartiennent à ce type d’effluents. 2. Proposer un modèle macroscopique décrivant les mécanismes de colmatage observés. [...] / During a membrane filtration process, a liquid medium is filtered through a membrane(selective barrier). The applied driving force between two sides of the membrane can be a gradient of pressure, temperature or a chemical/electrical potential.In pressure driven filtration processes (application of a pressure gradient as driving force between two sides of the membrane), certain components of the liquid medium pass through the membrane, while others are retained at the membrane surface. These processes are widely used as separation techniques in different industrial fields like waste water treatment, biotechnology, food and pharmacy. Compared to conventional techniquesof separation (distillation, crystallization, ...), membrane processes offer more compact installations with more optimized operational costs. Moreover, membrane processes are mainly performed in absence of chemical additives and phase change. In this work we focus on the pressure-driven microfiltration membrane processes.The main disadvantage of these processes is the continuous accumulation of particles on the membrane surface. This affects the membrane selectivity, modifies the quality and the quantity of the liquid passing through the membrane and leads to an increase of energy costs. Membrane fouling occurs in all types of membrane processes and therefore is known as the major obstacle for widespread use of these processes. Different techniques are used to overcome the effects of fouling on the membrane performance : physical-chemical treatment of used membranes, modification of the operational conditions (tangential flow of the feed solution to the membrane is often applied for minimizing the particle accumulation to the membrane surface), use of membranes less susceptible to fouling, etc. Depending on the nature of the treated solutions, the deposited particles are highly variable. Microorganisms, natural organic matter such as proteins, polysaccharides, humid substances, inorganic oxides and salts contribute notably to membrane fouling.It should be noted that membrane fouling problem is a multi-physics (hydrodynamics,mass transport, physics, chemistry), multi-scale (different length scales are involved:molecules, pores and membrane surface) and time dependent (evolution of the membrane microstructure and the molecule-surface interactions) phenomena.In the last decades, a huge number of experimental studies have been invested to understand fouling mechanisms. It has been pointed out that membrane physicochemical properties, solution chemistry and operational conditions are the three major factors affecting the fouling mechanisms. In parallel, theoretical models have been proposed to confirm/describe the experimental observations.Modeling of membrane fouling is an essential tool for assessing the fouling mechanisms. It helps predicting the membrane performance and consequently finding adapted strategies to prevent their modification during the filtration process.In general, the models can be classified into two main categories: mass transport models which focus on solute permeation during the filtration process, and fouling models based on particle or solute blocking within the membrane porous structure. In most of the cases, models depend strongly on the empirical or semi-empirical parameters and thus remain phenomenological. Two main objectives have been set for the present work: 1. Get a better understanding of the membrane fouling mechanisms during filtration of a liquid medium containing suspended microorganisms. It should be pointed out that several Industrial streams and wastewaters belong to this kind of effluents.2. Propose a macroscopic model describing the observed fouling mechanisms. [...]

Predicting the dissolution rates of immiscible contaminants in groundwater is crucial for developing environmental remediation strategies, but quantitative modeling efforts are inherently subject to multiple uncertainties. These include unknown residual amounts of non-aqueous phase liquids (NAPL) and source zone dimensions, inconsistent historical monitoring of contaminant mass discharge, and the mathematical simulation of field-scale mass transfer processes. Effective methods for simulating NAPL dissolution must therefore be able to assimilate a variety of data through physical and scalable mass transfer parameters to quantify and reduce site-specific uncertainties. This investigation coupled upscaled and numerical mass transfer modeling with uncertainty analyses to understand and develop data-assimilation and parameter-scaling methods for characterizing NAPL source zones and predicting depletion timeframes. Parameters of key interest regulating kinetic NAPL persistence and contaminant fluxes are residual mass and saturation, but neither can be measured directly at field sites. However, monitoring and characterization measurements can constrain source zone dimensions, where NAPL mass is distributed. This work evaluated the worth of source zone delineation and dissolution monitoring for estimating NAPL mass and mass transfer coefficients at multiple scales of spatial resolution. Mass transfer processes in controlled laboratory and field experiments were analyzed by simulating monitored dissolved-phase concentrations through the parameterization of explicit and lumped system properties in volume-averaged (VA) and numerical models of NAPL dissolution, respectively. Both methods were coupled with uncertainty analysis tools to investigate the relationship between data availability and model design for accurately constraining system parameters and predictions. The modeling approaches were also combined for reproducing experimental bulk effluent rates in discretized domains, explicitly parameterizing mass transfer coefficients at multiple grid scales. Research findings linked dissolved-phase monitoring signatures to model estimates of NAPL persistence, supported by source zone delineation data. The accurate characterization of source zone properties and kinetic dissolution rates, governing NAPL longevity, was achieved by adjusting model parameterization complexity to data availability. While multistage effluent rates accurately constrained explicit-process parameters in VA models, spatially-varying lumped-process parameters estimated from late dissolution stages also constrained unbiased predictions of NAPL depletion. Advantages of the numerical method included the simultaneous assimilation of bulk and high-resolution monitoring data for characterizing the distribution of residual NAPL mass and dissolution rates, whereas the VA method predicted source dissipation timeframes from delineation data alone. Additionally, comparative modeling analyses resulted in a methodology for scaling VA mass transfer coefficients to simulate NAPL dissolution and longevity at multiple grid resolutions. This research suggests feasibility in empirical constraining of lumped-process parameters by applying VA concepts to numerical mass transfer and transport models, enabling the assimilation of monitoring and source delineation data to reduce site-specific uncertainties. / Doctor of Philosophy / Predicting the dissolution rates of immiscible contaminants in groundwater is crucial for developing environmental restoration strategies, but quantitative modeling efforts are inherently subject to multiple uncertainties. These include unknown mass and dimensions of contaminant source zones, inconsistent groundwater monitoring, and the mathematical simulation of physical processes controlling dissolution rates at field scales. Effective simulation methods must therefore be able to leverage a variety of data through rate-limiting parameters suitable for quantifying and reducing uncertainties at contaminated sites. This investigation integrated mathematical modeling with uncertainty analyses to understand and develop data-driven approaches for characterizing contaminant source zones and predicting dissolution rates at multiple measurement scales. Parameters of key interest regulating the lifespan of source zones are the distribution and amount of residual contaminant mass, which cannot be measured directly at field sites. However, monitoring and site characterization measurements can constrain source zone dimensions, where contaminant mass is distributed. This work evaluated the worth of source zone delineation and groundwater monitoring for estimating contaminant mass and dissolution rates at multiple measurement scales. Rate-limiting processes in controlled laboratory and field experiments were analyzed by simulating monitored groundwater concentrations through the explicit and lumped representation of system properties in volume-averaged (VA) and numerical models of contaminant dissolution, respectively. Both methods were coupled with uncertainty analysis tools to investigate the relationship between data availability and model design for accurately constraining system parameters and predictions. The approaches were also combined for predicting average contaminant concentrations at multiple scales of spatial resolution. Research findings linked groundwater monitoring profiles to model estimates of contaminant persistence, supported by source zone delineation data. The accurate characterization of source zone properties and contaminant dissolution rates was achieved by adjusting model complexity to data availability. While monitoring profiles indicating multi-rate contaminant dissolution accurately constrained explicit-process parameters in VA models, spatially-varying lumped parameters estimated from late dissolution stages also constrained unbiased predictions of source mass depletion. Advantages of the numerical method included the simultaneous utilization of average and spatially-detailed monitoring data for characterizing the distribution of contaminant mass and dissolution rates, whereas the VA method predicted source longevity timeframes from delineation data alone. Additionally, comparative modeling analyses resulted in a methodology for scaling estimable VA parameters to predict contaminant dissolution rates at multiple scales of spatial resolution. This research suggests feasibility in empirical constraining of lumped parameters by applying VA concepts to numerical models, enabling a comprehensive data-driven methodology to quantify environmental risk and support groundwater cleanup designs.

Additive manufacturing (AM) has largely improved design freedom compared with traditional manufacturing processes such as casting and milling. The layer-by-layer workflow makes it possible to produce objects with much more complex shapes and structures. This feature is of particular interest for turbine blade manufacturing since internal cooling channels with higher thermal efficiency can be achieved toimprove the overall efficiency of a gas turbine. One feature of AM, especially for Laser Power Bed Fusion (LPBF) working on metal powders, is the relatively large surface roughness (SR), which will affect both heat transfer and pressure loss. Its geometry is also unique with the very randomly distributed spherical-shaped structures. This randomness makes the correlations for heat transfer and pressure loss based on sand grain roughness not applicable anymore. More in-depth research is needed to investigate the roughness effects. In this study, the AM roughness is modelled by a statistical distribution of spheres with different diameters using an upscale ratio of 77.4. An infrared (IR) camera was used to record the temperature distribution on the rough plates subjected to heated airflow. Three Re in the turbulent region (15000, 20000, 25000) were tested and the data from the IR camera were used to calculate the heat transfer coefficient (HTC) on the rough plate through a 3D finite element calibration solver. The results of averaged HTC agree well with data of the real Inconel 939 AM channel from which the upscaled rough plates are modelled. Also, the general patterns of HTC distributions matched the fluid dynamics analysis. Moreover, the results of arranging smooth and rough plates together shows that the heat transfer enhancement from SR is due to both the induced turbulent flows and the increased surface area. / Additiv tillverkning har i hög grad förbättrat designfriheten jämfört med traditionella tillverkningsprocesser som gjutning och fräsning. Att bygga lager för lager gör det möjligt att tillverka föremål med mycket mer komplexa former och strukturer. Denna egenskap är av särskilt intresse för tillverkning av turbinblad eftersom bättre interna kylkanaler kan uppnås för att förbättra den totala verkningsgraden hos en gasturbin. En egenskap hos additiv tillverkning är den relativt stora ytojämnheten, som påverkar både värmeöverföring och tryckförlust. Dess geometri är också unik med mycket slumpmässigt fördelade sfäriskt formade strukturerna. Denna slumpmässighet gör att de korrelationer för värmeöverföring och tryckförlust som baseras på sandkornens grovlek inte längre är tillämpliga. Mer djupgående forskning behövs för att undersöka grovhetseffekterna. I den här studien modelleras den additiva tillverkningens grovhet med en statistisk fördelning av sfärer med olika diametrar med ett uppskalningsförhållande på 77,4. En infraröd (IR) kamera användes för att registrera temperaturfördelningen på de skrovliga plattorna som utsattes för ett uppvärmt luftflöde. Tre Re i det turbulenta området (15000, 20000, 25000) testades och data från IR-kameran användes för att beräkna värmeöverföringskoefficientenpå den grova plattan genom en 3D finita elementkalibreringslösare. Resultaten av den genomsnittliga värmeöverföringskoefficienten stämmer väl överens med data från den verkliga additivt tillverkade Inconel 939-kanalen från vilken de uppskaladeg rova plattorna är modellerade. Även de allmänna mönstren för fördelningen av värmeöverföringskoefficienter stämmer överens med den fluiddynamiska analysen. Dessutom visar resultaten av att arrangera släta och grova plattor tillsammans att värmeöverföringsförbättringen från ytjämnhet beror på både de inducerade turbulenta flödena och den ökade ytarean.