In this work we compare the performance of a multilayer neural network with the classic GARCH model when both are applied to a regresion problem of a univariate time series. To make sound decisions it is of main importance To make sound decisions, it is of main importance to predict the future behaviour of the main variables. The disposal of huge amounts of data and the non linearity of the correlation of the time series require the use of neural networks, whose architecture demands machine learning techniques. For the input layer the ACF and the PACF correlation function are needed, and for the number of neurons in the hidden layers the version of the method of cross validation for time series is used. We present the results with real data that show better performance of the resultant neural network in comparison to the GARCH model.
Uncertainty estimation is a necessary step in any geophysical study before drawing interpretations. In particular for seismic slip inversion studies, inaccurate slip assessments can lead to misleading interpretations of the tectonic kinematics. In recent decades, seeral strategies for uncertainty evaluation, under different theoretical frameworks, have been proposed, but they have not yet been systematically compared. In this work, we propose specific indices for four different strategies to assess the uncertainty in the seismic slip inversion. These strategies include: (1) the marginal distributions obtained via Bayesian inference, (2) the Hessian matrix, (3) the resolution matrix, and (4) the mobile checkerboard tests. These strategies were applied to two canonical synthetic cases: the first representing a reverse fault, and the second, a strike-slip fault. Additionally, the uncertainty analysis was performed for a real case: the 2021 Acapulco earthquake (Mw 7.0). The results reveal a similarity between the uncertainty distributions obtained using Bayesian inference and those derived from the Hessian matrix. Likewise, comparable results are observed between the mobile checkerboards and the resolution matrix. These similarities are attributed to the theoretical foundations and methodological approaches underlying each strategy. Although the employed strategies do not produce identical uncertainty distributions, all indicate lower uncertainty in regions with higher station density and proximity to the fault. The results confirm that uncertainty is not driven by the data themselves, but rather by the fault geometry and the distribution of measurement instruments.
We aim to demystify PINNs and enhance their practical utility. A simplified model of the Tacoma Narrows Bridge oscillations---formulated as an ordinary differential equation (ODE)---serves as our central, recurring example. While this is an ODE, the methodological framework, challenges, and solution strategies we discuss are directly applicable to the more general class of partial differential equation (PDE) problems, treating ODEs as a particular case of PDE where there is no space-dimension, in this particular case the only variable is time. We argue that the conventional interpretation of PINNs as regularized regression is mathematically imprecise. Instead, we posit that they fundamentally solve PDE-constrained optimization problems. This perspective allows us to unify concepts from Tikhonov regularization, numerical optimization principles, and modern PINN architectures. Through a consistent case study with four distinct methods---Baseline (standard penalty method), Tikhonov regularization, Augmented Lagrangian, and Curriculum Learning---we demonstrate the practical implications of this viewpoint. Our computational experiments, conducted with a unified PINN architecture featuring Fourier feature embeddings and trained for 20,000 epochs across all methods, reveal that the Augmented Lagrangian method achieves superior performance with an L2 error of $3.59\times10^{-4}$, outperforming the baseline by a factor of 2.89$\times$ (i.e., nearly three times more accurate). While this improvement is substantial, it is more modest than theoretical extremes, demonstrating that the baseline penalty method already performs reasonably well on this problem with appropriate tuning. These results validate that techniques grounded in classical optimization can address known limitations of the standard penalty-based formulation, though the margin of improvement depends on problem characteristics. The article offers actionable implementation guidelines, analyzes optimization challenges, and suggests future research directions bridging numerical analysis, optimization, and machine learning, thereby establishing a principled framework for PINN development.
A common test for the diagnosis of type 2 diabetes is the Oral Glucose Tolerance Test (OGTT). Recent developments in the study of OGTT tests have framed it as a Bayesian inverse problem. These data analysis advances promise great improvements in the descriptive power of OGTTs. OGTT tests are typically done with invasive, bothersome, and somewhat expensive venous blood tests. A natural question is whether improved data analysis techniques would allow for less invasive and cheaper glucometer measurements to be used. In this paper we explore this question. Using one dynamic model, we develop an error model for glucometer capillary blood sugar measurements and compare results of venous blood sugar tests for 65 patients, finding a match in over 90\% of observed cases. Our conclusion suggests that this model (or one much like it) may permit capillary glucose to be used with reasonable accuracy in performing OGTTs.
This article provides a theoretical analysis of the concept of educational quality from a systemic perspective, integrating institutional, pedagogical, sociocultural, critical, and assessment approaches. Through a literature review, it highlights how the notion of educational quality has evolved beyond academic efficiency to encompass key dimensions such as equity, inclusion, and social impact. Based on this analysis, a new multidimensional definition of educational quality is proposed, incorporating strategic management, pedagogical relevance, and cultural diversity. This definition differs from traditional models by prioritizing the participation of all educational stakeholders. Finally, the article discusses the challenges of implementing this vision in real educational contexts and its potential influence on policies and practices to ensure more inclusive, equitable, and sustainable learning experiences.
In this work is applied the robust stability for a PI-Posicast control with constant perturbations. Using the Lyapunov-Krasovskii approach, it is introduced a functional that leads to perturbation conditions and robustness levels. The functionals are applied to the estimation of the domain of attraction and of the system solutions. Finally, an illustrative example is given and in conclusion, it is observed that the robustness results obtained through temporal analysis using complete type functionals are much more conservative than those obtained in the frequency domain.
Este estudio tiene como objetivo analizar cómo diferentes patrones del tiempo de entrega de un proveedor impactan al costo del inventario. Se presentan 6 distribuciones diferentes de probabilidad del tiempo de entrega: determinística, normal, uniforme, exponencial, triangular y lognormal. La metodología para solucionar esta problemática ha sido la simulación, obteniéndose con ella para un modelo de inventario de revisión continua (Q, R), que hace un nuevo pedido por Q unidades cada que el nivel del inventario baja al punto de reorden R. Para un caso ilustrativo se aplica la simulación para obtener los valores de Q y R que optimizan el costo del inventario. Dentro de este costo se incluyen los rubros de pedidos, mantenimiento, faltantes y el de compra de los artículos, dado que el proveedor incluye descuentos por volumen. Asimismo, se considera una restricción en cuanto al nivel de servicio mínimo deseado. Para cada una de las 6 distribuciones de probabilidad, Q ha resultado constante, definida en su valor mínimo con el que el proveedor de los artículos oferta su mejor precio. Por su parte R sí ha cambiado para las diferentes distribuciones de probabilidad y varía lineal y directamente con la desviación estándar del tiempo de entrega. La distribución exponencial ha sido la de máxima variabilidad y costo del inventario, en particular en los rubros de mantenimiento y faltantes. Además, no ha habido una relación directa y lineal entre el costo del inventario y la desviación estándar del tiempo de entrega. Con esto queda claro que una buena gestión del inventario implica considerar la incertidumbre de varios parámetros, como es el caso de la demanda de ítems y el tiempo de entrega del proveedor.
En este estudio se investiga la confiabilidad de trabes de concreto reforzado sometidas a momento flexionante debido a cargas gravitacionales. La resistencia a la compresión simple del concreto se caracteriza probabilísticamente con muestras de un laboratorio localizado en Aguascalientes, México. No existiendo normatividad oficial en dicha ciudad para diseño de elementos de concreto, se consideran dos versiones de las normas de la Ciudad de México para determinar las funciones de estado límite y obtener las confiabilidades mediante metodologías distintas para dos tipos de secciones. Se determinó que la resistencia a compresión simple del concreto sigue una distribución de extremos tipo Weibull. Se concluye que el usar el formato de un solo factor de carga tanto para carga muerta como para carga viva, como en una versión anterior del reglamento de la Ciudad de México, puede conducir a valores no uniformes de confiabilidad para diferentes cocientes entre la carga viva y la carga muerta. Se concluye también que el formato del ACI y de la versión actual de las normas de la Ciudad de México, conducen a niveles mas uniformes del índice de confiabilidad, y se propone un formato que mejora aún mas dicha uniformidad, mediante una relación que es función del cociente entre carga viva y carga muerta. En lo concerniente a una sección “T” investigada, se observó una disminución en la confiabilidad estructural, lo que podría atribuirse al método empleado. Los resultados muestran que la confiabilidad de trabes de concreto podría depender de las características de materiales y resistencias de la región en estudio, e.g. el Bajío, y que distintos formatos de factores de carga y resistencia, así como expresiones para determinar la capacidad a flexión podrían usarse para conocer la confiabilidad implícita de elementos existentes, y para diseñar elementos de concreto nuevos para una confiabilidad objetivo.
This paper presents the results of the stabilization and regulation of a driverless two-wheeled vehicle considering a Linear Parameter-Varying (LPV) mathematical model. The controller design is based on the linear and time invariant systems theory, extending its analytical procedure towards an LPV model, and achieving the control of the vehicle at its upright position by means of the gain scheduling methodology considering the translational velocity of the vehicle as the scheduling variable. The development of a CAD model designed in SOLIWORKS and communicated to MATLAB-SIMULINK through the Simscape tool is presented, allowing the validation of the controllers designed in simulation. Unlike previous works around two-wheeled vehicles, two torques are applied to control the vehicle in a desired angular position, which is considered the main contribution of this paper.
El propósito de este artículo es mostrar el desarrollo de una aplicación móvil codificada en Java para Android implementando Realidad Virtual, la cual que tiene como objetivo el poder almacenar información tanto de estudiantes de Nivel Medio Superior (NMS), profesores de matemáticas, así como datos esenciales de la institución para poder ser medidos y expuestos a docentes y directivos. La aplicación es capaz de realizar la técnica estadística denominada Análisis de Componentes Principales (PCA), la cual sintetiza la información con la que la base de datos fue alimentada y extrae características generando vistas representativas, las cuales son combinaciones lineales de las originales logrando con ello reducir el número de dimensiones de las variables permitiendo identificar factores correlacionales. Además, realiza la proyección de los datos simplificados mediante gráficos 2D y RV haciendo uso de los cardboard generando una visualización inmersiva de los datos. Por lo tanto, esta aplicación pretende facilitar la detección de indicadores de aprendizaje de las matemáticas en estudiantes de NMS, de esta manera los usuarios podrán concentrarse en aquellos factores que más influyen. Este trabajo está limitado a mostrar el algoritmo que se codifico y la interfaz de la aplicación para la identificación de factores de aprendizaje en matemáticas.
Generative design, based on genetic algorithms, creates multiple design solutions for complex structural problems by setting parameters, constraints, objectives, and performance criteria. The aim of this work is to evaluate, through numerical methods, the lightweighting of components manufactured in structural steel by modifying the material, geometry, and the process of manufacture in a generative design system. By characterizing the mechanical properties of an existing structure using finite element methods, objectives and constraints were established to optimize the components through genetic algorithms. In this study, these components were evaluated under static compression loading conditions with different materials, ASTM A36 Steel and Polyphenylene Sulfide. The results showed maximum stresses of 11.20 MPa and 10.76 MPa respectively for each material, with a difference of 0.44 MPa. The optimization process generated 14 iterations for each component, meeting the algorithm's objectives to reduce mass and be manufacturable through injection molding. Resulted in a maximum stress of 11.27 MPa in the optimized components, without exceeding yield strength. Furthermore, a significant mass reduction was achieved, with an 88% reduction in component "A" and 79% in "B." The originality of the study lies in the use of advanced design techniques for optimization not only in terms of geometry, but also material and manufacturing processes. This provides new perspectives and solutions for the fabrication and design of lightweight components manufactured at high volumes. It is concluded that lightweighting of components is feasible through numerical methods and generative design, reducing the mass of the studied components by 80% without sacrificing their mechanical properties, and simultaneously generate geometries feasible for plastic injection molding.
Mexico's growing energy demand necessitates exploring sustainable generation options. This study performs a Life Cycle Assessment (LCA) of various generation technologies in the Mexican context, assessing the environmental impact of photovoltaic, wind, geothermal, hydroelectric, nuclear, biogas, coal, gas, fuel oil, and diesel systems throughout their life cycle stages, from material extraction and manufacturing to operation and disposal. The LCA results are then integrated into an Optimal Unit Commitment method on the IEEE 39-bus test system in the PowerFactory software to analyze the economic and environmental viability of the generation plants. The analysis shows that the most economically and environmentally viable technologies are wind and solar, both with zero operating costs and zero polluting emissions. As a complementary alternative, hydropower stands out for its reduced operating cost, which is competitive when considering the pollution fines associated with other technologies. The results also recommend reducing the capacity of coal power plants from 1,224 MW to 157 MW, to mitigate 〖CO〗_2 emissions and avoid additional pollution costs. The findings aim to inform policymakers and stakeholders in selecting environmentally responsible generation solutions that contribute to a sustainable and diversified energy mix for Mexico.
This study presents the development and validation of a low-cost wearable device designed to evaluate running technique and quantify training loads in amateur and semi-professional runners. The system incorporates four flex resistive sensors to measure the flexion angles of the hips and elbows, providing real-time biomechanical feedback. Additionally, it integrates heart-rate-based metrics to objectively assess training intensity using the Training Impulse (TRIMP) method. Four volunteers participated in the study: three semi-professional track athletes and one amateur runner. During training sessions, joint flexion data were acquired by an Arduino Nano microcontroller and transmitted via Bluetooth to a custom-developed mobile application for initial visualization. Post-training, data were processed using a MATLAB algorithm to calculate TRIMP values. Heart rate signals were obtained from a commercially available health band worn by each participant before and during the exercise protocol. The results demonstrate that the device is capable of reliably capturing joint movement patterns and estimating training intensity. Biomechanical differences were observed between participants, such as asymmetries in hip flexion and variations in arm movement, which were consistent with the coach's visual analysis. The heart rate data and resulting TRIMP scores also reflected the varying physical demands experienced by each runner, confirming the utility of the system for training quantification. This wearable solution offers a practical and accessible tool for monitoring both movement mechanics and physiological effort during training. Its ease of use, low cost, and capacity to provide objective feedback make it particularly suitable for athletes and coaches seeking to optimize performance while reducing the risk of injury. The integration of biomechanical and physiological data into a single platform represents a promising step toward smarter and more personalized training strategies in sports science.
The experimental study conducted in an automotive company aimed to optimize the process flow and redesign service lines in the final assembly area, thereby creating physical space for the integration of new lines. Three pivotal tools were employed: the Guerchet methodology, lean manufacturing techniques, and process simulation before physical alterations. The outcomes revealed a substantial increase in production, ranging between 17% and 20%, following the implementation of the proposed changes. Furthermore, there was a reduction of 25 square meters in the physical space occupied by the final assembly service lines. This productivity enhancement signifies a significant advancement in the automotive industry, marking a milestone in service parts efficiency. In addition to the productivity impact, these changes brought about economic savings for the company. Decision-making relied on positive simulation results, allowing physical modifications solely upon their proven effectiveness in a virtual environment. This strategy not only drove productivity improvements but also positioned the company on the path towards the digitalization of its processes, as a prelude to a transition towards Industry 4.0. This strategic approach not only demonstrates enhancements in operational efficiency but also underscores the value of prior planning and the gradual adoption of digital technologies in the current manufacturing landscape. The integration of these tools offers not only immediate improvements but also a long-term vision toward a more efficient and adaptable manufacturing environment.
En el presente trabajo se realiza el análisis esfuerzo - deformación del socket de una prótesis transtibial, que soporta cargas generadas por el peso corporal en los estados estáticos y dinámicos o marcha. Para el análisis de marcha se considera un ángulo de dorsiflexión de 10 grados, se modela y simula mediante elementos finitos FEA el socket del muñón de un paciente de 37 años, 1.74 m y 83.4 kg de peso corporal. Se consideran los materiales más comunes utilizados en la fabricación del socket y sus propiedades mecánicas se obtuvieron de literatura especializada del tema. El trabajo permite determinar, a través del diseño y selección del material para fabricar el socket, el espesor mínimo de las paredes que permita transmitir las cargas a los demás componentes de la prótesis transtibial y evitar la generación de fisuras durante la marcha. Se determina que cada tipo de material utilizado tendrá su propio factor de seguridad FS y que no es posible establecer uno general para todos los distintos tipos de material empleados. El espesor determinado permitirá ampliar el periodo de uso del socket, mejorando la comodidad y estabilidad del usuario al caminar, así como también, es importante tener en cuenta la flexibilidad que proporciona el material seleccionado para su fabricación.
El análisis y diseño de elementos de concreto reforzado requiere la curva esfuerzo-deformación (f-ɛ) y el módulo elástico del concreto simple (Ec), este influye en el cálculo de desplazamientos y en las propiedades dinámicas. Con el objetivo de caracterizar mecánicamente el concreto simple f'c=250 kg/cm2, empleado en Ometepec Guerrero, fabricado con agregados pétreos de los ríos Santa y Ometepec se elaboraron 105 probetas cilíndricas de 15x30cm divididas en 4 combinaciones, el cemento utilizado fue CPC 30 R, RS. Una vez conocidas las características físicas de los componentes del concreto, se diseñaron las mezclas mediante el método de volúmenes absolutos, obteniendo las cantidades de materiales requeridos. Los especímenes fueron ensayados en una prensa automatizada a una velocidad constante de acuerdo con la norma NMX-C-128-ONNCCE-2013. Las curvas esfuerzo-deformación(f-ɛ) fueron graficadas y se obtuvo el factor numérico K (relación módulo elástico y resistencia a compresión), el promedio de este fue superior a 14,000 y la deformación unitaria promedio resultó inferior a 0.003, ambos valores definidos por la NTC-DCEC-2023. Las curvas f-ɛ promedio de las combinaciones mostraron que el concreto elaborado con cantos rodados, presentaron un mayor Ec y una menor deformación (Ꜫo) comparado con las combinaciones donde se empleó agregado triturado. El modelo de Hognestad representa adecuadamente la curva f-ɛ para concretos con cantos rodados con el parámetro B igual a 2.00, y 2.20 en agregados triturados. El valor medio del módulo elástico fue 241,556.44 kg/cm2, mayor en 8% al valor normativo para agregado de origen basáltico (223,693.54 kg/cm2 ) con una resistencia media a compresión de 255.30 kg/cm2.
El presente trabajo describe el desarrollo de un sistema híbrido que combina plantas y tecnología electrónica avanzada para monitorear y analizar las señales bioeléctricas generadas por las plantas en respuesta a estímulos externos. Este sistema innovador utiliza un convertidor analógico-digital (ADC) ADS1256, controlado por un microcontrolador AVR Mega2560, lo que permite adquirir datos eléctricos de alta precisión desde las plantas, siendo la especie Epipremnum aureum la principal sujeta de estudio. Una característica destacada del sistema es la incorporación de nanopartículas de plata para mejorar la conductividad eléctrica de las plantas, lo que resulta en una mayor eficiencia en la captura de señales bioeléctricas. Los datos obtenidos se procesan en tiempo real mediante software desarrollado en Python, lo que facilita el análisis continuo de variables ambientales como la temperatura y la luz. Este avance tiene amplias aplicaciones potenciales en áreas como la agricultura inteligente, el monitoreo ambiental y el desarrollo de biosensores basados en plantas. A diferencia de los métodos convencionales de sensado, el uso de plantas como biosensores ofrece una solución más sostenible y con menor impacto ambiental. Además, el sistema abre nuevas posibilidades para la investigación de las interacciones bioelectrónicas en plantas y su adaptación a diferentes condiciones ambientales. El trabajo presenta un enfoque novedoso dentro del campo emergente de la Cyborg Botany, lo que sugiere que en el futuro estas tecnologías podrían integrarse en redes de monitoreo a gran escala para ofrecer soluciones más ecológicas en la gestión del medio ambiente.
The Engineering Tower is a ten-story building, which includes a six-story office block in Ciudad Universitaria, in the south of Mexico City, and has been fully operational for the last twenty years. Passive design is eminently sustainable in the use of water and energy and in the control of comfort. The building includes a central atrium, which induces a permanent air current, generating all the necessary air changes without mechanical assistance. Occupants have adopted practices to operate windows in a manner that ensures a satisfactory interior comfort temperature, resulting in highly sustainable year-round thermal and comfort operation. A database of air temperature and humidity on various floors and offices, as well as outside, has been recorded since the beginning of commercial operation, confirming some of the best architectural practices, such as proper porous external skin performance, induction of vertical currents in the atrium and air flow controls to limit air exchange with the environment. A mathematical model is proposed that correlates the main parts of the building and their energy flows, and the results are compared with measurements from several years. It is proposed that the model, which is appropriately validated, will allow architects to design future buildings in a similar climate, with the same degree of sustainability. Overall, the study validates the use of an atrium to provide year-round passive comfort
El lavado industrial de botellas de vidrio es esencial para las industrias de alimentos y bebidas, asegurando la limpieza y la sanitización de los envases. Asimismo, manteniendo la calidad del producto final. Sin embargo, este proceso consume grandes cantidades de agua, planteando desafíos significativos para la sostenibilidad y conservación de recursos hídricos. La propuesta del sistema diseñado y fabricado que se presenta en este trabajo, es completamente automático y es capaz de manipular variables, como; velocidad de motores, tiempo de cepillado y enjuague. Esta línea de lavado, actualmente procesa 720 botellas por hora, reduciendo significativamente el consumo de agua, en comparación con métodos tradicionales. Así como, se aplicaron tecnologías avanzadas de automatización y programación, con el objetivo principal de mejorar la eficiencia y sostenibilidad del proceso.
The article discusses the issues of malnutrition and food insecurity worldwide, high lighting the significant problem of food waste. We propose a technological solution to address this specific issue. The use of computer vision, particularly Deep Learning (DL) models, is proposed as a means to identify the maturity state of food, thereby deter mining its consumability and preventing premature waste. This manuscript explores various applications of deep learning in topics related to the food industry, focusing on research aimed at mitigating food waste. Three datasets were constructed to identify the most commonly wasted foods, as well as to determine the maturity levels of these foods. A deep learning model was developed to identify three different food items with an effectiveness rate of 77%. Additionally, three deep learning models were created to assess the maturity degree of each food, achieving an average effectiveness of 70%. These results are comparable to similar studies that have employed deep learning models. This research shows the feasibility of using deep learning models to determine the maturity levels of three types of food that are frequently wasted in México. This approach is novel in research areas dedicated to developing methods, techniques, or technologies for avoiding or reducing food waste.
