ABSTRACT. This paper presents a hybrid methodology for harmonic state estimation in electric power systems, combining time-domain and frequency-domain analysis. The proposed approach is based on the use of the Fast Fourier Transform (FFT) to extract harmonic content from waveform measurements, and the application of the Kalman filter to improve estimation accuracy in the presence of uncertainty. The methodology is validated through simulations on a 5-bus test system, demonstrating that with only 11 measurements, a 90% error reduction can be achieved. The results show that this technique is efficient, accurate, and suitable for real-time applications, especially in networks affected by nonlinear loads.

ABSTRACT. This work proposes a data-driven model predictive control (MPC) strategy grounded in the Koopman operator framework for thermal regulation on the Temperature Control Lab (TCLab) platform. The Koopman-based model is identified from closed-loop data that include both forced and free responses. System stability is ensured by estimating the Koopman eigenvalues through constrained optimization, while a control-affine structure captures the influence of system inputs on the dynamics. To enable integration with a linear observer and an MPC, the complex-valued lifted model is converted to a real-valued form. Both the controller and observer are tuned using Bayesian optimization to minimize the Integral of Time-weighted Absolute Error (ITAE). Experimental results demonstrate that the Koopman-MPC approach provides reliable reference tracking in real-world conditions, despite the presence of process noise and ambient temperature fluctuations.

ABSTRACT. This article presents an approach for validating protection settings in an electrical system using COMTRADE files and experimental tests with a CMC356 injector. The IEEE 14-bus model was used, selecting three buses to conduct short circuit studies and adjust intelligent electronic devices associated with transmission lines. The study included configuring current and potential transformers to ensure proper sizing, as well as adjusting overcurrent protection settings using functions 50, 50N, 51, and 51N. Operating currents and trip times were calculated based on the system’s characteristics. COMTRADE files, generated through simulations in PowerFactory, were injected into the intelligent electronic devices using the CMC356 injector and Test Universe software. The recorded trip times were then compared with static and dynamic simulations. The results showed acceptable error margins according to the IEEE 3004.1-2013 standard and confirmed that COMTRADE files are effective tools for validating protection settings by generating representative system signals.

ABSTRACT. Abstract— Optimal Power Flows (OPF) are an optimization tool used to analyze the operation of a Power Electric System (PES). The solution obtained provides the scheduling and dispatch of generating units, ensuring the conditions for proper energy management. The use of OPF is increasingly important due to its ability to address various scenarios in the current operational environment of a PES. This problem involves the optimization of an objective function that can take various forms, while satisfying a set of operational and physical constraints of the PES, such as the transmission network and the generators connected to it. This work presents the formulation, simulation, and results of two dynamic OPF methodologies. The first methodology consists of a conventional AC-OPF model using Nonlinear Programming (NLP). The second methodology is a DC-OPF approach that linearizes transmission losses using a piecewise linear model and linearizes the cost curves of thermal generators, formulated through Mixed-Integer Programming (MIP). Both methodologies are validated using a case study with a one-day-ahead horizon, considering different levels of penetration of variable wind and solar generation. The objective is to minimize generation costs, which in turn maximizes renewable generation, promoting the most efficient use of energy through proper dispatch of thermal, solar, and wind units. The formulation is tested on the IEEE 30-bus system, and both methodologies are compared, yielding an optimal solution for all proposed scenarios.

ABSTRACT. The increasing global integration of Electric Vehicles (EVs) in recent decades has significantly raised energy demand in Electric Distribution Systems (EDS), as these systems are primarily responsible for supplying power during EV charging. This growing demand can pose a risk to the operational limits of the electrical grid, affecting parameters such as power losses, voltage fluctuations, and ultimately the continuity of electric power service, impacting not only EV users but also other nearby consumers. This study evaluates the impact of various EV charging profiles on conduction losses and voltage profiles. The analysis places particular focus on the most electrically distant point in the network, where such effects are typically more pronounced. The evaluations are conducted using the quasi-dynamic simulation tool available in the software DIgSILENT PowerFactory, applied to the IEEE 13- and 37-node test feeders.

ABSTRACT. Space vector modulation (SVM) techniques are widely used to transfer energy from a DC source to an AC load in low, mid and high power. In mid and high power three-phase active neutral point clamped (ANPC) inverter is often preferred which is mainly controlled with SVM methods. In this paper, an analysis of three conventional SVM techniques for the threephase ANPC inverter is carried out. The Nearest-Three Vectors (NTV), Two Medium Vectors (M2V) and Three-Medium Vectors (M3V) are the techniques implemented by numerical simulations and validated using an emulation system Typhoon Hardware In the Loop (HIL). Common Mode Voltage (CMV), Common Mode Current (CMC), THDi and THDv are the parameters measured in this study. The results demonstrate that the NTV-SVM presents high CMC while M2V and M3V presents low CMC

ABSTRACT. This paper presents a real-time simulation of a three-phase boost rectifier with an LCL filter. The grid current is controlled using state feedback with integral action, tuned by the Linear Quadratic Regulator (LQR) criterion, while the DC bus voltage is regulated by a conventional Proportional-Integral (PI) controller. The study includes the mathematical modeling of the LCL filter, considering leakage resistances, and derives a state-space model in the synchronous reference frame. Additionally, the DC bus is modeled as a transfer function. To incorporate integral action, an augmented state-space system is formulated, and controller gains are computed using MATLAB’s LQR function. For the DC bus voltage controller, gains are determined based on the Nyquist stability criterion. Real-time simulations are conducted on the OPAT-RT platform, demonstrating effective DC bus voltage regulation. This study highlights the advantages and limitations of the LQR approach: it achieves high performance in managing Multi-Input Multi-Output (MIMO) highly coupled systems with low overshoot and fast response; however, without an observer, LQR requires numerous sensors, and its tuning process can be complex and time-consuming. Finally, future research directions are proposed, including the integration of state observers and refinement of tuning methods.

ABSTRACT. This paper presents the design of a three-stage Solid-State Transformer (SST) with a photovoltaic (PV) array connected at the low-voltage DC (LVDC) link and the evaluation of its performance. The PV system supplies a fraction of the load demand, while the rest is drawn from the medium-voltage AC (MVAC) grid. Grid-forming strategies—droop control and Virtual Synchronous Machine (VSM)—are implemented in the rectifier and inverter to enhance system flexibility. The SST is evaluated under MVAC voltage sags, PV generation loss, and abrupt load changes to compare the control responses. The simulation results show that both methods maintain system stability. VSM offers smoother dynamic behavior during voltage sags, while droop control supports modular operation by allowing independent stage control. In contrast, VSM requires synchronization between stages, which may reduce modular flexibility.

ABSTRACT. This paper presents the modeling, digital control design and emulation of a Three-Phase Active-Front-End Rectifier. The proposed system employs a Voltage-Oriented Control (VOC) strategy with synchronized Sinusoidal Pulse Width Modulation (SPWM), where the phase angle is obtained through a Synchronous Reference Frame Phase-Locked Loop (SRF-PLL). As part of the control design, a systematic tuning methodology is proposed for selecting the parameters of the current and voltage loops in the frequency domain. The control algorithm is implemented on a TMS320F28335 DSP and validated through PSIM simulations and Hardware-In-the-Loop (HIL) emulation using the Typhoon HIL 402 platform, demonstrating effective regulation of the DC link voltage and suitability for grid-connected energy conversion applications.

ABSTRACT. This paper presents the development and validation of a control strategy for converters in an alternating current (AC) microgrid, focusing on the reconfiguration capability between grid-forming (GFR) and grid-following (GFD) converters. A cascaded control scheme is proposed for the GFR and a current control loop for the GFD, enabling stable operation under load variations and compatibility with nonlinear loads. The microgrid is modeled using averaged equations that describe the system’s electrical behavior. Simulation results demonstrate that, in the event of a GFR failure, the GFD can dynamically reconfigure itself as a GFR, ensuring uninterrupted operation. This functionality enhances the system's flexibility, resilience, and autonomy, particularly in distributed generation scenarios powered by renewable energy sources such as photovoltaic systems.

ABSTRACT. Active sensors such as LiDAR offer accurate 3D reconstructions, but their cost and complexity limit their adoption in low-budget mobile robots. This paper presents a modular navigation pipeline based solely on a monocular RGB image, without requiring SLAM or additional sensors. The system combines metric depth inference using UniDepth, structured point cloud filtering, occupancy map generation, and trajectory planning with Artificial Potential Fields. The generated trajectory is converted to physical commands executable by an autonomous mobile robot. The method was quantitatively validated by evaluating depth accuracy, computational performance and reconstruction of real trajectories from encoder data. The results show consistent and safe trajectories in real scenarios, highlighting the method’s potential for autonomous navigation tasks.

ABSTRACT. Detecting anomalies caused by potential cyberattacks through network traffic analysis is a challenging task, especially in the Internet of Medical Things (IoMT) environments, due to their heterogeneous nature and wide range of vulnerabilities. Consequently, unsupervised ML models have been explored for anomaly detection. However, their dependence on manually hand-crafted network traffic features limits their ability to effectively detect attack patterns that may be overlooked during feature engineering. Therefore, this study introduces a novel method for characterizing network traffic flows through embeddings generated by a Transformer-based model, and subsequently proposes an unsupervised anomaly detection based on Isolation Forest that operates directly on these learned embeddings, effectively mitigating the need for manual feature engineering. According to the experiments, the method obtained competitive results regarding to state of the art proposals.

ABSTRACT. The confidentiality and integrity of sensitive data are increasingly critical due to the growing number of transactions conducted over the Internet using Internet of Things (IoT) devices. Ensuring these security properties requires the implementation of robust technical controls. Cryptography is one such control, providing techniques to transform sensitive data into encrypted information, thereby forcing potential attackers to either break the algorithm or perform exhaustive brute-force searches. This paper analyzes and proposes the use of random combinations of classical encryption techniques—specifically Vigenère, columnar transposition, and reverse ciphers—for IoT devices. To evaluate the effectiveness of this random encryption method, we applied three-step, five-step, and ten-step encryption sequences and compared the results with those obtained using a 256-bit substitution-permutation network cipher. The ten-step configuration produced the most uniform distribution, effectively eliminating repeated patterns while requiring less than one-tenth the processing time of the substitution-permutation network.

ABSTRACT. Data collection in sensory analysis research has traditionally relied on manual paper-based methods, which introduces human error and limits the scalability of studies. Existing software solutions often lack the flexibility and modularity necessary to adapt to different products or evaluation protocols without incurring complex custom development. This article presents the design, development, and implementation of a native Android mobile application, developed as a component within a comprehensive research information system, applying fundamental software engineering principles such as modular design, component reuse, service-oriented architecture, and user-centered development. The main innovation lies in an architecture that allows for the dynamic generation of sensory evaluation forms, configurable from a web platform, enabling their use in multiple contexts without changes to the application source code. The solution was validated through a usability evaluation in a real-world wine tasting environment, with the participation of 35 panelists. The results, obtained through a 16-question questionnaire on a 7-point scale, reflect high levels of satisfaction, ease of use, and a rapid learning curve (overall average of 5.96 out of 7). The proposal constitutes a significant contribution to the field of Software Engineering applied to scientific research, providing a robust, scalable, and validated tool for the structured collection of sensory data.

ABSTRACT. The Prony method is a signal processing technique that is used to decompose a signal into a sum of damped complex exponentials. The key strength of the classical Prony method lies in converting a non-linear approximation problem into a linear one, which involves solving a system of linear equations and then a root finding problem for the resulting polynomial. Prony analysis is highly effective for estimating amplitude, damping,and frequency from data (sampling). However, it has limitations, including frequency resolution, noise sensitivity, and analytical complexity. This paper aims to propose a novel methodology thatintegrates genetic algorithms in the Prony method to improve signal reconstruction from noisy and sparsely sampled data using a simple mathematical analysis. The results reveal that the proposed method outperforms three traditional Prony-like methods, Least Squares (LS), Total-Least-Squares (TLS) and Matrix Pencil Method (MPM) in terms of accuracy and tolerance to noise.

ABSTRACT. Modern wind energy conversion systems (WECS) can emulate the inertial response of conventional generators by implementing additional control loops. Inertial control methods, such as Optimized Power Point Tracking (OPPT) and its variants, and Torque-Limited Inertia Control (TLIC), control the kinetic energy of the rotating masses in the wind turbine, allowing WECS to assist in frequency regulation in the event of an imbalance between the generated power and the power demanded by the grid-connected load at a given time. These inertia control methods modify the operating point of the wind turbine to provide frequency support without the need for energy storage systems. By adjusting the operating point, the wind turbine changes its rotational speed. Modifying the turbine operating point causes the blades to be subjected to forces that can reduce their useful life. This article analyzes wind turbine rotational speed time series obtained from simulations of contingency events in power systems with a high participation of wind generation. WECS implements inertial control systems based on OPPT, OPPT Extended, and TLIC. The power system simulation environment is MATLAB/Simulink. Finally, the possible impact on the forces acting on the wind turbine blades was evaluated based on the time series of the rotor speed obtained with and without these frequency control strategies using the transient structural analysis in ANSYS Workbench.

ABSTRACT. A computational model is developed to analyze the performance of sodium battery cells intended for microgrid applications. Using COMSOL Multiphysics, a cell composed of a hard carbon anode, a NaFeO₂ cathode, and a NaClO₄ electrolyte is studied. Four scenarios with different discharge rates (1, 5, 10, and 12 A/m²) are evaluated, observing how they influence the capacity and voltage of the cell. The results indicate that lower rates allow for greater energy efficiency, which is ideal for stationary slow-discharge applications such as microgrids. In addition, it is highlighted that efficient cell design and the correct selection of materials can improve the integration of intermittent renewable energies into these networks. The study concludes that sodium batteries are a competitive option compared to lithium, due to their cost, availability of materials, and adaptability to variable thermal conditions present in real environments.

ABSTRACT. This paper presents an extended harmonic domain (EHD) modeling of a photovoltaic (PV) generation system. The system consists of a PV module model, a Boost converter, a three-phase inverter, a three-phase transformer, and a grid equivalent. The obtained model is validated through simulations using its corresponding time-domain model. The EHD model is then used to develop a methodology based on a non-linear least-squares method to estimate the parameters of the three-phase inverter. The goal is to improve the DC bus voltage and enhance the dynamics of the power electronic converters used. Subsequently, this methodology is applied to calculate key parameters of the solar module, such as its resistances, using its non-linear model, the maximum power point, and another operating point. The results indicate that the proposed methodology enables the estimation and optimization of system component parameters to improve the system’s dynamic response. Furthermore, it allows the calculation of other relevant parameters, such as the resistances of the PV model, which are necessary to obtain its characteristic curves.

ABSTRACT. This study presents a systematic review and detailed probabilistic analysis of hourly wind speed and solar irradiance data for the La Rumorosa, Baja California, Mexico. Monthly wind speed time series were evaluated over a 10-year period, comprising 87,360 hourly data points, to identify the most suitable probability distribution function (PDF) for modeling the data distribution. The Weibull distribution is widely accepted by the scientific community and recommended by International Electrotechnical Commission (IEC 61400-12-1) for wind resource assessments due to its ability to accurately represent wind speed variability and frequency but other PDFs should be explored. Therefore, statistical characteristics were analyzed using historical wind data from NASA POWER and the most commonly adopted PDFs in the literature, including log-normal, Weibull, Gamma, Rayleigh, and Normal distributions, enabling comparative performance assessment. Once the most appropriate distribution for wind speed data was identified, wind energy potential can be reliably estimated. To validate the model fits, the Kolmogorov-Smirnov (K-S) goodness-of-fit test was applied, and statistical metrics including root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and coefficient of determination (R²) were calculated. Results indicated that the Weibull distribution provided the best fit for most months (RMSE: 0.0069, MAE: 0.0042, R²: 0.984, K-S: 0.027), but the Normal distribution achieved superior performance in certain months (RMSE: 0.0035, MAE: 0.0025, R²: 0.996, K-S: 0.010), offering optimal representation of wind regimes based on seasonal variability. For solar irradiance analysis, 51,100 data points were examined, corresponding to 14 daily hours of available solar radiation, with the summer solstice used to define the range of useful solar hours. The same unimodal probability distributions applied to wind data were initially evaluated for solar irradiance modeling. Subsequently, bimodal analysis was implemented to capture intraday variability and mixed atmospheric conditions (partially cloudy and clear-sky days). The bimodal mixture approach significantly improved model fit, enabling more accurate representation of hourly solar irradiance distributions. The innovation of this work lies in developing a robust methodological framework for precise sizing of hybrid wind-solar systems, specifically designed to mitigate intermittency issues inherent in renewable energy sources. The results provide fundamental statistical tools for probabilistic characterization of renewable resources in La Rumorosa, Baja California, with direct applications in designing operational strategies, energy storage systems, and hybrid system optimization to ensure enhanced reliability in energy supply.

ABSTRACT. In recent years, the progressive removal of large synchronous generators and the rapid expansion of inverter-based renewable resources have reduced system inertia to historically low levels, degrading its ability to damp frequency disturbances and prolonging recovery after faults or abrupt load changes. Against this backdrop, a low-inertia nine-bus network —entirely supplied by inverter-based resources— is analyzed in PSCAD under two operating paradigms: Grid-Following (GFL) and Grid-Forming (GFM). Three GFM schemes —Droop, Power Synchronization Control (PSC) and Virtual Synchronous Machine (VISMA)— are evaluated through a step load increase contingency. Frequency nadir, zenith and rate of change of frequency (ROCOF) are quantified, along with system recovery time and oscillation damping. While GFL inverters provide a rapid active-power response, they exhibit under-damping and pronounced frequency oscillations at high penetration levels. In contrast, GFM configurations maintain nominal frequency more robustly across all scenarios, with VISMA delivering the greatest transient stability by minimizing nadir deviations and oscillatory behavior via synthetic inertia emulation. These findings inform grid-code specifications and operational guidelines for secure integration of high shares of inverter-based renewables in low-inertia power systems.

ABSTRACT. Multi-focus image fusion has been extensively explored to extend the depth of field in conventional optical systems, addressing the loss of information in blurred regions. In this work, we present a deep learning-based approach that repurposes the MobileNetV2 architecture by removing its classification head and extending it symmetrically with a decoder for image reconstruction. The encoder is used as a frozen feature extractor, while the decoder incorporates transposed convolutions and skip connections to recover fine spatial details. Despite being trained for only 50 epochs on a custom multi-focus dataset, the network achieves promising fusion results, demonstrating a favorable trade-off between output quality and architectural simplicity compared to more complex state-of-the-art methods.

ABSTRACT. This work presents the implementation of EfficientNet-B2, a pretrained convolutional neural network model based on the ImageNet dataset, using transfer learning to classify fundus photographs. A dataset composed of five public databases focused on glaucoma detection was used. The proposed methodology was structured into four main stages: (1) selection and integration of the dataset, (2) data analysis and preprocessing, (3) selection, adaptation, training and validation of the pretrained model, and (4) evaluation of the obtained results. During preprocessing, techniques such as normalization and one-hot encoding were applied to properly prepare the input data. The results include performance metrics and confusion matrix, allowing for tan assessment of the architecture's effectiveness and the formulation of strategies to improve classification efficiency.

ABSTRACT. Vibration regulation is essential to improve the accuracy and prolong the lifetime of mechanical components. In this sense, jerk plays a key role due to its direct association with the system’s residual vibrations. This work presents an efficient methodology for the real-time detection of jerk peaks from the analysis of vibration signals obtained from a triaxial accelerometer. Unlike previous approaches that require complex dynamic models, this study introduces a scheme based on the numerical derivative of acceleration and an adaptive threshold computed using moving statistics. This accurate approach allows for precisely identifying relevant events, even in noise. The proposed methodology was experimentally validated using trapezoidal profiles in velocity designed to present jerk pulses at known time instants. The results of these experiments demonstrate a high temporal correspondence between the theoretical pulses and the detected peaks, thereby validating the approach as a valuable tool for experimental analysis of motion profiles.

ABSTRACT. Low-cost robotic arms are increasingly used in educational environments to teach motion planning and kinematics, but often lack safe and intuitive controllers tailored to novice users. This study presents the design and validation of a custom servomotor control platform based on an STM32 microcontroller, intended to simplify user interaction and reduce hardware-related risks in collaborative robotics. The platform includes a four-layer PCB with eight optically isolated PWM outputs, a USB-UART interface, and firmware capable of storing up to 64 motion trajectories in internal flash memory. Commands are received as ASCII strings over a serial connection and executed in real time. A graphical user interface allows users to control each joint from 0° to 180° using sliders and buttons, abstracting low-level signal handling. The system was validated on a six-degree-of-freedom robotic arm, demonstrating accurate angular control, reliable trajectory playback, and robust USB communication without data loss. Tests confirmed the utility of software-defined angular limits for protecting hardware and improving setup time. The proposed platform supports hands-on learning by prioritizing ease of use, safety, and modularity, and is suitable for integration into remote labs and future human–robot interaction research.

ABSTRACT. This work presents a rigorous comparison between Python and C for real-time signal processing on the BeagleBone AI-64, using MATLAB executed on a PC as a precision reference. The Discrete Wavelet Transform (DWT) with five levels and Daubechies 4 (db4) was implemented in both languages, running directly on the embedded BeagleBone AI-64 platform to process signals with disturbances. Precision was evaluated through the Mean Absolute Percentage Error (MAPE), and computational efficiency was assessed via real-time processing times per sample. Results show that Python and C achieve practically identical precision with MAPE errors below 0.008%, but C processes each sample in less than 100 microseconds, meeting strict real-time requirements. Python, however, is only suitable for less demanding scenarios or offline processing due to its longer execution times. This study provides key evidence for language selection in critical embedded systems, emphasizing the importance of validating precision and efficiency directly on the target hardware. It also serves as an experimental reference for future developments in power quality monitoring and control on advanced embedded platforms.

ABSTRACT. The synthesis of realistic electromyographic (EMG) signals has become a key challenge in the development of robust machine learning models for biomedical applications, particularly in scenarios where collecting large-scale real datasets is difficult or costly. In this study, a Deep Convolutional Generative Adversarial Network (DCGAN) architecture was adapted to generate synthetic one-dimensional EMG signals corresponding to hand opening and closing movements. The generator model transforms latent noise vectors into EMG-like waveforms, while the discriminator employs a parallel multi-domain strategy analyzing time, frequency, envelope, and time-frequency representations, as well as leveraging a minibatch discrimination mechanism to enhance diversity. The quality and fidelity of the generated signals were systematically evaluated using statistical metrics (root mean square, absolute mean, variance, skewness, kurtosis), amplitude distribution, autocorrelation, and frequency domain analysis via FFT. Experimental results demonstrate that the synthetic EMG signals exhibit high similarity to real data across multiple domains, supporting their potential use for data augmentation and algorithm validation in EMG-based biomedical research.

ABSTRACT. The integration of intelligent systems for continuous process monitoring is a key objective of Industry 4.0. Induction motors (IMs), which account for over 50% of global energy consumption, are critical in industrial environments, making their condition monitoring essential to prevent downtime, reduce unnecessary maintenance, and improve energy efficiency. Bearing failures are the primary cause of IM malfunctions; however, most existing studies focus solely on metallic bearings. This work proposes a methodology for detecting outer race faults in metallic, ceramic, and hybrid bearings. The approach involves extracting statistical time-domain features from both vibration and stator current signals, followed by dimensionality reduction using Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Fault classification is then performed using a neural network-based classifier to identify the presence of outer race faults across different bearing types. The proposed method is validated using an experimental dataset, and its performance is compared with that of conventional techniques.

ABSTRACT. This work proposes a method for detecting faults in induction motors using the Hilbert Transform (HT), a technique that enables precise analysis of the variation in frequency and amplitude of electrical signals in the time domain. Through this approach, defects can be identified by analyzing the harmonic components of current signals. The study also presents an algorithm for estimating the Fourier coefficients of these signals, providing an efficient tool for fault diagnosis without the need for large volumes of data. The results show that the use of the HT improves the accuracy and speed of diagnostics, which could benefit predictive maintenance of industrial motors by enabling earlier fault detection. The coefficient algorithm performs a transformation from the time domain to the frequency domain by means of the Fourier coefficients in complex form. The result is two sets of coefficients, convoluted with the values of the positive and negative frequency components, applying the duality property. The results obtained with this algorithm allow comparison with the method based on the discrete HT, considering both the real and imaginary parts. The application of this algorithm enables the diagnosis of the health status of electric motors, providing information in terms of spectral density and harmonic frequencies associated with faults, within a detectable range from 40 dB to 10 dB.

ABSTRACT. Inter-turn short-circuit (ITSC) faults in induction motors represent one of the most common and critical failure modes in industrial environments. If not detected at an early stage, these faults can escalate rapidly, resulting in severe equipment damage, costly downtime, and significant production losses. Therefore, accurate and early detection of ITSC faults is essential to improve system reliability and ensure operational continuity. This work proposes a methodology for classifying ITSC faults using stator current signals, which combines signal envelope analysis, multiscale signal decomposition, and deep learning. Initially, amplitude variations are extracted from the raw current signal via a sliding window integral, effectively isolating modulated components associated with fault-induced sidebands. The resulting signal is then processed using a multilevel Discrete Wavelet Transform (DWT). The extracted features are normalized and fed into a custom one-dimensional Convolutional Neural Network (1D-CNN) designed to learn fault-specific patterns from time-series data. The methodology was validated under six fault severities and varying mechanical load conditions to explore robustness. The proposed approach achieved accuracy rates up to 99.41% in five-fold cross-validation, demonstrating generalization performance.

ABSTRACT. Induction motors are present in industry due to their low cost and robustness. Fault detection on induction motors has been studied using various motor signals, including current, vibration, and stray flux monitoring. Another noninvasive technique currently used is thermography analysis. Thermography has gained importance in fault detection for induction motors due to the convenience of taking a thermal image and visualizing the temperature representation of the motor. This has enabled the development of various methods for image analysis, with a focus on fault detection. This work presents a case study on the detection of induction motor bearing damage using thermographic images and the 2D Fast Fourier Transform. The thermographic images were taken from a healthy motor and two damaged conditions (bent, cross) on the front bearing. The obtained results indicate that the 2D-FFT is a viable option for detecting bearing damage.

ABSTRACT. This article presents a model for determining the loadability capacity of electrical systems to analyze reactive power compensation while considering alternating current power flow constraints. The proposed approach calculates the maximum load coefficient that the network can handle, explicitly incorporating reactive power compensation within the reactive power balance constraint. The model, which is nonlinear, is applied to IEEE test systems with 9, 14, 30, and 39 buses, demonstrating that operating under maximum demand conditions can significantly impact the system's voltage profiles. The article emphasizes the importance of maintaining voltages within established operating limits to address this issue, ensuring the power network's safe, reliable, and efficient operation.

ABSTRACT. The control centers at the Federal Electricity Commission have the task of repairing and maintaining the electric power distribution system in a functional state. Considering how power outages affect consumers, it’s essential to analyze factors such as SAIDI, SAIFI and CAIDI, and have a better understanding of the effectiveness of each maneuver done to restore power distribution. Dashboards were created using Excel and PowerBI to display index values and other data in an intuitive design. Each dashboard is uploaded to a main web page named “Contenedor de Distribución” (SDD) found in a local server, and this web page was designed and organized in a way that each user could navigate it with ease.

ABSTRACT. This article addresses one of the most important problems faced in modern electrical distribution systems: The optimal reconfiguration of primary feeders that contain distributed generation systems and Reactive compensation systems. Both aspects require optimal strategies for opening and closing switches, as well as adjusting reactive compensation. The problem is solved by using efficient power flow techniques for radial systems that allow calculating different operating scenarios with the aim of minimizing the number of switch operations and also establishing the optimal level of reactive compensation through linear sensitivities. The proposed algorithm takes into account the hourly variation of the load and thus the variation in renewable energy sources. Case studies are presented.

ABSTRACT. This paper presents the implementation of models to represent the generator/converter interface and electrical controls of Inverter-Based Renewable energy sources (IBR), corresponding to the second generation of generic models proposed by the Western Electricity Coordinating Council (WECC) in a time-domain simulation program. A model suitable for representing variable-speed type IV-B wind turbines (WTG) and solar photovoltaic (PV) power plants is described, and included in TRANSTAB, an in-house program for electrical system stability studies developed by the Research Group on Dynamic Phenomena of Electrical Machines and Interconnected Networks of the Graduate Program in Electrical Engineering at SEPI-ESIME-Zacatenco. The implementation of the models was successfully validated using a test protocol developed by EPRI for this purpose, comparing the results of TRANSTAB simulations with the ones of two popular commercial simulation programs.

ABSTRACT. This paper analyzes the basics of the Dot Product criteria, and proposes some improvements to correctly use it, when analyzing power systems with detailed modeling, including an Inverter Based Renewable energy resource (IBR) plant. It verifies the performance of the dot product criteria in the original IEEE 9-bus, 3-machine test power system, and in the same system, when generator at bus 3 is replaced by an IBR power plant of an equivalent capacity. This has a twofold objective: verifying the effect of an IBR in the stability assessment of electric power systems, and explaining the performance of the dot product with the Equal Area Criterion (EAC). Results show that the IBR power plant affects, indeed, system stability, but it does not interfere with stability assessment, while the point at which dot product detects instability, in which the dynamic trajectory of the system crosses the potential energy boundary surface of the unstable cases, is precisely the same at which the EAC detects instability.

ABSTRACT. Electromagnetic energy represents an alternative to diagnose, prevent or discover some human disease through the microwave medical imaging devices, a non-invasive monitoring device, which is applied to detect tumors, ischemic strokes or bone fractures healing, whose operating principle is based on the electromagnetic radiation through the damaged tissue or bone fracture, in this sense, a fundamental component of these medical devices is the antenna, which it should comply with a small size, gain, bandwidth and low cost, in this way, the objective of this work is the design and simulation of a microstrip array antennas, proposing its frequency band operation at 28 GHz Millimeter Wave Band, in conjunction with the beamforming technique.

ABSTRACT. In this paper, a comparison of two estimation models for the temperature variables of a distillation process is presented. A data set collected from a distillation process t0 . . . t6; the data t0 . . . t5 from the process is normalized and used as input to a model to estimate t6 and compare the result with a model that uses the data using Principal Component Analysis (PCA). The number of the selected principal components determines the number of input values for the estimation model. The Root Mean Square Error (RMSE) is used as validation metric, and a graphical and quantitative comparison of the results is performed.

ABSTRACT. This paper presents a comprehensive theoretical and experimental analysis of ADC selection for harmonic analysis applications, comparing successive approximation (SAR), sigma delta, and integrating architectures. Communication protocols (SPI, I2C, UART) are evaluated, and the noise behavior in high resolution sigma-delta converters, specifically the ADS1263, is characterized. Theoretical specifications are validated through experimental voltage measurements, showing increased noise during active conversion (0.16 Vpp vs. 0.018 Vpp in the STM32 successive approximation ADC). Design recommendations inclu de proper decoupling, ground plane implementation, and filter configuration to mitigate noise in precision measurement systems.

ABSTRACT. This study presents a strategy based on optical simulation to evaluate the photonic performance and energy efficiency of multispectral LED lighting systems applied in vertical farming environments. The analytical model was based on a hydroponic rack-type system, in which critical physical parameters such as luminaire mounting height, cultivation area dimensions, and effective radiation coverage were defined. Three distinct spatial configurations, designated as SD1, SD2, and SD3, were developed to analyze the influence of distribution patterns on irradiance uniformity. The light sources considered corresponded to Samsung LH351H series LEDs: the blue emitter, with a central wavelength of 450 nm, radiant flux of 750 mW, and quantum efficacy of 2.86 μmol J⁻¹, and the deep-red emitter, with a wavelength of 660 nm, radiant flux of 1060 mW, and efficacy of 4.04 μmol J⁻¹. The emitters were modeled in the TracePro® simulation environment as Lambertian sources with a conical angular distribution, using the manufacturer’s specifications. The red/blue spectral ratio was fixed at 3:1, following guidelines reported in the scientific literature to enhance photosynthesis under controlled conditions. Each scenario was simulated using ray-tracing techniques with a grid of more than 16,000 sampling points, enabling the determination of the spatial distribution of photosynthetic photon flux density (PPFD), photonic uniformity, and angular variability within the cultivation area. All simulations were conducted under equivalent energy conditions, with constant input power maintained to ensure an objective technical comparison across designs.

ABSTRACT. This paper presents a machine learning approach for the automatic classification of synthetic Jovian radio emissions using spectrogram analysis. As a continuation of previous work involving the design of a dipole antenna and the development of a Python-based acquisition system centered at 20.1 MHz, this study focuses on generating a synthetic dataset that replicates the Io-A, Io-B, and Io-C radio emissions of Jupiter. Each class was modeled using Python scripts that incorporated variability in position, intensity, shape, and noise within the 16–24 MHz frequency range. Two classification models were trained and evaluated: a convolutional neural network (EfficientNet-B0) and a vision transformer (ViT Tiny Patch16-224). Both models were tested using a 70/15/15 train-validation-test split and validated with 5-fold cross-validation. The ViT model achieved up to 100% classification accuracy across all folds, outperforming the CNN in consistency and generalization. This synthetic classification framework serves as a preparatory phase for future work involving the training and validation of these models using real labeled spectrograms extracted from the Juno/Waves database, which includes emission records from 2016 to 2023.

ABSTRACT. Evaporation is a key process in the hydrological cycle and plays a crucial role in water management, ecosystem development, and conservation. In recent decades, Artificial Intelligence (AI) has been proposed as an alternative to direct measurements, such as the Class A pan evaporimeter method for estimating the evaporation rate. This study compares three AI models —K-Nearest Neighbors (KNN), Random Rorest (RF), and Multiple Linear Regression (MLR)— to estimate the monthly pan evaporation rate in a semiarid region. The results show that KNN is the most accurate model, with a mean absolute error of 21.68 mm/month and a Pearson correlation coefficient of 0.89, outperforming the MLR and RF methods.

ABSTRACT. This study presents the design, implementation, and evaluation of an autonomous urban traffic monitoring system that integrates artificial vision with crowdsourced data. The system was deployed in the municipality of Corregidora, Querétaro, Mexico, to support mobility research in intermediate cities. High-definition cameras were installed at strategic urban locations to capture continuous video streams processed by real-time computer vision algorithms for vehicle detection, speed estimation, lane change recognition, and traffic density calculation. In parallel, real-time alerts from the Waze platform were incorporated to enhance situational awareness and validate visual observations. The resulting multimodal dataset was analyzed using time series modeling, multivariate statistical analysis, and geospatial visualization.

The system achieved high accuracy in vehicle detection (96.2\% daytime, 91.8\% nighttime), reliable lane change identification (88.6\% true positive rate), and robust speed estimation (mean absolute error of 5.2\%) across various traffic scenarios. The integration of vision-based indicators and user-generated alerts improved incident classification and enabled the identification of critical congestion segments. These results validate the effectiveness of the system as a scalable solution for data-driven traffic monitoring and predictive mobility planning in real-world urban environments.

ABSTRACT. This work presents a non-contact methodology for estimating oxide scale thickness on AISI 1045 steel during high-temperature oxidation. Radiometric infrared thermography was used to capture the thermal evolution of steel specimens subjected to a controlled Joule heating process. Each thermal image was flattened and processed via Principal Component Analysis (PCA) to extract low-dimensional thermal features. These features were used to train machine learning regressors, specifically Random Forest and XGBoost, to predict oxiderelated dimensional changes. The methodology was validated using micrometer-based thickness measurements synchronized with the thermal data. Results demonstrated strong predictive performance with mean absolute errors below 0.007 mm. This approach enables real-time tracking of thermally induced surface evolution without invasive sensors, opening possibilities for smart manufacturing applications.

ABSTRACT. A mechanical deformation refers to the change on a material when this is subjected to external force. Commonly, mechanical sensors as strain gauges, load cells, etc., can be applied to measure these deformations, converting mechanical energy into an electrical signal. Another device that capable of measuring mechanical deformations is the optical interferometer. Specifically, a shearing interferometer can generate interference patterns by comparing an unperturbed wave (sample at rest) with a deformed wave (sample under stress). In this study, through optical interferometry, the deformations suffered by an electrical insulator of ceramics when it is subjected to mechanical forces (compression 250kg) will be analyzed. Applying this force it will be able to see the deformation of this ceramic. The optical test implemented using a shearing interferometer provides reliable results for detecting changes under compression. However, future work could improve these results by optimizing the optical configuration, enhancing interferogram quality, improving mechanical stabilization, and performing quantitative validation through comparison with alternative types of mechanical sensors.

ABSTRACT. The Breast Cancer Wisconsin (Diagnostic) Data Set (WDBC) has been widely used for breast cancer classification, employing machine learning techniques ranging from linear models to neural networks. However, the actual need for complex models on this dataset remains an open question. In this work, a comparative analysis is performed between classical models (Logistic Regression, SVM, Random Forest) and more complex models (MLP), evaluating accuracy, training times, and interpretability. The results show that, in structured problems such as WDBC, classical models combined with dimensionality reduction can match the performance of complex architectures while offering lower computational costs.

ABSTRACT. Este artículo presenta una evaluación comparativa del desempeño de distintas configuraciones en serie de reactores de síntesis para la producción de metanol, tomando como base la tecnología licenciada por Lurgi. Si bien esta tecnología es ampliamente utilizada en la industria, su diseño tradicional con un único reactor puede estar limitado por restricciones termodinámicas que reducen la conversión de los reactivos.

En este estudio se modelaron y simularon configuraciones con uno, dos y tres reactores empleando el entorno AVEVA Process Simulation, utilizando como alimentación la composición típica del gas natural proveniente del yacimiento de Camisea (Malvinas). El análisis se enfocó en parámetros clave como la conversión global de hidrógeno, la conversión de metano en el reformador, las emisiones de CO2, la relación entre la producción de metanol y el consumo de gas natural, así como la demanda energética del sistema.

Los resultados permitieron identificar las ventajas técnicas, energéticas y ambientales de cada configuración. Se destaca que la incorporación de un segundo reactor mejora significativamente la eficiencia y el desempeño global del proceso, mientras que la adición de un tercer reactor aporta beneficios marginales con un impacto operativo desfavorable.

Cabe señalar que esta evaluación se ha centrado exclusivamente en criterios técnicos, energéticos y ambientales relacionados con la conversión. No se ha considerado un análisis económico detallado, ya sea de inversión (CAPEX) o de operación (OPEX), por lo que las conclusiones deben entenderse como una aproximación conceptual que respalda la configuración de dos reactores como la opción más equilibrada y recomendable.

ABSTRACT. This article presents the feasibility, design, configuration, and dynamic and transient operation analysis of offshore wind farms for the efficient supply of electric energy. The proposed location is in region of Mxico, in the Gulf of Tehuantepec in Oaxaca, Mxico, which presents great potential and ideal characteristics for its installation and operation. This region offers significant advantages compared to others, according to data and studies provided by Mexican public sector institutions in charge of energy regulation and complementarity studies. It should be noted that this research represents an initiative for the installation of renewable energy generation systems in Mxico (wind power), taking advantage of the strong and constant wind currents in the area. The region is characterized by its favorable climate, its proximity to the ocean, and its strategic location for the installation of off-shore wind farms. This document analyzes the environmental and operational advantages, as well as the technical challenges, in the modeling and simulation of different case studies, using the PSCAD/EMTDC® simulator to verify its application.

ABSTRACT. The detection of structural damage in wind turbine blades is essential for ensuring operational reliability and preventing unexpected failures. This work presents a method based on the Markov Transition Field (MTF) and Convolutional Neural Networks (CNNs) for automatic classification of blade conditions using vibration signals. MTF transforms time-series data into two-dimensional representations that preserve temporal transition patterns, enabling CNNs to extract spatial features associated with structural degradation. The vibration signals were acquired from the Y-axis, which has shown greater sensitivity to damage-related information in rotating systems. Four severity levels were considered: healthy, light, intermediate, and severe. Analysis was conducted at 7 revolutions per second (rps), selected as an intermediate point between the start-up (3 rps) and maximum operational speed (12 rps), allowing for the capture of representative dynamics under partial load conditions. The proposed MTF-CNN model achieved 100% classification accuracy, demonstrating its effectiveness for identifying damage severity in wind turbine blades and its potential for integration into condition monitoring systems

ABSTRACT. Nowadays, there are various proposals to promote the use of alternative energies globally. In this regard, wind turbines present a prominent option because they are a clean and renewable energy source, which contributes to reducing environmental impact. Wind turbines are systems that convert the wind's kinetic energy into electrical energy; however, for efficient performance, the system must be in healthy condition. Therefore, it is important to have monitoring and diagnostic strategies in place to detect wind turbine failures. This paper presents a proposal for the characterization of blade fracture failures in low-power wind turbines. The proposal is based on the use of Hjorth’s parameters (HPs) to analyze triaxial vibration signals. A multivariate analysis of variance (MANOVA) and comparison of means are performed to statistically validate the HPs' ability to characterize the signals. An experimental validation of the proposal was performed with a low-power wind turbine in a controlled environment with a wind tunnel, where different levels of failure severity were analyzed (healthy, light, intermediate and severe). The results show the effectiveness of HPs as indicators capable of characterizing vibration signals according to the case studies considered.

ABSTRACT. Energy consumption in industrial machines has become a significant factor in the reduction of production costs. The design of trajectories and the manipulation of motion dynamics have proven to be a viable solution for reducing energy consumption in industrial machinery. In this study, an analysis of the current consumption of a servomotor utilizing two motion dynamics was conducted, which demonstrated a 30% reduction in current consumption of the profile with reduced jerk levels.

ABSTRACT. This paper presents the Hardware-in-the-Loop (HIL) implementation of the grid-side converter (GSC) in a wind energy conversion system. The implementation is performed in the OP5600 real-time simulator based on the RT-Lab/SimulinkTM simulation platform, and the TMS320F28335 Digital Signal Processor (DSP). The model includes the dynamics of the three-phase grid connection, RL filter, and DC-link capacitor. The real-time implementation and validation include a synchronous reference frame phase-locked loop (SRF-PLL), current and voltage loop PI regulation, and anti-windup techniques. The results demonstrate stable synchronization, effective DC-link voltage control, and protection against overcurrent during capacitor charging. The HIL approach allows for safe and accurate validation of control algorithms before deployment in physical systems.

ABSTRACT. This paper presents the implementation of the Hardware-In-the-Loop (HIL) real-time simulation technique applied to the traction system of an electric vehicle. The control algorithm is implemented in a Texas Instruments digital signal processor (DSP), which communicates with a real-time simulator from OPAL-RT through analog and digital channels. The controller is evaluated in a real-time simulation environment using a driving cycle composed of three stages: acceleration, constant speed, and deceleration. The signals analyzed include angular velocity, electromagnetic torque developed, and the three-phase stator currents of the machine.

ABSTRACT. The evolution of technological paradigms such as the Internet of Things (IoT) and data processing through Artificial Intelligence (AI) techniques has given rise to the concept of the Artificial Intelligence of Things (AIoT). This approach has contributed to the development of solutions in sectors such as agribusiness and packaging centers, where repetitive operational processes—such as object identification and counting—are common. This article proposes a scalable and accessible AIoT architecture, developed by integrating Transfer Learning techniques and implementing the Google Teachable Machine tool, for the automated identification and counting of objects passing in front of a camera. The system integrates a classification model, a webcam, and the MQTT protocol to display real-time data on the Ubidots platform. The result is an accurate and replicable system that can help improve operational efficiency and traceability in critical processes.