http://www.bdtd.uerj.br/handle/1/3684 Fri, 17 Jul 2026 20:17:56 GMT 2026-07-17T20:17:56Z http://www.bdtd.uerj.br/handle/1/25624 Título: Emprego de métodos de aprendizado de máquina automático em tarefas de classificação e em análise de séries temporais aplicada em caso de pandemias Autor: Andrade, Bárbara Martins de Primeiro orientador: Luna, Aderval Severino Abstract: Data classification in chemometrics, combined with automated machine learning (AutoML), emerges as an innovative approach for estimating appropriate models and parameters in different contexts. This work proposes a methodology for classifying real chemical data, combining algorithms selected through meta-learning and specific preprocessing techniques to ensure robustness and reproducibility of results. Diverse datasets available in the literature were analyzed, including information on Brazilian artisanal cheeses, Chinese porcelain, drinking water quality, and organic waste from various animals. Exploratory analyses identified trends, distributions, and similarities in the data, in addition to highlighting the need for balancing techniques due to the presence of imbalanced classes. Data processing was performed using the Lazy Predict framework, which automates the application of multiple machine learning models, objectively recommending the best algorithms and parameters. Twenty-seven models were tested on artisanal cheese data, with algorithms such as Extreme Gradient Boosting, Random Forest, Bagging, and Extremely Randomized Trees achieving up to 98% accuracy, significantly outperforming conventional assumption-based models, which achieved 84%. In the remaining datasets, consistent gains in efficiency and reduced computational time were observed, demonstrating AutoML's potential to optimize results and accelerate analytical processes. In an exploratory approach, the methodology was also applied to time series of disease incidence using the AutoTS framework to evaluate the viability of AutoML on this type of dynamic data, especially in pandemic scenarios. This approach enabled the automation of modeling and algorithm optimization, offering more accurate predictions and reducing the time required for analysis, albeit initially and experimentally. This study demonstrates that the combination of AutoML with chemical and time series data analysis offers an efficient, accessible, and reproducible approach, allowing users without technical training to explore complex information and obtain relevant data to guide scientific and strategic decisions. The reproducibility of the results instills confidence in the methodology's reliability and its potential to consistently deliver accurate and insightful data analysis. Instituição: Universidade do Estado do Rio de Janeiro Tipo do documento: Tese Thu, 14 Aug 2025 00:00:00 GMT http://www.bdtd.uerj.br/handle/1/25624 2025-08-14T00:00:00Z http://www.bdtd.uerj.br/handle/1/25528 Título: Estudo da interação de diferentes bactérias na indústria do petróleo e seus metabólitos: monitoramento por microbiologia clássica e biologia molecular Autor: Montez, Gustavo Fabbri Primeiro orientador: Costa, Antonio Carlos Augusto da Abstract: The biogenic production of sulfide is one of the main problems of the oil and gas industry, causing corrosion in storage tanks and pipes. This is possible by injecting seawater during secondary oil recovery. In the present work, high levels of sulfate-reducing bacteria and acid-producing bacteria were detected in water / oil samples from various locations in the oil industry, as well as several other microbial groups. 35 samples were analyzed in a preliminary phase and in a second moment another 13 samples, where a wider range of microbial cells was detected. The water and oil samples that showed the highest microbial growth were submitted to metagenomic analysis, which confirmed the presence of a diversity of microorganisms, indicating the complexity of the consortium in the production of sulfide, based on the activity of acid-producing cells and associated species. It was also identified, through chromatographic analyzes, formic acid as a metabolite for the growth of acid-producing bacteria that will serve as a substrate for sulfate-reducing bacteria, thus creating a corrosion- friendly microbiome. Instituição: Universidade do Estado do Rio de Janeiro Tipo do documento: Tese Fri, 12 Feb 2021 00:00:00 GMT http://www.bdtd.uerj.br/handle/1/25528 2021-02-12T00:00:00Z http://www.bdtd.uerj.br/handle/1/25511 Título: Determinação de elementos-traço em matrizes complexas por ICP OES e ICP-MS Autor: Costa, Marina Araujo João Lopes da Primeiro orientador: Gois, Jefferson Santos de Abstract: The determination of trace elements in complex matrices presents challenges in chemical analysis due to interferences from major matrix components that compromise result accuracy and precision. This work presents analytical strategies to address such challenges using Inductively Coupled Plasma Optical Emission Spectrometry (ICP OES) and Inductively Coupled Plasma Mass Spectrometry (ICP-MS). The thesis is divided into three chapters, each focusing on a specific matrix and related analytical issues in trace element determination: seawater, edible insects, and glycerin. The first study focuses on the preconcentration of As, Cu, and Pb in seawater, a matrix with high levels of dissolved salts that interfere with analysis. Using manganese oxide molecular sieve (OMS-2) as an adsorbent, the preconcentration process was optimized, achieving high enrichment factors and suitable detection limits for accurate metal quantification. The second study investigates the bioaccessibility of essential trace elements (Fe, Zn, and Cu) in different edible insect species using simulated in vitro digestion followed by ICP-MS detection. Results showed interspecies variability, with mealworms presenting the highest values. Complementary studies using Size Exclusion Chromatography (SEC) coupled with ICP-MS provided insights into the molecular weight of Fe, Zn, and Cu associated species during digestion, suggesting these ions are primarily bound to low-molecular-weight organic compounds. The third study developed a robust "dilute-and-shoot" method for trace metal determination in glycerin, a biodiesel byproduct. Due to its high viscosity and organic nature, glycerin requires an approach that minimizes non-spectral interferences. The validated method proved effective, ensuring the precision and accuracy needed for compliance with regulatory standards. Overall, the research offers effective solutions to analytical challenges in trace element determination in complex matrices, contributing to progress in industrial and scientific fields. Instituição: Universidade do Estado do Rio de Janeiro Tipo do documento: Tese Fri, 22 Aug 2025 00:00:00 GMT http://www.bdtd.uerj.br/handle/1/25511 2025-08-22T00:00:00Z http://www.bdtd.uerj.br/handle/1/25204 Título: Estudo da reforma do biogás, utilizando-se catalisadores a base de Ni suportados em zeólitas: Efeito da hierarquização da estrutura e da síntese a partir de fontes alternativas de Si Autor: Guimarães, Amanda de Carvalho Pereira Primeiro orientador: Henriques, Cristiane Assumpção Abstract: The increasing search for sustainable energy sources and the urgency to reduce the impacts of climate change have stimulated the use of biogas. In this context, dry reforming of methane (DRM) emerges as a promising approach to convert biogas, composed of methane (CH4) and carbon dioxide (CO2), into syngas, a versatile and high value-added product. However, DRM is an endothermic process that requires high temperatures, leading to the deactivation of nickel (Ni) catalysts due to sintering and carbon formation. In this work, catalysts of nickel (Ni) supported on hierarchical ZSM-5 and USY zeolites were prepared, using desilication and dealumination/desilication methods, respectively. The study investigated how the hierarchical porous structure of these supports affected the dispersion of Ni particles, the metal-support interaction and, consequently, the catalytic activity and stability during DRM. The use of sugarcane bagasse ash (SCBA), an agricultural residue, as a source of silicon for zeolite synthesis is also considered. For the characterization of the catalysts, a series of analyses were used, such as Nitrogen Adsorption, X-ray Fluorescence Spectroscopy (XRF), X-ray Diffraction (XRD), Temperature-Programmed Reduction (TPR), Temperature-Programmed Desorption of CO2 (TPD-CO2), Scanning Electron Microscopy (SEM), Transmission Electron Microscopy (TEM), Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS), and Thermogravimetric Analysis (TGA). The catalytic tests were performed at 800 °C, under atmospheric pressure, and initially with a CH4/CO2 ratio of 1. For the best-performing catalyst, additional tests were conducted using different CH4/CO2 ratios and with the addition of O2. The results revealed that the creation of mesopores in both zeolites increased the initial conversion of CH4 and CO2. It was observed that the catalyst supported on the hierarchical ZSM-5 zeolite demonstrated superior performance, exhibiting greater stability and less carbon formation and sintering. This result was attributed to a better dispersion of the NiO/Ni2+ species, the optimization of the metal-support interface and the strong basicity of the material, with an increase in the accessibility of the active Ni sites. The CH4/CO2 ratio and the addition of O2 influenced the conversion, syngas selectivity, and carbon formation for this catalyst. On the other hand, the catalyst supported on the hierarchical USY zeolite presented inferior performance, due to its high microporosity, which limited the reducibility and the accessibility of the reactants to the catalytic sites. The preparation of new catalysts, modifying only the sequence of Ni impregnation in relation to the treatment steps, confirmed this behavior. The effect was more pronounced when Ni impregnation was carried out before or between the zeolite modification stages. The catalyst prepared with the USY zeolite synthesized with SCBA showed an increase in catalytic stability, with the presence of silica in its composition. The creation of hierarchical zeolites from residues, such as SCBA, offers a promising path for the development of more efficient and sustainable catalysts for DRM. Instituição: Universidade do Estado do Rio de Janeiro Tipo do documento: Tese Fri, 24 Oct 2025 00:00:00 GMT http://www.bdtd.uerj.br/handle/1/25204 2025-10-24T00:00:00Z