Doctorado en Ciencias de la Ingeniería (DOCI)

Permanent URI for this collection

https://repositorio.uandes.cl/handle/uandes/989

Browse

Browsing Doctorado en Ciencias de la Ingeniería (DOCI) by Issue Date

Filter results by year or month
Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    A novel fluid dynamic study of the gas-liquid flows in biotrickling filters through CFD simulations and digital imaging techniques
    (Universidad de los Andes, 2022) Carreño López, Felipe Antonio; Moreno Casas, Felipe; Vergara Fernández, Alberto
    Daily, tons of volatile organic compounds (VOCs) which negatively affect the environment and human health are emitted into the atmosphere from anthropogenic and natural sources. Biotrickling filtration (BTF) is becoming one of the most promising treatment technologies for odor control. Within the last decades, the treatment of pollutants have been studied, and diverse numerical models for predicting the mass transfer have been intensively developed. However, the current state of the art is mainly based on using the two-film, penetration, and surface renewal theories which do not account for local variations of the fluid velocities, physical properties, or flow regimes. To account for variations on the local physical processes, a detail description of porous media, the multiphase fluid dynamics, and the biomass film is required. This work investigates and extends a three-dimensional computational fluid dynamic (CFD) model coupled with computerized tomography (CT) with the novel incorporation of a contrast agent as a first attempt to assess the local biofilm formation inside a realistic porous structure used in biotrickling filtration of VOCs. The validation of these models was accomplished in terms of the gas and liquid phase residence time distribution (RTD), and the volumetric mass transfer coefficient. The gas phase RTD was obtained using a novel methodology based low cost MOx sensor; the liquid phase RTD was obtained from a methylene blue pulse method, while the mass transfer characterization was carried out by using the sulphite method. Finally, the column was operated for the treatment of toluene vapours and a contrast agent was added after reaching the steady state in order to obtain a 3D description of the local biofilm formation. These results were used to validate the CFD-CT models. The mean RTD and the normalized variance estimated in the simulation were 43.709 s and 0.326, respectively. Compared with the experimental results, a relative difference of 4.167% for the mean RTD and 32.515% for the normalized variance were found. The computed surface area was available for biodegradation was 0.366 m2. This work results in a validated gas RTD model, whereas for the liquid RTD and mass transfer coefficient the proposed approaches seem promising but requires additional computational resources to assess the steady state behavior. This methodology demonstrated the feasibility to obtain the local biofilm formation but additional imaging procedures are required to reconstruct the closed manifold geometry to use this image as a computational mesh.
  • Thumbnail Image
    Item
    Study of alternative metabolic pathways for the production of (R)-3-hydroxybutyric acid in polyhydroxybutyrate producing bacteria
    (Universidad de los Andes, 2022) Yañez Meneses, Luz Francy
    Considerable rich literature has accumulated concerning biochemical, physiological, and genetic aspects of polyhydroxybutyrate (PHB) intracellular accumulation in bacteria. The costs of substrates and processing, including the extraction of the polymer accumulated in intracellular granules, still hampers a more widespread use of this family of polymers. The PHB monomeric unit, (R)-3-hydroxybutyric acid (R3HBA) has found uses at the biomedical, chemical and supplement industries. The literature shows that two main process engineering and metabolic engineering strategies have been identified aimed at the production of chiral R3HBA: (i) production from the accumulated polymer (polymerization and depolymerization system, PDS); (ii) by bypassing the accumulation of PHB using metabolically engineered bacteria. The later includes the use of thioesterases (thioesterase shortcut system, TSS) that removes CoA from R3HBA-CoA, resulting in the R3HBA release to extracellular medium. This PhD thesis aims at broadening the understanding of the genetic and operational factors leading to PHB polymerization and R3HBA production in Azohydromonas lata DSM 1123, Cupriavidus necator H16 and Methylocystis parvus OBBP. Results showed that the growth associated PHB production observed in A. lata mimics an overflow metabolism, additionally, a successful PHB depolymerization in a two stage chemostat was obtained. The feasibility of producing R3HBA through in-vivo depolymerization of the intracellularly accumulated PHB in M. parvus was investigated. A PHB to R3HBA conversion of 77.2 ± 0.9% (R3HBA titer of 0.153 ± 0.002 g L⁻¹) can be attained in a mineral medium containing 1.0 g L⁻¹ KNO₃ at 30 °C with shaking at 200 rpm and a constant pH of 11 for 72 hours. Nitrogen deprivation, oxygen limitation, the supplementation with exogenous R3HBA and neutral or acidic pHs strongly reduced the excreted R3HBA concentration and yield. The implementation of the TSS system in M. parvus and C. necator by the construction of an expression vector containing tesB was hampered by inconsistencies in the constructed plasmids pLY01 and pLY02. Finally, the production of R3HBA by redirecting fluxes in the PHB metabolic pathway was investigated in C. necator; two mutant strains were constructed using the suicide vector pT18mobsacB: C. necator ∆phaC and C. necator ∆phaC ∆hbd, both unable to polymerize PHB and the last one incapable to transform R3HBA into acetoacetate. The mutant trains released pyruvate and R3HBA, suggesting that a native thioesterase of C. necator may play a role in the release of R3HBA by removing CoA from 3HBA-CoA. A protein homology on the genome of C. necator showed an enzyme encoded as WP_037025319.1 with a percent identity of 44 % in comparison with Ycia that may trigger R3HBA release. The results obtained in this work demonstrated the feasibility of R3HBA production by reducing or eliminating the fluxes of the reactions consuming R3HBA via operational manipulation as described in M. parvus and A. lata or via gene knock outs in C. necator.
  • Thumbnail Image
    Item
    Modelos de pronóstico con machine learning para apoyar la toma de decisiones
    (Universidad de los Andes, 2023) Karmy Diban, Juan Pablo; Pérez, Juan; Vairetti, Carla
    El analisis de series de tiempo es una tarea fundamental dentro de la ciencia de datos. Ser capaces de almacenar, procesar, y entender los datos históricos nos permite utilizarlos como insumo para el proceso de toma de decisiones. Normalmente las series de tiempo poseen, ya sea implícita o explícitamente, características que permiten realizar su análisis en base a una estructura jerárquica. Esta tesis se centra en el estudio de estas series de tiempo, y en la posibilidad de obtener mejores estimaciones de las mismas, mediante el estudio de sus jerarquías. Este trabajo presenta cuatro investigaciones. Dos de ellas poseen una contribución científica aplicada, mediante la predicción de series de tiempo jerárquicas (HTS) en la industria del Travel Retail y en un Contact Center, respectivamente. Las otras dos investigaciones entregan una contribución metodológica, mediante la incorporación de la estructura jerárquica de las series en un algoritmo de Machine Learning, buscando generar nuevos modelos especializados en la estimación de este tipo de series temporales. En la primera contribución aplicada, se generan estimaciones de venta de productos en la industria del Travel Retail mediante el enfoque clásico ε-SVR, pero adaptado a tres propuestas de algoritmos para su utilización en HTS. En este caso, nuestra metodología propuesta, SVR-BU, obtiene los mejores resultados, superando incluso a las técnicas clásicas ARIMA y Holt-Winters en términos de mean absolute percentage error (MAPE). En la segunda contribución aplicada, se realiza la estimación del volumen de llamadas entrantes en un Contact Center mediante los mismos algoritmos propuestos en el trabajo anterior, y nuevamente nuestra metodología es la que obtiene los mejores desempeños en términos de MAPE. En este caso, la contribución corresponde a la utilización de estos resultados en un modelo de optimización estocástica de staffing propuesto. Las dos contribuciones metodológicas corresponden a nuestras propuestas θ-SVR y KAT-SVR. Ambos métodos pensados como una extensión de ε-SVR para HTS. La idea es agregar información a través de los niveles jerárquicos, previniendo que las series de los niveles inferiores se desvían mucho de las series de los niveles superiores. Buscamos construir un modelo capaz de lidiar con el ruido intrínseco de los niveles inferiores. La gran diferencia entre ambas propuestas radica en que KAT-SVR incorpora métodos de kernel a los modelos reviamente presentados, con el objetivo de otorgarles mayor flexibilidad, mediante la posibilidad de modelar de mejor manera problemas más complejos, con características no lineales. Los modelos presentados en ambos trabajos, θ-SVR y KAT-SVR, obtienen los mejores desempe˜nos en términos de MAPE al ser comparados con ε-SVR tradicional, ARIMA, y Holt-Winters.
  • Thumbnail Image
    Item
    Analítica prescriptiva basada en inteligencia computacional para maximizar la rentabilidad
    (Universidad de los Andes, 2024-09) Meza Angulo, Armando De Jesús; Pérez R., Juan.
    En el contexto actual, las organizaciones enfrentan desafíos en su proceso de toma de decisiones. La dinámica del mercado, la variabilidad en la demanda y la competencia exigen enfoques más sofisticados. La toma de decisiones debe considerar múltiples criterios y gestionar la incertidumbre en la información. Por lo tanto, la capacidad de implementar modelos analíticos y metodologías robustas es fundamental para optimizar operaciones y maximizar rentabilidad. Esta tesis propone modelos y metodologías para mejorar la gestión de productos sustitutos y las estrategias de retención de clientes en entornos empresariales dinámicos con información limitada. Se presentan dos investigaciones interrelacionadas que convergen en optimizar la toma de decisiones, además de una tercera en desarrollo. La primera investigación optimiza la utilidad del consumidor mediante un modelo dinámico de lot-sizing, para maximizar las ganancias al optimizar precio, producción y almacenamiento. El modelo emplea un algoritmo de Optimización por Enjambre de Partículas (PSO) para determinar óptimamente estos parámetros, a través de ecuaciones de punto fijo. Se abordan conjuntamente decisiones de almacenamiento, producción y fijación de precios de productos sustitutos en un marco innovador. La contribución es científica aplicada, desarrollando un modelo que integra un análisis multiproducto y considera la demanda no fija y probabilística, contrastando con modelos tradicionales de demanda determinista. Este enfoque representa más realísticamente las decisiones de los consumidores, mejorando la respuesta empresarial, proporcionando soluciones prácticas para la gestión de inventarios y precios en mercados competitivos. La segunda investigación diseña y optimiza campañas de retención de clientes, mediante un Sistema de Inferencia Difusa tipo Mamdani y modelos de optimización que consideran el abandono de clientes (churn) en función de un umbral de decisión. Este enfoque maximiza la utilidad de las campañas basándose en el análisis de la información del cliente y los retornos marginales. Esta contribución es metodológica, desarrollando técnicas para categorizar clientes según su propensión a abandonar, permitiendo estrategias de retención más efectivas. Al ajustar intervenciones según el comportamiento previsto, esta metodología maneja la incertidumbre en la información, clave en la toma de decisiones en entornos dinámicos. Los resultados en simulaciones numéricas demuestran la efectividad y adaptabilidad de ambos modelos en distintos contextos, destacando su relevancia en la toma de decisiones. Finalmente, se proponen futuras extensiones como algoritmos paralelizados y validación empírica. En resumen, esta tesis ofrece herramientas aplicables en el ámbito empresarial para mejorar la toma de decisiones y optimizar la utilidad.
  • Thumbnail Image
    Item
    Efficient uncertainty quantification and propagation in performance-based earthquake engineering
    (Universidad de los Andes, 2025-04) Birrell Arangua, Matías; Astroza Eulufí, Rodrigo
    In recent decades, the constant deterioration of existing infrastructure and the increasing exposure to natural hazards driven by geological processes and changing climate conditions have motivated the development of a new philosophical approach to structural engineering, known as performance-based engineering. Its goal is to provide a rigorous, science-based framework through a comprehensive assessment of structural risk, ultimately delivering a decision variable that is useful for practical decision-making. To this end, performance-based engineering establishes a probabilistic framework that aims to address uncertainty regarding (i) the hazards to which the structure is exposed, (ii) the actual behavior of the structure versus that predicted by the engineering model, and (iii) the damage caused when certain intensity levels are exceeded. At each of these stages, properly quantifying uncertainty and subsequently propagating it through the following stages is critical for a successful risk assessment. In this context, methodological progress has been gradual, supported by technological advances that have enabled the implementation of probabilistic methods. However, the cost of adopting a probabilistic framework has been high, especially due to the need for largescale simulation of finite element models, which requires significant computational and time investment. For this reason, developing methods that enable efficient uncertainty quantification and propagation in performance-based engineering remains an open challenge and a key area of current research. This thesis presents two approaches aimed at providing efficient methods for uncertainty quantification and propagation by supporting structural simulations with machine learning surrogate models using Gaussian processes. The first approach focuses on quantifying and decomposing parameter-induced uncertainty in structural responses under specific hazard scenarios. Its goal is to support probabilistic sampling-based analyses, including model calibration and updating, iterative performancebased design, and sensitivity analysis. The second approach focuses on the quantification, propagation, and decomposition of uncertainty in structural vulnerability assessment under a broad range of seismic events. This approach implements and discusses the performance-based engineering framework from a philosophical standpoint, although applied to a real-world case study. Available definitions of damage states in bridge components and the relationships between these and their consequences are discussed. Both approaches are developed in a fully probabilistic setting, including probabilistic seismic hazard analysis, probabilistic structural modeling, and uncertainty decomposition.