In the presented paper, the applicability of pressure-decay methods to determine the diffusivities of gases in hydraulic fluids is analysed. First, the method is described in detail and compared to other measurement methods. Secondly, the thermodynamics and the mass transfer process of the system are studied. This results in four different thermodynamic models of the gaseous phase in combination with two diffusion models. Thirdly, the influence of the models on the pressure-decay method is evaluated computationally by examining the diffusion process of air in water as all system parameters are available from literature. It is shown that ordinary pressure-decay methods are not applicable to gas mixtures like air and therefore a new method for calculating the diffusivities is suggested.

Black soldier fly larvae (BSFL) waste treatment is becoming a more sought-after organic waste treatment since it produces larvae for animal feed as well as fertilizer, the need for prediction and optimization of treatment performance increases. This study aims to predict biomass conversion efficiency (BCE) for waste going into treatment, using only the nutritional data from other studies as it could give a good performance estimation while using available data. First, macronutrient fractions, based on a normalized total weight of volatile solids (VS), were calculated from nutritional data of other studies and of this present study.  Using the results, coordinates of these samples were calculated and put on the ternary plot, revealing trends for different macronutrient compositions. To validate these trends, selected nutrient compositions were created as waste mixtures containing fish, bread and vegetable waste and then treated by BSFL. After treatment, BCE and waste reduction in both VS and wet weight basis were noted.  The results showed that a composition high in carbohydrate and low in fat content is favorable for a high BCE value while protein content remained similar across all treatments and only seemed to slightly increase BCE. Treatment A5 held the highest BCE%VS (32.6 ± 3.5) with a composition of 73%VS carbohydrates, 19%VS protein and 8%VS fat and control treatment (A0) with no fish waste held the highest BCE%WW  (21.2 ± 1.8).   When comparing with the results of the other studies used to make the trends in the ternary plot, predictions of BCE were more accurate for a composition  of  70%VS carbohydrates, 20-25%VS protein and < 8%VS fat and also for compositions with similar waste substrates, as their percental changes were similar. This suggests that it is possible to predict high BCE  values from nutritional data in a ternary plot and could be of use when optimizing larval production. / Fluglarvskompostering med den amerikanska vapensflugan (Hermetia illucens) är en avfallsbehandling för organiskt avfall som producerar larver och gödselmedel som kan används som djurfoder respektive gödselmedel.  Många studier har undersökt hur makronäringsämnen påverkar larvproduktion eller behandlingseffektiviteten och kommit till olika resultat. I dagens läge skulle det underlätta om man kunde uppskatta den högsta möjliga effektiviteten baserat på data om avfallet som redan finns utan att behöva gå igenom en längre studie först. Den här studien har som syfte att undersöka om det är möjligt att förutspå biomassomvandlings-effektivitet (BCE) från makronäringsdata tagen från andra studier i ett triangulärt diagram. I denna studie användes studierna av Lopes et al., (2020) och Isibika et al., (2021) som basdata i triangulär diagrammet.  För att validera det trender som sågs i det triangulära diagrammet utfördes tester på valda kompositioner beståendes av ansjovis, bröd och grönsaksavfall med fluglarvkompostering. För att kunna applicera makronäringsämnernas data till triangulär diagrammet, beräknades mängden protein, kolhydrater och fett som en andel av den totala mängden glödförlust (VS).  Efter behandling beräknades BCE på våtvikt (WW) och VS basis.  Resultaten tyder på att en avfallskomposition med hög halt kolhydrater och låg halt fett (70%VS kolhydrater, 20–25%VS protein and <8%VS fett) bidrog till ett högt BCE värde, där proteinhalten var snarlik i alla behandlingar och verkade ha ett litet bidrag till ett högt BCE värde. Behandlingen A5 hade det högsta BCE%VS värdet (32.6 ± 3.5) med en komposition av 73%VS kolhydrater, 19%VS protein och 8%VS fett och kontroll behandlingen A0 (utan ansjovis avfall) hade högst BCE%WW värde (21.2 ± 1.8).   Vid jämförelse mellan de andra studierna och resultaten, var uppskattningar av BCE mer precisa för en komposition med ca 70%VS kolhydrater, 20–25%VS protein and <8%VS fett och för uppskattningar baserad på data från rapporter med liknande avfallssubstrat, då deras procentuella förändringar var liknande varandra. Detta stärker hypotesen att det är möjligt att kunna förutspå högt BCE värde med hjälp av näringsdata från avfallet och skulle kunna användas för att optimera av larvproduktion.

Afin d’améliorer le recyclage des thermoplastiques, de nouveaux mélanges sont étudiés pour élargir les débouchés offerts à ces matériaux. Ils visent à combiner efficacement les contributions de deux dispersions de polyéthylène (PE) et de polyamide (PA) pour renforcer les propriétés d’une matrice polypropylène (PP). Des stratégies de compatibilisation spécifiques à chaque phase minoritaire sont mises en place. Elles permettent d’élaborer en une seule étape d’extrusion réactive un mélange de deux dispersions nodulaires submicroniques et indépendantes. Grâce aux interphases de copolymères polyoléfines entourant le PE, ces particules facilitent l’écoulement plastique des macromolécules de PP avoisinantes à la place de la fissuration de la matrice. Entre le PP et le PA, deux compatibilisations différentes sont appliquées. Celle utilisant le polypropylène greffé d’anhydride maléique (PPgMA) forme une interface rigide et cohésive. Le module et la tenue en température supérieure du polyamide se répercutent alors sur les propriétés du mélange sans pour autant provoquer de fragilisation du matériau à température ambiante grâce à l’effet toujours efficace des particules de PE permettant de dissiper les contraintes concentrées autour du PA. Une compatibilisation alternative à base de poly(styrène-bloc-(éthylène-co-butylène)-bloc-styrène) greffé d’anhydride maléique (SEBSgMA) forme une interphase souple autour des particules de PA. Au lieu d’augmenter le module des mélanges, cette dispersion se comporte alors elle aussi comme un modifiant choc souple et augmente la déformabilité du mélange. Il est possible de tirer profit de la morphologie hétérogène du matériau pour localiser préférentiellement des particules de talc dans la phase PA des mélanges ternaires et ainsi minimiser le contact entre la charge et la matrice. Ainsi, les cavités formées par décohésion de l’interface charge/polymère restent confinées au sein d’une phase dispersée ce qui minimise leur impact sur la déformabilité du mélange. Le module d’Young de la phase PA est par contre augmenté ce qui se fait sentir sur le composite dans son ensemble. Une telle démarche permet de combiner rigidité et ténacité dans un mélange hétérogène de polymères. / In order to increase thermoplastic polymer recycling, new blends are studied to open-up new prospects for such materials. The objective is to efficiently combined the effects of two minor phases of polyethylene (PE) and polyamide (PA) to tune the properties of a polypropylene matrix (PP). Specific compatibilisation strategies adapted to each minor phase are established. They allow the self-arrangement in only one step of reactive-extrusion of a controlled morphology with two nodular and independent particle populations finely dispersed in the PP based blend. Thanks to the polyolefin copolymer interphases around PE, these particles promote PP shear yielding instead of matrix crazing. Two different compatibilisation routes are applied at PP/PA interface. The one using maleic anhydride grafted polypropylene (PPgMA) forms strong bonds between the two homopolymers and a stiff interface. In this case, the higher PA Young’s modulus strengthens the entire blend without inducing brittleness at room temperature thanks to the presence of the PE particles acting as stress dissipater around PA nodules. The alternative compatibilisation based on maleic anhydride grafted styrene-ethylene-buthylene-styrene bloc copolymer (SEBSgMA) creates a soft interphase around PA particles. Instead of heighten the blend modulus, this compatibilizer makes the PA particles appear very flexible and act as an impact modifier, and therefore increases the deformation capability of the blend. It is possible to take advantage of the heterogeneous morphology of the ternary polymer blend to design particular composites. Inorganic particles such as talc can be preferentially localized in the PA particles of the blend, minimizing contacts between the filler and the PP matrix. Thus, the cavities due to polymer/filler interface debonding stay confined within a dispersed phase which prevent their growth and the apparition of a catastrophic fracture. On the other hand, talc increases the Young’s modulus of PA particles and consequently the modulus of the entire composite. The design of such a heterogeneous material allows the increase of both strength and toughness simultaneously.

This paper discusses the potential of gasoline, ethanol and methanol ternary blend as an alternative passenger car fuel in Sweden. Sweden has set various targets aimed to reduce its GHG emissions and to increase the share of renewables in the transportation sector. Nevertheless, the majority of the energy consumed in the road transportation sector still comes from fossil fuels. In order to replace the energy supply of fossil fuels by more renewable fuels, the potential of alternative renewable fuels needs to be explored. Therefore, the potential of a domestically produced ternary blend of Gasoline‐Ethanol‐ Methanol (GEM) fuel blend is analysed in this report. In order to test whether it has the potential to become a successful alternative fuel, an analysis is performed on the: methanol and ethanol production potential from domestic second‐generation feedstocks, the selection of the most suitable production pathways of the biofuels, the potential for a Swedish GEM fuel distribution infrastructure, the economic competitiveness of GEM fuel, and lastly on the environmental impact of the shift from cars running on neat gasoline to GEM fuel. In order to perform the analysis, two scenarios are developed for projecting the share of the GEM cars(cars running on GEM fuel blends) in the Swedish passenger car fleet, considering a time horizon from 2017 to 2030. In Scenario 1, a high share of passenger cars running on GEM fuel is obtained with 22 percent by 2030. In Scenario 2, a low share of cars running on GEM fuel is obtained with 17 percent by 2030. In both scenarios, the passenger cars running on GEM fuel take over the share of cars running on gasoline. The scenarios serve to project the energy demand for GEM fuels. By 2030, the projected energy demand for GEM fuels is 9.7 and 7.5 TWh for Scenario 1 and Scenario 2, respectively. From the biofuel potential studies, it can be concluded that the production potential of the alcohol fuels, derived from currently untapped domestic secondary resources, exceeds the projected energy demand of 9.7 and 7.5 TWh in 2030. According to this thesis, the production potential of 2nd generation ethanol and methanol are 36 and 61.1 TWh, respectively, by 2030. Moreover, the study shows that the majority of the existing fuel distribution network of E85 and gasoline, which is forecasted to have a significant overcapacity in the same time‐span as the scenarios, can be utilized in a GEM fuel distribution network. As a consequence, no major investments are required to develop a Swedish GEM fuel distribution network. Regarding the selection of the biofuel production pathways, this study indicates the most suitable way of producing methanol is by black‐liquor gasification. Regarding second‐generation ethanol, this thesis indicates that the fermentation forestry residues is the most beneficial production pathway. The biofuel production pathways are selected based on the energy yield ratios, the biofuel production cost and biomass feedstock cost. Moreover, this study demonstrates that under the current Swedish policies, GEM fuels blends are economic competitive with gasoline and E85. In order to test the economic competitiveness, a pay‐off curve was developed based on the pump price of gasoline and fuel economy of GEM fuel blends. This study shows the pump prices of GEM fuel blends pay‐off in comparison to gasoline. This analysis indicates that the pump prices of GEM fuel blends lays between 0.87 and 0.92 euro per liter. Regarding the environmental impact, this study indicates that the amount of GHG emissions avoided varies between 10.1 and 13.3 million metric tons CO2eq in Scenario 1. In Scenario 2, the amount of GHG emissions that can be avoided varies between 8.6 and 11.3 million metric tons CO2eq. Moreover, this study indicates that high methanol containing GEM fuel blend are more favourable in terms of biomass utilization, and high ethanol containing GEM fuel blends are more favourable in terms of economy and GHG savings.

In thisproject, solubility of butyl paraben in 7 puresolvents and 5 ethanol aqueous solvents has been determined at from 1 ℃to 50 ℃. Thermodynamic properties of butyl paraben have been measured by DifferentialScanning Calorimetey. Relationship between molar solubility of butyl paraben in6 pure solvents and thermodynamic properties has been analyzed. Thisrelationship suggests a method of estimating activity of solute at equilibrium fromcombining solubility data with DSC measurements. Then, activity coefficient accordingto the solubility at different temperatures can be estimated. Duringthe solubility measurements in ethanol aqueous solvents, it is found that whenbutyl paraben is added into aqueous solutions with certain proportion ethanol,solutions separates into two immiscible liquid layers in equilibrium. Water andethanol are primary in top layer, while the butyl paraben is primary in bottomlayer, but the solution turns to cloudy when two layers of solution are mixed. Theaim of this work was to present the phase behaviour of liquid-liquid-phaseseparation for (butyl paraben + water + ethanol) ternary system from 1 ℃ to 50 ℃at atmospheric pressure. Thearea of liquid-liquid-phase separation region in the ternary phase diagram increaseswith the increasing temperature from 10 ℃to 50 ℃. In thisstudy, more than several hundreds of nucleation experiments of butyl paraben havebeen investigated in ethyl acetate, propanol, acetone and 90% ethanol aqueoussolution. Induction time of butyl paraben has been determined at 3 differentsupersaturation levels in these solvents, respectively. Free energy ofnucleation, solid-liquid interfacial energy, and nuclei critical radius havebeen determined according to the classical nucleation theory. Statistical analysis ofinduction time reveals that the nucleation is a stochastic process with widevariation even at the same experiment condition. Butyl paraben nucleates most difficultlyin 90 % ethanol than in other 3 solvents, and most easily in acetone. The interfacialenergy of butyl paraben in these solvents tends to increasing with decreasemole fraction solubility in these solvents. Coolingcrystallizations with different proportions of butyl paraben, water and ethanolhave been observed by Focused Beam Reflectance Method, Parallel VirtualMachine, and On-line Infrared. The FBRM, IR curves and the PVM photos show someof the solutions appeared liquid-liquid phase separation during coolingcrystallization process. The results suggest that if solutions went throughliquid-liquid phase separation region during the cooling crystallizationprocess the distribution of crystals crystal was poor. Droplets from solutions withsame proportion butyl paraben but different proportions of water and ethanolhave been observed under microscope. Induction time of the droplets has been determinedunder the room temperature. Droplets from top layer or bottom layer of solutionwith liquid-liquid phase separation on small glass or plastic plates were alsoobserved under microscope. The microscope photos show that the opposite flows ofcloudy solution on the glass and the plastic plate before nucleation. The resultsof the cooling and evaporation crystallization experiments both revealed thatnucleation would be prevented by the liquid-liquid phase separation. / QC 20110630

D. Tech. (Department of Chemistry, Faculty of Applied and Computer Sciences), Vaal University of Technology / A novel low-cost ternary Mn-Fe-Cu (MFC) metal oxide nanocomposite adsorbent was fabricated using facile co-precipitation method and successfully applied for the sequestration of Cr(VI) and As(III) from simulated aqueous efflent. The central composite design (CCD) of the response surface methodology (RSM) optimization technique determined the optimal working parameters for the preparation of the ternary MFC metal oxide nanocomposite. The spectroscopic microstructural analysis of the ternary MFC metal oxide nanocomposite was performed using fourier transform infrared (FTIR) spectroscopy, transmission electron microscopy (TEM), scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS) and X-ray diffraction (XRD) spectroscopy. The spectroscopic analyses revealed a rough surface with hydroxyl groups and the presence of mixed metal oxides in different valence states. The BET surface area, pore volume and pore size of the nanostructured MFC ternary metal oxide composite were found to be 77.2427 m2/g, 0.2409 cm3/g and 14.7560 nm, respectively. The pH drift method determined that the pHpzc of the adsorbent was 6.75. The batch technique was employed to investigate the adsorption dynamics (effects of ionic strength, co-existing anions, adsorbent regeneration and reuse) and optimum parameters (solution pH, adsorbent dosage concentration, desorption) of Cr(VI) and As(III) adsorption onto the MFC nanocomposite. The fitting of non-linear kinetic (pseudo-first-order, pseudo-second-order and Elovich), diffusion (intraparticle and Boyd) and isotherm (Langmuir, Freundlich and Dubinin-Radushkevich) models to the Cr(VI) and As(III) experimental adsorption data gave an insight into the adsorption mechanisms. The Langmuir adsorption capacities, qm (mg/g), were 168.71 at solution pH 3 and 35.07 at solution pH 9 for Cr(VI) and As(III) adsorption, respectively. The adsorption of Cr(VI) onto the ternary MFC metal oxide nanocomposite was physical and formed outer-sphere surface complexes through electrostatic interactions, while the removal of As(III) was specific due to inner-sphere surface complexation and ligand/ion exchange reactions. The results from XPS and FTIR analysis after the adsorption of Cr(VI) and As(III) showed that the surface hydroxyl groups on the MFC nanocomposite interacted with the Cr(VI) and As(III) species during the formation of the surface complexes. To facilitate ease of adsorbent removal from the treated simulated aqueous effluent, the ternary MFC metal oxide system was co-precipitated onto biochar support.

In den letzten Jahren hat sich maschinelles Lernen als hilfreiches Werkzeug zur Vorhersage von simulierten Materialeigenschaften erwiesen. Somit können aufwendige Berechnungen mittels Dichtefunktionaltheorie umgangen werden und bereits bekannte Materialien besser verstanden oder sogar neuartige entdeckt werden. Eine zentrale Rolle spielt dabei der Deskriptor, ein möglichst interpretierbarer Satz von Materialkenngrößen. Diese Arbeit präsentiert einen Ansatz zur Auffindung von Deskriptoren für periodische Multikomponentensysteme, deren Eigenschaften durch die genaue atomare Anordnung mitbeinflusst wird. Primäre Features von Einzel-, Paar- und Tetraederclustern werden über die Superzelle gemittelt und weiter algebraisch kombiniert. Aus den so erzeugten Kandidaten wird mittels Dimensionalitätsreduktion ein geeigneter Deskriptor identifiziert. Zudem stellt diese Arbeit Strategien vor bei der Modellfindung Kreuzvalidierung einzusetzen, sodass stabilere und idealerweise besser generalisierbare Deskriptoren gefunden werden. Es werden außerdem mehrere Fehlermaße untersucht, die die Qualität der Deskriptoren bezüglich Genauigkeit, Komplexität der Formeln und Berücksichtung der atomaren Anordnung charakterisieren. Die allgemeine Methodik wurde in einer teilweise parallelisierten Python-Software implementiert. Als konkrete Problemstellungen werden Modelle für die Gitterkonstante und die Mischenergie von ternären Gruppe-IV Zinkblende-Legierungen "gelernt", mit einer Genauigkeit von 0.02 Å bzw. 0.02 eV. Datenbeschaffung, -analyse, und -bereinigung werden im Hinblick auf die Zielgrößen als auch auf die primären Features erläutert, sodass umfassende Analysen und die Parametrisierung der Methodik an diesem Testdatensatz durchgeführt werden können. Als weitere Anwendung werden Gitterkonstante und Bandlücken von binären Oktett-Verbindungen vorhergesagt. Die präsentierten Deskriptoren werden mit den Fehlermaßen evaluiert und ihre physikalische Relevanz wird abschließend disktutiert. / In the last years, machine learning methods have proven as a useful tool for the prediction of simulated material properties. They may replace effortful calculations based on density functional theory, provide a better understanding of known materials or even help to discover new materials. Here, an essential role is played by the descriptor, a desirably interpretable set of material parameters. This PhD thesis presents an approach to find descriptors for periodic multi-component systems where also the exact atomic configuration influences the physical characteristics. We process primary features of one-atom, two-atom and tetrahedron clusters by an averaging scheme and combine them further by simple algebraic operations. Compressed sensing is used to identify an appropriate descriptor out from all candidate features. Furthermore, we develop elaborate cross-validation based model selection strategies that may lead to more robust and ideally better generalizing descriptors. Additionally, we study several error measures which estimate the quality of the descriptors with respect to accuracy, complexity of their formulas and the capturing of configuration effects. These generally formulated methods were implemented in a partially parallelized Python program. Actual learning tasks were studied on the problem of finding models for the lattice constant and the energy of mixing of group-IV ternary compounds in zincblende structure where an accuracy of 0.02 Å and 0.02 eV is reached, respectively. We explain the practical preparation steps of data acquisition, analysis and cleaning for the target properties and the primary features, and continue with extensive analyses and the parametrization of the developed methodology on this test case. As an additional application we predict lattice constants and band gaps of octet binary compounds. The presented descriptors are assessed quantitatively by the error measures and, finally, their physical meaning is discussed.

La prédiction de valeurs de tension interfaciale des fluides est capitale dans de nombreuses applications industrielles. Les techniques de simulation moléculaire et l’évolution rapide des moyens de calcul intensif permettent depuis quelques années de prédire des valeurs de tension interfaciale pour des systèmes complexes. Des travaux concernant des équilibres liquide / vapeur des corps purs SO2, O2, N2 et Ar montrent que les modèles choisi pour chaque molécule peuvent influencer la qualité des prédictions de tension interfaciale. Des simulations d’équilibres gaz acide / alcane de type CO2 / n-butane, CO2 / n-décane et H2S / n-pentane ont ensuite été réalisées. Elles ont mis en évidence l’efficacité des méthodes de simulation de type Monte Carlo pour la prédiction des tensions interfaciales pour de tels systèmes. L’étude de systèmes ternaires H2O / N2+CH4 et H2O / CO2+H2S a par ailleurs montré que le recourt à la dynamique moléculaire pouvait faciliter l’équilibration des systèmes simulés, rendant plus efficace la prédiction des tensions interfaciales. L’étude d’équilibres liquide / vapeur de saumures de chlorure de sodium a permis de mettre en évidence l’efficacité de certains potentiels non-polarisables pour la prédiction de l’évolution de la tension interfaciale avec la molarité de sel. Les modèles polarisables de type core-shell choisis ne permettent de prédire ni les masses volumiques, ni les tensions interfaciales. Enfin, l’étude d’équilibres eau / alcane en présence de sel ou de méthanol a montré que les méthodes de dynamique moléculaire permettaient de prédire quantitativement des valeurs de tension interfaciale pour ce type d’interface. L’effet de l’alcool abaissant la tension interfaciale a bien été observé, tout comme son placement préférentiel à l’interface. / The prediction of interfacial tension of fluids is critical for many industrial applications. Advances in molecular simulation, and the recent evolution of supercomputing calculations allow for some years to predict the values of interfacial tension for complex systems. A work involving liquid / vapor equilibrium of pure compounds SO2, O2, N2 and Ar show that the models used can impact the quality of the prediction. Simulations of acid gas / alkane equilibrium such CO2 / n-butane, CO2 / n-decane or H2S / n-pentane were then performed. They have demonstrated the performance of methods of Monte Carlo simulations for the reproduction of interfacial tensions for such system. The study of ternary systems H2O / N2+CH4 and H2O / CO2+H2S has also shown that using molecular dynamics could help the equilibration of the simulated systems. The study of liquid / vapor equilibrias of sodium chloride brines show that certain non-polarizable models perform very well to predict the changes in the interfacial tension with the molarity of salt. The core-shell polarizable models based on the Drude oscillator model chosen did not allow to predict brines densities and the interfacial tensions. Finally, the study of water / alkane equilibria in the presence of salt or methanol showed that the molecular dynamics methods allow to predict quantitatively interfacial tension values for this kinds of interfaces. The effect of alcohol lowering the interfacial tension has been observed : this small surfactant populate the interfacial region at weak concentration.

Le but de cette thèse est de développer des méthodes pratiques plus performantes pour la détection d'instances de classes d'objets de la vie quotidienne dans les images. Nous présentons une famille de détecteurs qui incorporent trois types d'indices visuelles performantes - histogrammes de gradients orientés (Histograms of Oriented Gradients, HOG), motifs locaux binaires (Local Binary Patterns, LBP) et motifs locaux ternaires (Local Ternary Patterns, LTP) - dans des méthodes de discrimination efficaces de type machine à vecteur de support latent (Latent SVM), sous deux régimes de réduction de dimension - moindres carrées partielles (Partial Least Squares, PLS) et sélection de variables par élagage de poids SVM (SVM Weight Truncation). Sur plusieurs jeux de données importantes, notamment ceux du PASCAL VOC2006 et VOC2007, INRIA Person et ETH Zurich, nous démontrons que nos méthodes améliorent l'état de l'art du domaine. Nos contributions principales sont : Nous étudions l'indice visuelle LTP pour la détection d'objets. Nous démontrons que sa performance est globalement mieux que celle des indices bien établies HOG et LBP parce qu'elle permet d'encoder à la fois la texture locale de l'objet et sa forme globale, tout en étant résistante aux variations d'éclairage. Grâce à ces atouts, LTP fonctionne aussi bien pour les classes qui sont caractérisées principalement par leurs structures que pour celles qui sont caractérisées par leurs textures. En plus, nous démontrons que les indices HOG, LBP et LTP sont bien complémentaires, de sorte qu'un jeux d'indices étendu qui intègre tous les trois améliore encore la performance. Les jeux d'indices visuelles performantes étant de dimension assez élevée, nous proposons deux méthodes de réduction de dimension afin d'améliorer leur vitesse et réduire leur utilisation de mémoire. La première, basée sur la projection moindres carrés partielles, diminue significativement le temps de formation des détecteurs linéaires, sans réduction de précision ni perte de vitesse d'exécution. La seconde, fondée sur la sélection de variables par l'élagage des poids du SVM, nous permet de réduire le nombre d'indices actives par un ordre de grandeur avec une réduction minime, voire même une petite augmentation, de la précision du détecteur. Malgré sa simplicité, cette méthode de sélection de variables surpasse toutes les autres approches que nous avons mis à l'essai.

In this work, the solubility of butyl paraben in 7 pure solvents and in 5 different ethanol-water mixtures has been determined from 1 ˚C to 50 ˚C. The solubility of ethyl paraben and propyl paraben in various solvents has been determined at 10 ˚C. The molar solubility of butyl paraben in pure solvents and its thermodynamic properties, measured by Differential Scanning Calorimetry, have been used to estimate the activity of the pure solid phase, and solution activity coefficients. More than 5000 nucleation experiments of ethyl paraben, propyl paraben and butyl paraben in ethyl acetate, acetone, methanol, ethanol, propanol and 70%, 90% ethanol aqueous solution have been performed. The induction time of each paraben has been determined at three different supersaturation levels in various solvents. The wide variation in induction time reveals the stochastic nature of nucleation. The solid-liquid interfacial energy, free energy of nucleation, nuclei critical radius and pre-exponential factor of parabens in these solvents have been determined according to the classical nucleation theory, and different methods of evaluation are compared. The interfacial energy of parabens in these solvents tends to increase with decreasing mole fraction solubility but the correlation is not very strong. The influence of solvent on nucleation of each paraben and nucleation behavior of parabens in each solvent is discussed. There is a trend in the data that the higher the boiling point of the solvent and the higher the melting point of the solute, the more difficult is the nucleation. This observation is paralleled by the fact that a metastable polymorph has a lower interfacial energy than the stable form, and that a solid compound with a higher melting point appears to have a higher solid-melt and solid-aqueous solution interfacial energy. It has been found that when a paraben is added to aqueous solutions with a certain proportion of ethanol, the solution separates into two immiscible liquid phases in equilibrium. The top layer is water-rich and the bottom layer is paraben-rich. The area in the ternary phase diagram of the liquid-liquid-phase separation region increases with increasing temperature. The area of the liquid-liquid-phase separation region decreases from butyl paraben, propyl paraben to ethyl paraben at the constant temperature. Cooling crystallization of solutions of different proportions of butyl paraben, water and ethanol have been carried out and recorded using the Focused Beam Reflectance Method, Particle Vision and Measurement, and in-situ Infrared Spectroscopy. The FBRM and IR curves and the PVM photos track the appearance of liquid-liquid phase separation and crystallization. The results suggest that the liquid-liquid phase separation has a negative influence on the crystal size distribution. The work illustrates how Process Analytical Technology (PAT) can be used to increase the understanding of complex crystallizations. By cooling crystallization of butyl paraben under conditions of liquid-liquid-phase separation, crystals consisting of a porous layer in between two solid layers have been produced. The outer layers are transparent and compact while the middle layer is full of pores. The thickness of the porous layer can reach more than half of the whole crystal. These sandwich crystals contain only one polymorph as determined by Confocal Raman Microscopy and single crystal X-Ray Diffraction. However, the middle layer material melts at lower temperature than outer layer material. / <p>QC 20130515</p> / investigate nucleation and crystallization of drug-like organic molecules