IEEE Latin America Transactions

An Effective Switching-Table-Based DTC of Five-Phase PMSM for Reduction of Torque Ripples

2026-06-11T01:58:48+00:00

Higher torque and flux ripples exist in conventional direct torque control (DTC) of five-phase permanent magnet synchronous motor (5-ϕ PMSM) drive due to application of single voltage vector over entire control period. Duty-cycle based DTC techniques are proposed for 5-ϕ PMSM recently to achieve superior steady-state responses. However, most of the duty cycle-based schemes are complex. To suppress the ripples in flux and torque, an effective switching table-operated DTC is proposed in which selection of active voltage vector (AVV) for independent torque and flux control is introduced for 5-ϕ PMSM. Secondly, the selected AVVs for torque and flux control are applied along with a null vector in two consecutive samples, respectively based on effective duty control scheme. The duty cycles of selected AVVs are determined to meet the reference flux and torques in respective control intervals, which ensures reduced flux and torque ripples. When voltage vector for torque control is applied, the effect on flux is negligible and vice versa, which achieves a decoupled flux and torque control. The proposed DTC is verified experimentally and compared with conventional DTC, conventional model predictive current control (MPCC) scheme, and a duty-cycle control based MPCC for 5-ϕ PMSM.

Modeling and Analysis of a Liquid-Cooled Heat Sink for Inverters Used in Hybrid Electric Vehicles

2026-07-06T09:01:40+00:00

This study presents the experimental characterization and dynamic thermal modeling of a liquid-cooled inverter heatsink for electric vehicle traction applications. The proposed approach considers the entire inverter housing as the heatsink and experimentally validates its thermal behavior under realistic operating conditions. Specifically, the heat transfer system was modeled using a Cauer equivalent circuit. The characterization yielded a thermal resistance of 16.75 m°C/W and a thermal capacitance of 6444.3 J/°C. The model validation was performed on a dynamometer bench with the inverter operating at an input power of 65.86 kW. Under these conditions, the system dissipated 2.34 kW of total power losses, resulting in average IGBT case temperatures of approximately 77 °C. The comparison between the experimental measurements and the model estimates, which predicted a case temperature of 80.48 °C, demonstrates the accuracy and reliability of the developed thermal model for this inverter-heatsink configuration.

Hardware in the Loop Marine Turbine Emulator

2026-06-23T16:33:27+00:00

Climate change and energy demand in coastal areas have strongly motivated ocean energy generation. Mexico has significant untapped marine potential, partly due to the high costs of accessing and commissioning on-site equipment. This paper accurately demonstrates a marine turbine emulator system in the laboratory, where the system ensures its performance by controlling the frequency and voltage in an induction machine so that the mechanical torque and speed of PMSG generator in the laboratory accurately reflects the real parameters of a hydro-generator installed at sea. This is accurately matched to a model of a Gorlov turbine operating in real time, with hardware control and inertia models of the mechanical system. Physical validation in three operating scenarios has ensured the integrity of the system under dynamic and stable conditions. This tool is undoubtedly the most innovative compared to the state of the art and is highly beneficial for control and mass integration studies of marine power generation in electricity grids.

EdgeFormerNet++: A Hybrid Attention-Guided Transformer Network for Accurate Retinal Layer Segmentation in OCT Images

2026-06-04T08:22:12+00:00

Accurate retinal layer segmentation in OCT is crucial for the early detection and monitoring of retinal and neurodegenerative diseases, yet it remains challenging due to the thin, low-contrast structures and closely packed layers in multi-class settings. We propose EdgeFormerNet++, a hybrid framework that integrates Res2Net-based multi-scale encoding, lightweight Transformer bottlenecks for global context, and dual attention (CBAM and scSE) for spatial-channel recalibration, complemented by an edge attention module to refine layer boundaries. The model is evaluated on two public datasets (NR206 and MGU) covering macular and peripapillary regions, achieving state-of-the-art Dice scores of 92.34% and 83.28%, respectively. Qualitative and quantitative results demonstrate accurate delineation of complex retinal structures, supporting the use of EdgeFormerNet++ for automated OCT analysis and computer-assisted diagnosis.

Imagined Speech in Spanish: EEG Dataset Acquisition Protocol and Baseline Classification Results

2026-07-08T20:51:20+00:00

This work presents a new Spanish-language electroencephalography (EEG) dataset for imagined speech, designed to support research in brain–computer interface (BCI) applications for assistive communication. A structured experimental protocol was developed to guide the acquisition process, incorporating auditory comprehension, imagined speech, and articulated speech production stages to enhance cognitive engagement and enable signal validation. The dataset includes 16 participants (9 male and 7 female), each performing 14 linguistic prompts consisting of nine words and five vowels. EEG signals were recorded using an open-source, low-cost acquisition system (OpenBCI Cyton + Daisy) with 16 channels configured according to the international 10–20 system. The collected signals were preprocessed, segmented, and evaluated through a deep learning classification framework adapted from recent imagined speech decoding approaches. Five classification experiments were conducted to assess the discriminability of the imagined speech signals. The results showed accuracies above the chance level across all experiments, achieving 30.79% ± 4.76 for vowel classification, 20.81% ± 3.11 for word classification, and up to 74.61% ± 7.11 for binary word–vowel discrimination. Comparisons with public datasets demonstrated that the proposed dataset achieves competitive or superior performance despite using low-cost hardware. The code used in this work is available at https://github.com/GracielaRamirezA/Imagined-Speech-in-Spanish.git.

Predicting Travel Intention Using Machine Learning and SHAP-Based Explainability: A Virtual Tourism Approach

2026-05-28T02:22:35+00:00

Virtual tourism is transforming the way users explore and evaluate travel destinations, yet accurately predicting the intention to visit after engaging in virtual experiences remains a challenge. Existing approaches often lack predictive accuracy and interpretability, limiting their application in tourism decision making. To address this, we develop a machine learning framework that integrates explainable artificial intelligence (XAI) to predict the intention to travel after experience with high accuracy and transparency. We implement eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine (LightGBM), and Multilayer Perceptron (MLP), applying cross-validation for robustness and Optuna-based Bayesian optimization to maximize model performance. To ensure interpretability, we employ SHapley Additive Explanations (SHAP). Our results show that XGBoost outperforms all models, achieving an accuracy of 93.21% and a cross-validation accuracy of 94.08%, validating its
robustness. SHAP analysis reveals that psychological engagement, such as emotional involvement, enjoyment, and immersive flow states, are key drivers of travel intention, with individual SHAP values further elucidating user-specific decision patterns. These findings align with consumer behavior theories, reinforcing the role of psychological and experiential factors in travel decisions. Our study presents a highly accurate, interpretable, and scalable predictive model that advances virtual tourism analytics, providing actionable insights for destination marketing and strategic tourism management. Future research should explore real-time user interactions, adaptive learning techniques, and external variables (social sentiment, economic conditions) to improve predictive accuracy and practical applicability.

Person Detection in Low-Light Environments: Evaluation of an Annotation-Free Approach

2026-06-18T16:33:57+00:00

This work investigates how to modify an existing dataset to enhance the capability of a predictive model' to detect objects in low-light environments. From an initial set of images captured under daylight conditions, synthetic darkened versions were generated. Subsequently, various training scenarios were organized to evaluate the impact of image transformation techniques and data augmentation strategies on predictive performance. The experiments performed demonstrate that the combination of original images and artificially darkened versions, coupled with data augmentation and neutral background images, results in more stable and accurate models for low-light detection, with negligible performance degradation in visible-light scenarios. The results highlight that it is possible to improve detector performance in real-world low-light settings without the need for new image acquisitions or additional annotation.

System Identification of an Aeronautical Application Precooler Heat Exchanger Using Neural Networks in an E-TUNI Configuration

2026-06-29T15:00:29+00:00

Modeling heat and mass transfer in engineering applications is of great importance for the development of efficient systems, especially in aeronautics, where weight and performance are critical. Conventional modeling of heat exchangers using white-box and regression techniques relies heavily on manual adjustment of parameters for high and low fidelity models, which increases cost and limits the direct use of available data. In this article, we apply a black-box approach and present the Euler Type Universal Numeric Integrator (E-TUNI) framework, a novel method for identifying systems. We then compare the performance of 40 topologies of E-TUNI and NARX in reproducing a reference model of the heat exchanger, showing E-TUNI's better performance than conventional direct use of Neural Networks for time-series prediction.

A Neural Network-Based Impedance Synthesis for Multi-Probe Harmonic Tuners

2026-06-02T20:17:06+00:00

This work presents a deep neural network (DNN) approach to solve the inverse tuner control problem in multi-probe harmonic impedance tuners used in RF Load-Pull measurements. In this problem, the desired reflection coefficients at the tuner reference plane are specified and the corresponding probe positions must be determined. The proposed methodology addresses the complex nonlinear interactions inherent in multi-probe systems, where the movement of a single probe shifts the impedance across multiple harmonics, rendering traditional brute-force search methods impractical due to the large number of possible probe configurations. Two feedforward DNN architectures featuring six hidden layers were designed and trained to predict the motor steps required to synthesize target impedances at the tuner reference plane. Experimental validation using a Focus Microwaves iMPT-1818-TC tuner at a 3 GHz fundamental frequency demonstrated prediction accuracies of 96.61% in magnitude and 94.03% in phase. Furthermore, a multi-harmonic model simultaneously controlling 3 GHz and 6 GHz achieved accuracies of 90.15% in magnitude and 87.78% in phase at the fundamental, and 88.12% in magnitude and 74.92% in phase at the second harmonic. By replacing iterative VNA-based searches with a predictive model, the proposed approach significantly reduces tuner pre-characterization time and improves the efficiency of RF transistor Load-Pull measurements.

A Circular Polarization MIMO Antenna using Sequentially Rotated Feeding Network for WLAN Applications

2026-07-07T18:13:06+00:00

This work presents a wideband circularly polarized (CP) MIMO antenna operating around 5.5 GHz. The design consists of two identical 2 × 2 subarrays using aperture-coupled microstrip elements. A 5 mm air gap between substrates is introduced to enhance impedance bandwidth, while a sequential phase feeding network ensures stable CP performance across a wide frequency range. The antenna is analyzed through fullwave simulation software of CST Studio Suite and validated by measurements on a fabricated prototype. Results show a 62.9% impedance bandwidth (4.21–7.67 GHz) and a 40% 3-dB axial ratio bandwidth (5.1–7.3 GHz). The antenna achieves a peak gain of 11.5 dBi with inter-element isolation exceeding 25 dB. These results demonstrate its suitability for wideband CP MIMO applications.

Table of Contents October 2026

2026-07-03T17:01:33+00:00