The purpose of this investigation was to improve upon the overall design of a 15-plate fractionating assembly used in the separation of simple binary mixtures of isopropyl alcohol and water, to recalibrate all measuring devices used in the operation of the column, to develop an analytical method by which the composition of liquid samples from the column could be determined accurately and quickly, and, in general, to place the assembly in such condition that experimental runs can be made more easily with a higher degree of accuracy than has previously been possible. / Master of Science

Molecular distillation presents a method of separation for substances of high molecular weights which would suffer thermal decomposition at the higher temperatures that would be required for conventional distillation. In general, the separatory power of molecular stills is poor and in order to obtain relatively pure substances from mixtures, a series of redistillations may have to be performed. The distillation of individual substances under molecular distillation conditions is best characterized by an elimination curve which is a plot of the yield of the substance against the temperature of distillation. Composite curves, similarly, describe the elimination of a mixture of substances. It was the purpose of this investigation to study the distillation characteristics of several anthraquinone dyes by the determination of single and composite elimination curves, and to attempt to separate binary mixtures of dyes by various redistillation techniques. The glass centrifugal molecular still with a five-inch rotor and a ball-jar condensing surface available at the beginning of the investigation was almost completely reconstructed as an all-metal unit employing copper and brass as the materials of construction. Final design included a vacuum-enclosed drive for the rotor. Other modifications included a copper, magnetically-operated feed pump; copper tubing flow lines with silver-soldered connections; copper feed, residue, and distillate reservoir tanks; spark plug leads for heater and motor lead attachments within the vacuum system; feed and residue thermocouples employing copper tubing as part of the copper lead of the copper-constantan hot junction; a silicone gasket for seal of the bell jar to the base plate; a copper feed tube and nozzle; and a vacuum pumping system consisting of two vertical diffusion pumps connected in series and exhausted to two megavac pumps and a hyvac pump to evacuate the distillate tank. Initial tests were made to obtain a constant-yield oil and to standardize an operating technique. A mixture of light and heavy mineral oils was found to be a satisfactory constant-yield oil. Several tests were made using a continuous temperature rise technique in which the distillate fractions were collected as the temperature rose slowly over the collection increment. 1,4-Diethyl-diamineanthraquinone, 1,4-diisopropyldiaminoanthraquinone, and 1,4-dibutyldiaminoanthraquinone dyes were distilled using this method. A discontinuous temperature rise technique replaced the continuous technique when it became evident that this initial technique could not be satisfactorily standardized. Instead of a continuous temperature increase as distillation progressed, the temperature was held constant while fractions were collected for a predetermined time interval required for five nominal passes of the feed across the rotor. A standardized technique was then developed end used for all subsequent tests, The conditions of operation under the technique were as follows: pressure, 6 ± 1 microns of mercury, absolute; rotor speed, 2000 ± 50 revolutions per minute; feed rate, 65 ± 2 milliliters per minute; nominal number of passes of the feed across the rotor for each fraction, 2; and difference between the feed and residue temperatures, 10 degrees Centigrade. The oil solvent, constant-yield oil, was a mixture of 300 milliliters of light mineral oil to 350 milliliters of heavy mineral oil for each test. Three anthraquinone dyes, celanthrene red 3B, 1,4-diethyldiaminoanthraquinone, and 2,6-dimethyldiaminoanthraquinone, were distilled individually using the standardized technique. The elimination curves were plotted and the maxima found to be 121, 152, end 159 degrees Centigrade, respectively. Equal weight binary mixtures of celanthrene red 3B and 1,4-diethyldiaminoanthraquinone, celanthrene red 3B and 2,6-dimethyldiaminoanthraquinone, and 1,4-diethyldiaminoanthraquinone and 2,6-dimethyldiaminoanthraquinone were distilled to study the effect of difference in the temperature interval between elimination maxima on the composite curve. The effect of varying the weight ratio was studied by the distillation of binary mixtures of 2:1, 1:1, and 1:2-1/2 weight ratio of celanthrene red 3B to 1,4-diethyldiaminoanthraquinone. A tertiary mixture of equal weights of the three dyes used in binary mixtures was distilled to note the added complexity of the composite curve. Attempts to separate binary mixtures were made by extractive redistillation in which additional mineral oil was mixed with the cumulative fractions from the original distillation before redistillation and repeated redistillation in which the cumulative fractions were merely redistilled without addition of any more oil. It was concluded from the investigation that each dye in a mixture distills independently of the other dyes preset. The introduction of each two GHZ groups in the 1,4 position of the anthraquinone series was found to increase the elimination maximum by approximately 10 degrees Centigrade. From a study of the composite curves for the various binary mixtures, it was decided that identification of two substances in an unknown mixture was possible if the second substance was in quantity at least 1/6 that of the first substance and had an elimination maximum at least 20 degrees Centigrade different from the first substance. A comparison of the weight per cent of 1,4-diethyldiaminoanthraquinone removed from the mixture with celanthrene red 3B by the two redistillative techniques indicated that extractive redistillation provided better separation than repeated redistillation with 75.5 weight per cent removed from the cumulative fractions up to 140 degrees Centigrade compared to 68.7 and 52.8 weight per cents removed by the first and second redistillations, respectively. / Master of Science

Molecular distillation is that process of free transfer of molecules under high vacuum from an evaporator to a condenser, where the distance of travel is within the limits of the mean free path of the vapor molecules in the residual gas. Under these conditions, distillations can be made at temperatures 100 to 150°C. lower than ordinary distillations. Many of the so-called "undistillables“ can be distilled under conditions existing in the molecular distillation process. The natural fats and waxes, sugar derivatives, petroleum residues, plastics, and plasticizers now fall within the scope of molecular distillation. The use of the molecular distillation process is now limited by the high costs involved. Thermal efficiency is low, only 2% of the total input heat is utilized for distillation. The separating powers of molecular stills are low, never more than that of one ”theoretical plate“. Only those materials that can bear the high process costs are now being produced commercially, the production of vitamins being the most prominent. Considerable work has been done relative to the development of molecular stills. Advancement to the present day centrifugal stills has been accomplished relatively recently. Many problems are yet unsolved. Much work must be done before molecular distillation can be placed on a level with other methods of distillation. An increase in thermal efficiency, an increase in separating powers, a decrease in film thickness, and a lowering of the costs of the process are necessary. The purpose of this investigation was to design and construct a five inch centrifugal molecular still. Four interchangeable rotors were employed, each of different thickness, different diameter, and having a different angle of inclination. The effect of angle of inclination on film thickness can then be studied. Although no data has been published relative to still performance of five inch commercial stills, data obtained from the operation of the still constructed in this investigation can be compared when such data is published. / Master of Science

This dissertation is a study of the composition dynamics of a bench size sieve-plate distillation column. The column was 2-3/4 inches in diameter, 30 inches in length, and contained 10 sieve-plates spaced 2-1/4 inches apart. A benzene-toluene mixture was used as the feed to the column. The pulse testing method for obtaining dynamic information was used. The column was upset by introducing a rectangular pulse increase in benzene composition in the feed stream. The response of each plate was measured in the form of temperature and was converted to a composition response. Frequency response curves were determined for each plate from the pulse response. For the frequency response analysis, the forcing function of a plate was considered to be a weighted sum of the compositions of the liquid and vapor streams entering the stage, while the output function was considered to be the liquid composition on the plate near the exit downcomer. The resulting frequency response was relatively flat in the frequency region of primary interest. The flatness of the frequency response was attributed to poor liquid mixing on tha plates. Based on the experimental observations, a plate in the column may be mathematically represented by a dead time and a steady-state gain. / Ph. D.

Federated Learning (FL) is a sub-domain of machine learning (ML) that enforces privacy by allowing the user's local data to reside on their device. Instead of having users send their personal data to a server where the model resides, FL flips the paradigm and brings the model to the user's device for training. Existing works share model parameters or use distillation principles to address the challenges of data heterogeneity. However, these methods ignore some of the other fundamental challenges in FL: device heterogeneity and communication efficiency. In practice, client devices in FL differ greatly in their computational power and communication resources. This is exacerbated by unbalanced data distribution, resulting in an overall increase in training times and the consumption of more bandwidth. In this work, we present a novel approach for resource-efficient FL called emph{REFT} with variable pruning and knowledge distillation techniques to address the computational and communication challenges faced by resource-constrained devices. Our variable pruning technique is designed to reduce computational overhead and increase resource utilization for clients by adapting the pruning process to their individual computational capabilities. Furthermore, to minimize bandwidth consumption and reduce the number of back-and-forth communications between the clients and the server, we leverage knowledge distillation to create an ensemble of client models and distill their collective knowledge to the server. Our experimental results on image classification tasks demonstrate the effectiveness of our approach in conducting FL in a resource-constrained environment. We achieve this by training Deep Neural Network (DNN) models while optimizing resource utilization at each client. Additionally, our method allows for minimal bandwidth consumption and a diverse range of client architectures while maintaining performance and data privacy. / Master of Science / In a world driven by data, preserving privacy while leveraging the power of machine learning (ML) is a critical challenge. Traditional approaches often require sharing personal data with central servers, raising concerns about data privacy. Federated Learning (FL), is a cutting-edge solution that turns this paradigm on its head. FL brings the machine learning model to your device, allowing it to learn from your data without ever leaving your device. While FL holds great promise, it faces its own set of challenges. Existing research has largely focused on making FL work with different types of data, but there are still other issues to be resolved. Our work introduces a novel approach called REFT that addresses two critical challenges in FL: making it work smoothly on devices with varying levels of computing power and reducing the amount of data that needs to be transferred during the learning process. Imagine your smartphone and your laptop. They all have different levels of computing power. REFT adapts the learning process to each device's capabilities using a proposed technique called Variable Pruning. Think of it as a personalized fitness trainer, tailoring the workout to your specific fitness level. Additionally, we've adopted a technique called knowledge distillation. It's like a student learning from a teacher, where the teacher shares only the most critical information. In our case, this reduces the amount of data that needs to be sent across the internet, saving bandwidth and making FL more efficient. Our experiments, which involved training machines to recognize images, demonstrate that REFT works well, even on devices with limited resources. It's a step forward in ensuring your data stays private while still making machine learning smarter and more accessible.

The effect of high organic liquid wastes on the overall performance of anaerobic bioreactor landfills is not known. In this study three different liquid wastes were added to mix of office paper, newspaper, cardboard and plastic in 3 different concentrations under anaerobic conditions to determine their effect on degradation of organics. The addition of buffer chemicals was found to improve degradation compared to data from an earlier study done without the addition of initial buffering. Paint waste with a COD 237,500 mg/L added at a concentration of 10% did not show any negative effect on the overall performance. The distillation waste with a COD of 812,500 mg/L added at 5% and higher concentration resulted in accumulation of volatile fatty acids and strong inhibition. The surfactant waste added at concentrations up to 1500 mg/L showed some inhibition but the overall performance was good. The surfactant waste also appeared to improve lignin degradation. / Master of Science

Master of Science

Implantées depuis plus de dix ans dans l’industrie, l’intérêt des colonnes à cloison n’est aujourd’hui plus à démontrer. Suivant les applications envisagées, elles peuvent permettre d’important gain tant énergétique qu’économique. Cependant leur design est toujours complexe et les méthodes de conception développées ne s’adressent qu’à des applications avec des mélanges proches de l’idéalité. Le développement d’une nouvelle méthode de design d’une colonne à distillation de mélanges non-idéaux applicable aux colonnes à cloison constitue l’objet d’étude de cette thèse. Dans un premier temps, une procédure de faisabilité et de design d’une colonne classique basée sur les feuilles opératoires a été proposée. Les paramètres de design obtenus sont ensuite utilisés comme initialisation d’une simulation rigoureuse conduite au sein du logiciel Prosim Plus. Afin de tester cette procédure, des mélanges idéaux, non-idéaux et azéotropiques ont été utilisés. Il a pu être montré que les paramètres de design obtenus permettent d’accéder à un design plus fiable et plus efficient, aussi bien d’un point de vue énergétique qu’économique, que ceux obtenus par la bien connue méthode shortcut FUGK. Dans un second temps, la procédure a été adaptée aux colonnes à cloison. Les paramètres obtenus ont également servi à initialiser une simulation rigoureuse et la procédure a été testée avec les mêmes mélanges. Il a été mis en évidence que les paramètres de design obtenus permettaient d’obtenir une bonne initialisation de la colonne. Comparée avec une autre méthode développée précédemment par le laboratoire, la procédure développée s’est révélée plus fiable et a permis l’obtention de design plus économique tant du point de vue énergétique qu’en termes d’investissement. Enfin, une phase de validation expérimentale a également été réalisée sur une installation pilote. Dans un premier temps, l’instrumentation et le contrôle de la colonne ont été fortement améliorés. Dans un second temps, des résultats expérimentaux réalisés sur avec un mélange hétéro-azéotropique ont pu valider des simulations en terme de profils de composition et de température interne ainsi que les compositions et les débits de sortie de la colonne. / The constant increase of divided wall columns usage in the industry for more than ten years shows that its benefits are no longer to be demonstrated. Depending on the applications, it can provide significant energy and economic savings. However their design is still complicated and existing design methods are only applicable to ideal mixtures. The development of a new design method for divided wall distillation column for non-ideal and azeotropic mixtures is the subject of this thesis. Firstly, a feasibility and design procedure for classic distillation column based on the operation leaves was proposed. Calculated design parameters are then used to initialize a rigorous simulation run in ProSimPlus© software. The procedure was tested with ideal, non-ideal and azeotropic mixtures. Results show that calculated design parameters provide a more reliable and efficient column design, from an energetic and economic point of view, than those calculated with the well-known FUGK shortcut method. Secondly the procedure was adapted for divided wall column. Compared with a method developed previously in the laboratory, this new methodology is more reliable and provides more energy and cost efficient designs. Finally, an experimental validation work was made on a pilot plant. Firstly, both instrumentation and control was strongly improved. Secondly, experimental results with a hetero-azeotropic mixture were able to validate rigorous simulation in terms of compositions and temperature profiles inside the column as well as outlet compositions and flowrates of the column.

Submitted by Johnny Rodrigues (johnnyrodrigues@ufcg.edu.br) on 2018-03-12T16:25:13Z No. of bitstreams: 1 ARIOSTON ARAÚJO DE MORAIS JÚNIOR - TESE PPGEQ 2015..pdf: 4075446 bytes, checksum: 58c7745e3f0cf289b5f5d0a0a774ed88 (MD5) / Made available in DSpace on 2018-03-12T16:25:13Z (GMT). No. of bitstreams: 1 ARIOSTON ARAÚJO DE MORAIS JÚNIOR - TESE PPGEQ 2015..pdf: 4075446 bytes, checksum: 58c7745e3f0cf289b5f5d0a0a774ed88 (MD5) Previous issue date: 2015-11-06 / Neste trabalho é proposta uma metodologia para construção de sensores virtuais implementados em software, com objetivo de estimar e prever o comportamento de impurezas na corrente de base de uma coluna de destilação de alta pureza, do processo produtivo do 1,2 Dicloroetano (C2H4Cl2). A aquisição dos dados utilizados na construção dos sensores virtuais foi realizada através do modelo matemático do processo, simulado com dados reais de uma planta industrial. O estudo específico engloba a modelagem matemática/termodinâmica e avaliação do comportamento estacionário e dinâmico dessa torre, simulada aqui no software Aspen Plus e DynamicsTM. Desse modo, o modelo fornece os dados necessários para inferência das impurezas relacionadas, que são os teores dos compostos tetracloreto de carbono (CCl4) e clorofórmio (CHCl3), ambos devem ser mantidos, respectivamente, em valores ≤3000 e ≤400 ppm (partes por milhão). A metodologia também aborda dois algoritmos de seleção de variáveis secundárias, que utilizam técnicas estatísticas multivariadas (algoritmo de todas as regressões possíveis-TRP e da análise de componentes principais-PCA). Verifica-se também nos dados gerados quanto a real ou não necessidade de remoção de erros grosseiros (outliers), por isso é também inserida na metodologia uma etapa de pré-processamento de dados. Foram selecionados os dez melhores modelos de inferência para cada uma das saídas. Diante dessa informação, os melhores modelos produzidos não utilizavam as concentrações dos compostos das correntes de alimentação e sim medições de temperaturas ao longo da torre. Uma importante conclusão do ponto de vista de construção de sensores virtuais, porque na maioria dos trabalhos desenvolvidos essas variáveis são cruciais na produção de bons resultados. O treinamento dos sensores virtuais foi efetuado em um ambiente ruidoso, haja vista que foram simulados ruídos inerentes às medições (ruídos brancos Gaussianos). Na etapa final, os sensores virtuais são construídos utilizando uma técnica de modelagem empírica, redes neurais artificiais (RNA), onde foram utilizadas RNA do tipo Perceptron Multicamadas (MLP). Foram também avaliadas diversas variações quanto ao número de neurônios e camadas ocultas das RNA, empregando como critério de parada a técnica de validação cruzada. Os sensores virtuais desenvolvidos apresentaram erros satisfatórios do ponto de vista de engenharia, uma boa análise de regressão e um bom erro médio quadrático. Logo, com essas estimativas espera-se a minimização e a previsão do comportamento transiente dos compostos no referido processo. / This work proposes a methodology for software implementation to make soft sensors. The goal is to estimate and predict the behavior of impurities in the bottom current of a highpurity distillation column, for 1,2-Dichloroethane or 1,2-DCE (C2H4Cl2) production. The data acquisition used in the construction of soft sensors was performed through a mathematical simulation of the process, with real industrial data taken from an industrial plant. A specific study involves the mathematical modeling, thermodynamics, evaluation of the steady state and the dynamic behavior of this process, simulated here in Aspen Plus and Aspen DynamicsTM software. Thus, the model provides the necessary data to infer the contents of the carbon impurities mentioned above, Tetrachloride (CCl4) and Chloroform (CHCl3), which are to be fixed approximately below 3000 and 400 ppm (parts per million) respectively. The methodology also covers selection algorithms of secondary variables, using multivariate statistical techniques: All Possible Regressions (TRP) and Principal Component Analysis (PCA). The data generated was checked in order to know whether to include or not a step for removal of outliers, so it was also included in the methodology one preprocessing data step. The ten best inference models were selected for each output concentration. With this information, these models do not use concentrations measurements in the feed streams but measurements of the temperature along the column. This is an important conclusion from the point of view of virtual sensors building, because in most of the literature reported these variables are crucial in getting good results. The training of soft sensors was done in a noisy environment, considering that simulated noise was inherent to measurements (Gaussian noise). In the final step, the soft sensors devices are constructed using an empirical modeling technique of artificial neural networks (ANN), which were generated ANN type Multilayer Perceptron (MLP). Several variations were also evaluated on the number of neurons and hidden layers of networks, employing as a stopping criterion the cross-validation technique. The developed soft sensors presented satisfactory errors from the engineering viewpoint, a good regression and a good mean square error. Finally, with these estimations it is expected to minimize and predict the transient behavior of the compounds in the referred process.

Submitted by Fabio Sobreira Campos da Costa (fabio.sobreira@ufpe.br) on 2016-09-16T14:17:18Z No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Dissertação_Andréa Leão_Ver_Final.pdf: 2849890 bytes, checksum: cd4d0770cfce99861e2cf216c67ed0fb (MD5) / Made available in DSpace on 2016-09-16T14:17:18Z (GMT). No. of bitstreams: 2 license_rdf: 1232 bytes, checksum: 66e71c371cc565284e70f40736c94386 (MD5) Dissertação_Andréa Leão_Ver_Final.pdf: 2849890 bytes, checksum: cd4d0770cfce99861e2cf216c67ed0fb (MD5) Previous issue date: 2016-02-23 / ANP / O presente estudo propõe a separação entre os triacilgliceróis (TAGs), principais constituintes dos óleos vegetais, e ácidos graxos livres (AGLs) através do processo de destilação a vácuo. A possibilidade dessa separação consiste na considerável diferença de volatilidade entre os AGLs e os TAGs. Foram preparadas misturas modelo aos óleos residuais, ou seja, combinações em diferentes proporções em massa de óleo de soja refinado (OS) e o reagente ácido oleico p.a (AO). As misturas foram destiladas nas pressões reduzidas de 0,13; 0,40; 0,67 e 1,33 kPa, sendo o processo realizado em batelada em um único estágio, sem refluxo. As amostras foram caracterizadas antes e após o processo de destilação a vácuo quanto: à massa específica, ao teor de água, ao teor de AGL (%) e à composição de ácidos graxos por cromatografia gasosa. Posteriormente, foram definidos os componentes modelo de TAG e AO e a estimativa dos parâmetros termofísicos das misturas para a simulação do processo de destilação a vácuo no software Aspen Plus V8.8, utilizando tanques flash não adiabáticos em série. O modelo termodinâmico usado foi o Non-Random Two-Liquid (NRTL). As curvas de destilação experimentais da mistura de 89,45% de OS e 10,55% de AO em massa foram melhor representadas pelas curvas de destilação simuladas, apresentando o percentual de desvios médios quadráticos de 2,5; 2,2; 2,4 e 2,5% para operações realizadas em pressões reduzidas de 0,13; 0,40; 0,67 e 1,33 kPa, respectivamente. Os dados de massa específica a 20°C dos produtos das destilações, resíduos e destilados, foram semelhantes aos do OS e AO, respectivamente. Os teores de água dos destilados (0,024% a 0,059%) e dos resíduos (0,001% a 0,014%) estão em níveis adequados para o emprego das reações de esterificação e transesterificação para a produção de biodiesel. Os teores de ácidos graxos livres dos resíduos obtidos após as destilações ficaram dentro da faixa de 0,4% a 1,2%, indicando processo de separação por destilação a vácuo efetivo. / This study proposes the separation of triacylglycerols (TAG), main constituent of vegetable oils, and free fatty acids (FFA) through the vacuum distillation process. The possibility of this separation is the considerable difference in volatility between the FFA and the TAG. Mixtures were prepared with model waste oil, or combinations of different mass ratios of refined soybean oil (SO) and the reagent oleic acid (OA). The mixtures were distilled at reduced pressures of 0.13; 0.40; 0.67 and 1.33 kPa, the process being carried out in batches in a single stage, without reflux. The samples were characterized before and after the vacuum distillation process as: the specific gravity, the water content, the FFA content (%) and the fatty acid composition by gas chromatography. Subsequently, the components were defined template TAG and OA and the estimation of parameters of mixtures thermophysical to simulate the vacuum distillation process in Aspen Plus v8.8 software using non-adiabatic flash tanks in series. The thermodynamic model used was the Non-Random Two-Liquid (NRTL). Experimental distillation curves of mixing 89.45% of SO and 10.55% of mass OA were best represented by simulated distillation curves, showing the percentage of mean deviation squared of 2.5; 2.2; 2.4 and 2.5% for operations in reduced pressures of 0.13; 0.40; 0.67 and 1.33 kPa, respectively. The specific mass of data at 20°C the distillation of the products, residues and distillates were similar to SO and OA, respectively. The content of distilled water (0.024% to 0.059%) and waste (0.001% to 0.014%) were at adequate levels for the use of esterification and transesterification reactions for biodiesel production. The contents of free fatty acids from residues obtained after distillations were within the range of 0.4% to 1.2%, indicating a separation process by distillation under vacuum effective.