Maestría
Permanent URI for this communityhttps://hdl.handle.net/11285/551038
Colección de Tesis y Trabajos de grado presentados para obtener una Maestría del Tecnológico de Monterrey.
Browse
Search Results
- Evaluating Pre-trained Neural Networks in Deep Learning for Early Detection and Enhanced Screening of Cervical Pathology(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12) González Ortiz, Orlando; Muñoz Ubando, Luis Alberto; emimmayorquin; Raymundo Avilés, Arturo; Cerón López Universidad, Arturo Eduardo; School of Engineering and Sciences; Campus Monterrey; Ochoa Ruiz, GilbertoThis document presents a research thesis for the Master in Computer Science (MCC) degree at Tecnologico de Monterrey. Cervical cancer remains a leading cause of mortality among women, particularly in low-resource regions where screening tools such as the Pap smear often fall short in early detection. This research explores the application of deep learning and pre-trained neural networks for the binary classification of cervical pathology, focusing on detecting dysplasia, specifically CIN2 and CIN3, as a potential prevention tool. We im- plemented multiple neural network models, including DenseNet, EfficientNet, MobileNet, and ResNet. The models were evaluated on two distinct datasets: one from the International Agency for Research on Cancer (IARC) and another from the Zambrano Hospital. To as- sess the generalization capacity of these models, we employed a sequential training approach where the first batch was trained with IARC data and tested on a Zambrano Hospital batch, with subsequent tests progressively incorporating prior results. Each experiment was repeated over 10 iterations to calculate confidence intervals for the performance metrics. Our results demonstrate that DenseNet and EfficientNet outperformed other models, achieving superior sensitivity and accuracy compared to conventional Pap smear tests. These findings indicate that deep learning models hold promise as an affordable, effective cervical cancer screening tool in low-resource communities. Future work will focus on augmenting datasets through collaboration with healthcare institutions and exploring generative models such as GANs to improve model robustness and generalization.
- Assessment of Alzheimer's disease-related blood and urine biomarkers for wastewater-based epidemiological studies(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-11) Armenta Castro, A.; Aguilar Jiménez, Osear Alejandro; emimmayorquin; Montesinos Castellanos, Alejandro; Flores Tlacuahuac, Antonio; School of Engineering and Sciences; Campus Monterrey; de la Rosa Flores, Orlando DanielIncidence of Alzheimer's disease, the leading cause of dementia and the fifth cause of death among elderly patients, has been rapidly increasing in recent years due to continued demographic aging. However, access to diagnosis and adequate care remains limited, especially in low-to-middle income countries, leaving an approximate 41 million cases currently undiagnosed. Such limitations can crucially compromise the quality and availability of care that can be provided to those in need. Wastewater surveillance, which is based on the detection and quantification of biomarkers in wastewater samples, has emerged as a promising tool to assess public health in a time and resource-efficient manner, providing important information for public health authorities and healthcare providers when used in tandem with relevant socioeconomic data and clinical reports. While its potential for monitoring infectious diseases has been proven, efforts towards the integration of biomarkers of chronic and degenerative diseases into such surveillance platforms are still needed. This dissertation aims to evaluate the main biomarkers related to Alzheimer’s disease, including proteins, long non-coding RNAs, and oxidative stress biomarkers, for their integration into wastewater surveillance biomarkers. Moreover, machine learning-based algorithms to correlate the concentration of biomarkers in wastewater to the clinical reports of incidence of a disease were developed using SARS-CoV-2 surveillance in university campuses across Mexico as a relevant case study, to develop effective data analysis strategies to integrate wastewater surveillance data into epidemiological models that allow for public health risk assessment and forecasting. This dissertation contributes to the consolidation of wastewater surveillance as a tool for comprehensive public health risk assessment and data-driven decision-making by demonstrating a pipeline for the integration of new biomarkers into surveillance platforms and effective, easily-interpretable data integration.
- Uso de materiales alternativos en la fabricación de componentes de maquinaria(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-02) Roch Carbajal, Fernando; Elías Espinosa, Milton Carlos; emimmayorquin; Perdigón Lagunes, Pedro; Escuela de Ingeniería y Ciencias; Campus Ciudad de México; Miró Zárate, Luis AngelEn esta tesis se aborda uno de los principales problemas en el ámbito de los componentes de máquinas: las consecuencias de su falla. A menudo, estas fallas causan daños significativos en la maquinaria. Para mitigar este problema, se propone el uso de material flexible TPU como sustituto en la fabricación de estos componentes. Además, se sugieren distintos diseños de geometría interna que podrían influir en los resultados. Se realizaron pruebas con engranes impresos mediante FDM en TPU flexible, evaluando la deformación del material bajo peso. Asimismo, estos pesos se simularon en Simcenter NASTRAN para validar los datos experimentales. Los resultados de ambas pruebas, físicas y las realizadas por simulación, demostraron que este material es bastante resiliente. En situaciones de falla, las piezas fabricadas no presentaron las fallas comunes en engranes de otros materiales y retornaron a su forma original sin indicios de daño permanente.
- Development and testing of Spirometer with pulmonary rehabilitation for patients of Amyotrophic Lateral Sclerosis(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-04) Perez Ortiz, Claudia Xochitl; Antelis Ortiz, Javier Mauricio; emimmayorquin; Mendoza Montoya, Omar; School of Engineering and Sciences; Campus Monterrey; Caraza Camacho, RicardoAmyotrophic Lateral Sclerosis is one of the most aggressive neurological diseases affecting the lower and upper motor neurons, it diseases and eventually kills the motor neurons, leaving the patient unable to walk, move, talk, and eventually breathe. For this reason, the main cause of death in ALS is respiratory failure. However ALS patients usually only see specialized health assistants every 2 to 3 months in ALS clinics. Inspiratory Muscle Training (IMT) is a form of resistance workout for the lungs, and it has been found to increase survival in ALS individuals for 12 months. Spirometers, devices that measure lung capacity, could help patients measure their lung state, and adjust therapies accordingly at home. For this reason, an automatic spirometer prototype that can record the state of the lungs, adjust itself, and perform IMT rehabilitation in ALS patients is proposed. Results show that the proposed spirometer prototype could measure FVC and PEF with an average accuracy of 96.98% and 92.6% respectively, and could improve FVC by 13.7%, and FEV1 by 13.6% with inspiratory incentive training.
- Improving the design of multivariable milling tools combining machine learning techniques(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-05) Ramírez Hernández, Oscar Enrique; Olvera Trejo, Daniel; emipsanchez; Puma Araujo, Santiago Daniel; Martínez Romero, Oscar; School of Engineering and Sciences; Campus Monterrey; Fuentes Aguilar, Rita QuetziquelChatter in milling operations degrades surface quality, compromises dimensional accuracy, accelerates tool wear and may damage spindle components. One effective strategy to mitigate chatter while maintaining high productivity is the use of specialized milling tools, such as multivariable milling cutting tools (MMCT), designed with variable geometry in their pitch (𝜙) and helix (β) angles. However, identifying the combination of these angles remains challenging because of the absence of analytics models that link MMCT geometrical parameters with dynamic stability limits. This study proposes a novel approach that integrates analytical lobes calculation with machine learning to enhance tool design efficiency. We find optimal tool geometry (pitch and helix angles) and cutting conditions (spindle speed and axial depth) to maximize the Material Removal Rate (MRR) in milling of a single degree of freedom. Our approach employs a genetic algorithm (GA) combined with a pattern recognition neural network (NN) to predict whether specific parameter combinations will yield stable or unstable behavior. The Multilayer Feedforward Neural Network is trained using a database generated from simulation of a SDOF mathematical model of milling, a non-autonomous Delay Differential Equation. The solution to the DDE is approximated through the Enhanced Multistage Homotopy Perturbation Method (EMHPM). The database includes 23,606,700 observations, covering a catalog of 36,318 MMCT configurations and 650 cutting conditions (axial depth of cut and spindle speed) for each tool configuration. The NN training database uses an approach for handling variable cutting coefficients based on exponential fitting model to describe their variation. These coefficients were characterized at small radial immersion of 1.86 mm using cutting forces of five MMCTs with a diameter of 0.5 in. This approach accurately predicts cutting forces, achieving an NRMSE below 10% when compared with experimental signals. The trained NN estimates the stability of the milling process with an error of 3.3%. Additionally, the combined use of the NN and GA reduces computation time by 98% compared to the GA with EMHPM. The selection of five combinations of geometric parameters that maximize MRR in a range between 26% and 120%, compared to the MRR of a regular tool, which is 190,493 mm³/min, has been performed. The rate of increase in MRR depends on each of the five selected geometries (see Chapter 5). Moreover, without the proposed approach, identifying the improved geometry would require up to 25 days using an exhaustive search scheme, where a SLD is generated for 10,000 cutting conditions for every tool configuration.
- Development of polyethylene fibers using extrusion for the projection of its implementation on textiles(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12) Martínez Hernández, Saúl; Treviño Quintanilla, Cecilia Daniela; emimmayorquin; Lozano Sánchez, Luis Marcelo; Franco Urquiza, Edgar; Martínez Franco, Enrique; Burelo Torres, José Manuel; School of Engineering and Sciences; Campus Querétaro; Treviño Quintanilla, Cecilia DanielaThe global textile industry faces significant challenges due to unsustainable practices, including extensive resource consumption and substantial waste generation. This thesis investigates the development of polyethylene (PE) fibers using extrusion techniques to address the demand for durable, lightweight, and sustainable fibers. The choice of PE is driven by its favorable optical properties, availability, and compatibility with textile production requirements. This research optimized extrusion parameters—screw speed, heating zone temperature, cooling rate, and collection speed—to produce fibers with a target diameter of 15 μm, achieving final diameters of up to 8 μm. A coextrusion approach was utilized, creating core-shell fibers with PE as the core and polylactic acid (PLA) as the shell, enabling precise diameter control. The PLA shell was removed through chloroform dissolution. Fibers with and without the shell were characterized using differential scanning calorimetry (DSC), Fourier-transform infrared spectroscopy (FTIR), thermogravimetric analysis (TGA), and scanning electron microscopy (SEM) to evaluate their composition and structural integrity. The fibers were woven into textiles using a table loom, tested for wicking properties, and compared against existing textile alternatives. SEM analysis provided detailed structural insights into the woven samples. Results demonstrate the potential of these fibers as a sustainable alternative to conventional textiles, with promising performance in wicking tests. Further optimization and exploration of production methods are necessary to enhance their viability for industrial applications.
- Magnetic gripper design optimization for robotic bending cell using artificial intelligence clustering of sheet metal parts(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-11-25) Treviño Treviño, Ana Paula; Ahuett Garza, Horacio; emipsanchez; Urbina Coronado, Pedro Daniel; Orta Castañón, Pedro Antonio; School of Engineering and Sciences; Campus MonterreyThe manufacturing sector is currently facing unprecedented challenges in adapting to the constantly evolving demands of diverse product lines and rapid market changes. Conventional manufacturing systems are struggling to adapt to the increasing variety of production components, leading to notable inefficiencies and heightened expenses. In this context, Reconfigurable Manufacturing Systems (RMS) have emerged as a prominent strategy to boost the adaptability and responsiveness of production processes. Therefore, the design and optimization of grippers for robotic arms are deemed essential to improve efficiency and productivity. The project aims to enhance gripper design by using AI clustering techniques and dimensional analysis to cluster production components and define design parameters for novel gripper configurations. This approach aligns with the tenets of lean manufacturing and data-driven decision-making, empowering manufacturing engineers and designers. The project also aims to optimize internal design and manufacturing, reducing reliance on external suppliers, and improving long-term adaptability and competitiveness by leveraging the cost reduction that in-house processes represent. The case study examines 964 sheet metal production components, highlighting inefficiencies of manual classification, part allocation challenges, and design specification retrieval. Furthermore, it explores different scenarios to render the best cluster quality possible with the supplied dataset and the constraints that materialize when translating the design parameters into actual design properties of the grippers, as well as the gripper-part compatibility. The thesis introduces an innovative method for managing part variety in gripper design by seizing advanced technologies and data-driven decision-making. This results in substantial enhancements in time efficiency, cost reduction, safety optimization, and the eradication of inefficient workflows within the manufacturing sector.
- Design and Optimization of a Neuromorphic Controller for Ackermann Vehicles in Mixed Reality Environments(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-11-03) De La Trinidad Rendón, Julio Sebastían; Reyes Avendaño, Jorge Antonio; emimmayorquin; Carrillo Martínez, Luis Antonio; School of Engineering and Sciences; Campus Ciudad de México; González Hernández, Hugo GustavoDesigning autonomous navigation systems is a challenging task that must account for demands such as high power consumption, computational load, and system requirements, as well as potential constraints like limited resources. To address these challenges and the complexities of traditional control and artificial intelligence approaches—such as reliance on dynamic or kinematic models and the low explainability of AI—this thesis proposes the design, testing, and optimization of a neuromorphic algorithm for path tracking and obstacle avoidance in Ackermann-type vehicles. The testing and optimization process is supported by a Mixed Reality (MR) testing framework. The proposed algorithm is inspired by the whisker-mediated system found in rodents and other mammals. It leverages spiking neural networks (SNNs) to unify path tracking and obstacle avoidance under a single, interpretable framework. Optimizable control gains allow the algorithm to be tailored for specific hardware and improve responses to stimuli, enabling smoother navigation. The accompanying MR framework enhances testing by combining real sensor data with synthetic elements to generate semi-synthetic point clouds. This creates dynamic environments with both physical and virtual obstacles, eliminating the need for full-scale prototypes or costly setups. The framework also collects key data and metrics in real time, integrating environmental information through a digital twin. The experimental setup combines physical and virtual environments using a Quanser QCar, a 1:10 scale vehicle equipped with sensors such as a planar PRLiDAR. Positioning is provided by an OptiTrack Motion Capture System, enabling real-time data collection. The MR environment integrates sensor data with synthetic elements through a Unity-powered simulation environment and ROS communication. Optimization results demonstrate the neuromorphic algorithm’s ability to maintain trajectory adherence while responding effectively to obstacles, even under constrained conditions. The MR environment supports precise, repeatable testing, validating the system’s robustness. By addressing the dual challenges of algorithm complexity and testing scalability, this thesis contributes to efficient and accessible solutions for autonomous vehicle navigation and evaluation.
- The Rising Sea-Level Caused by Climate Change in Quintana Roo, Mexico: A Model-Based Study of Vulnerability and Adaptation(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12-02) Mondragón Muñoz, Valeria Fernanda; Cervantes Avilés, Pabel Antonio; emimmayorquin; Batalini de Macedo, Marina; Campus Ciudad de México; Huerta Aguilar, Carlos AlbertoQuintana Roo is a state in Mexico with a population of 1.9 million inhabitants. This state has an extensive coastal region widely known for its tourist destinations such as Cancún, Tulúm, and Playa del Carmen, among others. However, the impacts of climate change pose increasing challenges for this region, such as the rise in sea level. The main objective of this work was to determine the risks posed by rising sea levels and how they intersect with local economic and environmental factors. This study determined climatic anomalies using a climatic model (CMIP6), created a sea-level rise simulation, and assessed potential adaptation strategies. The research revealed that the temperature anomaly in a stabilization scenario could increase to 2°C by 2100, aligning with the goal of the United Nations in the Paris Agreement. Nevertheless, it can reach 4°C in a pessimistic scenario by 2100. Precipitation modeling results indicate that the rainfall may decrease by up to 10 mm/month in a stabilization scenario and up to 28 mm/month in a pessimistic scenario by 2100. The expected changes in these climatic conditions pose significant threats to the stability of the local ecosystems and communities in the area, such as the increase in the frequency and intensity of cyclones, heat waves, wildfires, coastal and fluvial flooding, and the elevation in sea level. About this last which is the aim of this work, the simulation of sea-level rise revealed that significant zones that hold both touristic and habitational infrastructure would be affected by this phenomenon. Considering that millions of people visit the destinations in Quintana Roo every year and that 85% of the state GDP depends on tertiary activities related to tourism, the activities related to this sector and its economy are unequivocally vulnerable to climate change hazards. To adapt the vulnerable areas against the potential adverse effects several actions in the coastal zones are considered. However, it is crucial to consider the knowledge gained through this study for making informed decisions and promoting effective strategies to protect the region and ensure sustainable development.
- Social media to predict the 2024 mexican presidential election: a three model approach(Instituto Tecnológico y de Estudios Superiores de Monterrey, 2024-12) Gutiérrez Valenzuela, Héctor Abel; Zareel, Mahdi; emipsanchez; Sánchez Ante, Gildardo; Biswal, Rajesh Roshan; School of Engineering and Sciences; Campus Monterrey; Brito, Kellyton(Only 1 page) The appearance and rise of social media have evolved the way people interact with each other. From interpersonal communication to mass media production, social media applications have shifted the approach to how an individual or a complete corporation could generate and propagate a message. It was just a matter of time before this new way of reason- ing communication influenced political messages too. Ever since Obama´s 2008 and 2012 victories, the role social media could play in a presidential election was evident. More re- cently, the 2016 Cambridge Analytica scandal ultimately defined how influential the content people see on social media could be. Numerous research has emerged aiming to foresee these political movements based on online performance and many methods have been proposed. The definitive, most common pattern in these works is sentiment analysis in social media posts. The process is simple: detect how many people ’like’ a candidate´s online presence, and how many don´t, and this will likely represent the outcome of an election. However, this approach has sparked both criticism and unsatisfying results. The following work considers a contemporary approach to predicting elections with social media and collected polls. This strategy has succeeded in the case of various Latin American countries such as Argentina, Brazil, Colombia, and Mexico. However, we aim to identify potential flaws and improvements in the method to prove a concrete methodology can work outside a single election exercise, a repetitive cause for concern for multiple experts in the field. Results show that some replicated experiments do not successfully predict the result of the 2024 Mexican presidential election, as in 2018. However, we prove concrete method- ologies and models, like the multi-layer perceptron model (MLP) can successfully predict electoral results in more than one election. Moreover, we propose the least absolute shrink- age and selection operator (LASSO) to construct better and more descriptive predictors for electoral results. These two utilized implementations accurately predicted the winner of the 2024 election but remained short of the official performance of the winning candidate, Claudia Sheinbuam. In the case of the second and third place, both models merely missed the official result by 3 points.
