Keywords: Power and Eneergy Systems, Smart Grids, Automation and Process Control
Abstract: The IEC 61850-6 documentation provides Unified Modeling Language (UML) diagrams to promote efficient and effective understanding of the System Configuration Description Language (SCL). This article further expands the use of these diagrams by proposing new ones for sections not yet covered by the normative documentation. To achieve this, the logic behind the standard’s UML diagrams was reproduced, and the relevant sections of the standard were thoroughly analyzed. This work presents new UML diagrams concerning the base types that define most of the SCL elements, and the SCL description file types meant to be used throughout the intended engineering process in IEC 61850-6. These contributions provide a coherent visual representation of critical SCL structures, enhance comprehension, reduce the learning curve for engineers and developers, and help prevent common misunderstandings. Ultimately, this work contributes to more efficient and reliable IEC 61850 system engineering by expanding the standard’s modeling resources.

Keywords: Robotics and Mechatronics, Industry Applications
Abstract: This work proposes a framework for designing human-robot interactions based on key aspects of the field, communication and the context of the application. The framework was assessed through the implementation of a gesture-based safety mechanism for stopping a robot in an emergency. The safety mechanism was tested with a simulated environment, and its performance was compared against conventional controller-based stop commands at varying distances. Results demonstrate that gesture-based control can reduce response times and extend safety capabilities, particularly when operators are not in immediate proximity to the robot or unable to reach its controls. This study highlights the viability of systematic gesture-based interaction design to enhance safety, usability, and responsiveness in industrial HRI, while also providing a useful framework for assessing the interaction design between humans and robots.

Keywords: Industrial Ultrasound Theory and Application, Industry Applications, Robotics and Mechatronics
Abstract: Numerous industrial applications require measuring liquid levels in wide tanks. Doing so in a non-invasive manner from the outer wall of the tank is either a requisite or an advantage in a large group of those applications. In this work, we evaluate an acoustic technique using transducers fixed on the same side in the outer face of the tank. The transducers operate in a pitch-catch configuration, propagating the acoustic wave within the metallic wall. For a given emitted pulse, the received signal energy varies according to the distance between the transducers, as well as the acoustic impedance of both the metallic wall material and the fluid inside thetank. This technique is applicable for tanks with only one accessible face and is independent of the shape or size of the base section. Experiments were conducted to model the relationship between energy attenuation and liquid level. This study evaluates the sensitivity of the response to detect both the liquid level and the type of liquid used. The tests were performed using commercial automotive oil (SAE 20W50) and water. The fitting of the experimental data allows modeling the energy versus liquid height curve using an exponential function. The exponent depends on the energy transmission coefficient calculated for each fluid, allowing differentiation between them. The experiments presented here contribute to the use of this inspection technique for applications in oil or petroleum tanks.

Keywords: Automation and Process Control, Robotics and Mechatronics, Internet of Things
Abstract: This work presents the modeling, simulation, and control of an inverted pendulum system, focusing on the application of optimal control techniques. Using the Lagrangian method, the system's equations of motion were derived and represented in state-space form, enabling the implementation of a Linear Quadratic Regulator (LQR) controller in MATLAB/Simulink. The system's performance was evaluated through simulations with various input types and under the influence of white noise, emulating real-world operating conditions. The results demonstrated the effectiveness of the LQR controller in stabilizing the pendulum and tracking reference signals, even in the presence of disturbances. As a continuation of this study, a physical prototype is being developed using 3D printing with PETG material, aimed at assistive applications to support individuals with reduced mobility.

Keywords: Automation and Process Control, Electrical Machines and Drives
Abstract: This paper presents an experimental performance comparing Generalized Predictive Control (GPC) and Proportional-Integral (PI) control for three-phase induction motor speed regulation. It highlights GPC’s advantages in steady-state error elimination, disturbance rejection, and flexibility. Through simulations and experiments, the research evaluates both controllers' performance, concluding with practical insights on their applicability in industrial control systems.

Keywords: Industrial Ultrasound Theory and Application, Virtualization, Simulation Techniques and Augmented Reality, Cyber Physical Systems, Digital Twins and Knowledge Systems
Abstract: This paper presents a digital twin-driven simulation framework for modeling and analyzing the inspection of ultrasonic phased arrays through fluid-solid interfaces, with a focus on the beamforming principles of FMC and TFM. The system simulates a phased array immersed in a coupling fluid and evaluates beam propagation and focusing behavior across planar boundaries using high-frequency ray theory. The pipeline integrates field simulation and Snell's law-based delay law computation, fully implemented in Python and grounded in canonical models from the literature. Quantitative evaluation is performed using FWHM metrics and delay law resolution analysis. The results confirm the ability of the framework to replicate key focus characteristics and quantify resolution degradation due to delay quantization. Multiple FWHM estimation methods are benchmarked, reinforcing the framework's utility to validate and tune FMC-TFM-based strategies in multilayer media environments.

Keywords: Additive Manufaturing, Personalized Products
Abstract: A impressão 3D tem se consolidado como uma ferramenta promissora no desenvolvimento de tecnologias assistivas e soluções de reabilitação, possibilitando a personalização de dispositivos de forma acessível e eficiente. Este estudo revisa aplicações da impressão 3D em áreas como próteses, órteses, mobilidade funcional e atividades de vida diária, discutindo sobre seus potenciais benefícios aos usuários, bem como os desafios para implementação na prática de reabilitação. Neste sentido, o estudo destaca o papel da interdisciplinaridade, aliando conhecimentos de saúde, design e tecnologia na promoção de colaborações entre instituições de reabilitação e de ensino no estabelecimento de estratégias de promoção de conhecimento e treinamento de uso das tecnologias. Destaca-se também a importância de abordagens colaborativas com a participação dos usuários para garantir soluções alinhadas às suas necessidades, bem como da capacitação de profissionais para o uso eficaz dessas ferramentas na prática clínica.

Keywords: Virtualization, Simulation Techniques and Augmented Reality
Abstract: A utilização de veículos blindados dentro do Brasil Exército remonta à era pós-Segunda Guerra Mundial, marcada pela aquisição do Veículo Blindado de Transporte de Pessoal M113. Contudo este veículo não atendeu totalmente aos requisitos para combate montado, comprometendo assim a segurança e eficácia das operações. Em 2009, o Exército introduziu o Tanque de Batalha Principal Leopard 1A5, modernizando seu Tanque Regimentos. Atualmente, a substituição do M113 BR Veículo blindado de transporte de pessoal pelo veículo blindado de combate | para Soldados de Infantaria (IFV) visa melhorar a operacionalidade capacidades das Forças-Tarefa Blindadas (FT Bld). Este novo veículo apresenta diferenças fundamentais que exigem a adaptação de doutrina, táticas, técnicas e procedimentos para Forças-Tarefa de Infantaria Blindada e Cavalaria. Este artigo analisa o desempenho de uma Tarefa Blindada Força equipada com o IFV, destacando o antecipado impactos operacionais. Espera-se que a transição incorporar novos recursos e melhorar as operações eficiência, considerando as limitações e capacidades de a nova plataforma em comparação com a anterior. O O Estado-Maior do Exército aprovou os Requisitos Operacionais para o IFV em 2020, significando um passo significativo na modernização das Forças Armadas brasileiras.

Keywords: Additive Manufaturing, Virtualization, Simulation Techniques and Augmented Reality
Abstract: Additive manufacturing, particularly 3D printing, has proven to be an efficient solution for low-scale production applications. In this study, components for 65 units of forklift simulators were fabricated using 3D printing technologies, such as Selective Laser Sintering (SLS) and Stereolithography (SLA). The objective of this work is to estimate the service life of these simulator components and determine the required quantity of replacement parts. To achieve this, a fatigue life analysis was implemented using the Juvinall method, adapted for the materials employed in the printing processes. The calculated fatigue life and part quantities were subsequently validated through Finite Element Method (FEM) simulations. The results showed a strong correlation between both approaches, indicating that all components can withstand the programmed operational design life of five years.

Keywords: Optimization Heuristics and Methods, Additive Manufaturing, Industry Applications
Abstract: Topology optimization is an efficient tool in structural and product engineering, enabling the design of efficient, cost-effective, and high-performance components. Therefore, this study presents a topology optimization-enhanced redesign process for industrial components of an additively manufactured 3D handbrake instrument and subsequent its structural evaluation using Finite Element Analysis (FEA). Hence, the aim is to enhance structural performance and manufacturing adaptability in a competitive industrial context – specifically by reducing production time and material consumption for a brake lever bar and its base support. In this path, optimization was conducted using both commercial software Altair and standard MATLAB implementation. The methodology involved defining and optimizing the initial models with each approach, converting the results into three-dimensional CAD models, and subsequently evaluating their mechanical performance via numerical simulations in FEA/CAE software (ANSYS). Finally, the designs were sent for additive manufacturing by Fused Filament Fabrication (FFF). The results demonstrate the potential of topology optimization in improving structural efficiency while addressing manufacturing requirements.

Keywords: Power and Eneergy Systems, Renewable Energy
Abstract: Electric power systems have undergone significant transformation with the widespread implementation of Distributed Generation (DG), especially in distribution networks. In Brazil, the development of photovoltaic (PV) DG has notably affected energy suppliers, potentially causing reverse power flow in distribution feeders, overvoltage issues and problems with coordination in protection systems. Consequently, it is essential to calculate the maximum Hosting Capacity (HC) allowed for PV systems connected to medium voltage networks, ensuring proper operation under normal conditions. The main objective of this paper is to develop a methodology to determine the HC of distribution systems with PV-DG, considering two critical factors: reverse power flow, and overvoltage. The feeders are modelled using OpenDSS, integrated with a Python interface to facilitate enhanced analysis and enable a large number of simulations. The results suggest that the developed methodology is a promising tool for determining the HC of distribution systems.

Keywords: Power and Eneergy Systems, Artificial Intelligence
Abstract: Electricity consumption in the residential sector has increased over the last few decades. With the introduction of new technologies, different types of loads have been used and consumer behaviour has changed. Therefore, it is extremely important to improve electricity consumption models for residential units to incorporate changes in consumer habits and uncertainties. This article proposes the use of an Interval Type-2 Fuzzy Inference System (IT2FIS) to estimate the electricity consumption of residential units, taking into account aspects of consumer behaviour, climatic conditions and socio-economic aspects. The database used to evaluate the models contains 2073 hourly consumption samples, obtained from the 2019 Survey of Ownership and Usage Habits of Electrical Equipment in the Residential Class, carried out by the Programa Nacional de Conservação de Energia Elétrica - Procel. The models were implemented using the Pyhton programming language. The results were evaluated using Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Mean Square Error (MSE) and Root Mean Square Error (RMSE). The results obtained were satisfactory and demonstrate that IT2FIS models can correlate human behavioural aspects, social and economic factors, and climatic conditions in order to estimate hourly electricity consumption in the residential sector.

Keywords: Power and Eneergy Systems, Optimization Heuristics and Methods, Industry Applications
Abstract: Non-technical losses (NTLs) cause significant financial impacts to power utilities. They threaten public safety, compromise the energy quality, and reduce the power grid's reliability. Studies related to NTLs commonly use machine learning algorithms to generate a list of consumer units (CUs) with suspected irregularities, without considering the costs of field inspections. In this context, this study aims to optimize the cost-benefit ratio of field inspections. In this sense, a genetic algorithm is applied to trace optimized routes, defining an optimal sequence of CUs to be inspected by power utilities. A moving average algorithm estimates the theft energy by an irregular CU. In this way, inspections are prioritized in CUs whose estimated value of lost energy exceeds the operational cost of the inspection. The results are presented through subtours, representing the optimal inspection routes for a subset of CUs. For each subtour, the route to be taken and the estimated financial return are provided, contributing to minimizing the operational cost of inspections and making it possible to combat NTLs.

Keywords: Power and Eneergy Systems, Electrical Vehicle and Energy Storage, Renewable Energy
Abstract: Electric power systems have been undergoing the implementation of new technologies, increasing their operational complexity in terms of monitoring, processing, and automation. In this scenario, the growth of distributed generation stands out, especially photovoltaic and wind generation, along with the popularization of Distributed Energy Resources (DERs). However, these advances bring challenges related to system stability, as DERs are not dispatchable, which can cause voltage fluctuations and reduce system inertia, making disturbance compensation more difficult. Additionally, the increasing adoption of electric vehicles represents a volatile load with complex behavior, impacting voltage levels in the grid. This study aimed to understand the effects of a classical Volt-Var control algorithm applied to storage systems based on supercapacitors in managing voltage variations caused by high levels of photovoltaic generation and significant load fluctuations.

Keywords: Power and Eneergy Systems, Renewable Energy, Power Quality
Abstract: O crescente aumento da demanda energética global tem impulsionado diversos setores da academia e da indústria a desenvolver tecnologias voltadas ao uso eficiente da energia. Atualmente, diferentes tecnologias vêm sendo integradas como Recursos Energéticos Distribuídos (REDs) nas redes de distribuição, exigindo, assim, novos modelos de negócio que favorecem a maior penetração desses recursos nos sistemas elétricos. Entre essas tecnologias emergentes, destacam-se os veículos elétricos, os quais, por meio do conceito Vehicle-to-Grid (V2G), podem trazer benefícios tanto para os consumidores quanto para as concessionárias de energia. Neste contexto, o presente artigo tem como objetivo analisar a viabilidade técnica e econômica da inserção de veículos elétricos em uma rede de distribuição utilizando o modelo Vehicle-to-Building (V2B) utilizando o software OpenDSS. A análise considera os impactos dessa inserção na rede elétrica e propõe estratégias para sua implementação, visando ampliar a participação ativa dos usuários no sistema. Os resultados obtidos permitiram identificar cenários promissores para a aplicação do modelo V2B, evidenciando benefícios como a redução dos custos operacionais, melhoria da gestão da demanda e maior resiliência do sistema elétrico, reforçando assim o potencial dos veículos elétricos como agentes ativos na transição energética.

Keywords: Artificial Intelligence, Automation and Process Control, Resource Efficienty & Circular Economy Tracking
Abstract: O avanço da Indústria 4.0 e a crescente adoção de fontes renováveis ampliaram a relevância das baterias de íons de lítio, tornando necessária a adoção de estratégias inteligentes de controle térmico. Nesse contexto, este trabalho apresenta o desenvolvimento de uma lógica de controle adaptativa aplicada a um sistema preditivo de resfriamento, baseada em previsões de temperatura obtidas por redes neurais treinadas com dados gerados em simulações de ciclos de carga e descarga de baterias, utilizando a biblioteca PyBaMM em conjunto com algoritmos de Machine Learning. Foram avaliadas arquiteturas como CNN-1D e Elman-RNN, analisando-se o desempenho tanto na previsão da temperatura quanto na resposta adaptativa da lógica simulada para o acionamento de um ventilador em ambiente virtual. Os resultados evidenciam a viabilidade dos modelos de redes neurais para aplicações embarcadas, contribuindo para o aumento da eficiência e da segurança em sistemas de armazenamento de energia.

Keywords: Artificial Intelligence, Automation and Process Control, Electrical Machines and Drives
Abstract: This work proposes an automatic method for classification of partial discharge signals in insulation systems of high-voltage electric generators, using a computer vision technique based on the Bag of Features algorithm in a database com- posed by images of Phase-Resolved Partial Discharge of corona, internal and surface discharges in the electrical insulation of the stator bars of high voltage generators. The algorithm was implemented in a computational environment, combining the extraction of characteristics from images through the Bag of Features technique with the supervised learning process, Support Vector Machine, in the training actions and classification of signals of the automated routine. The proposed method achieved a very high classification accuracy for the three types of partial discharge signals. The results demonstrate the feasibility of integrating this approach into predictive maintenance systems for high-voltage electrical equipment, in this case, high-voltage electric machines, enhancing operational reliability and reducing costs in maintenance processes.

Keywords: Artificial Intelligence, Industry 4.0, Automation and Process Control
Abstract: A indústria siderúrgica desempenha um papel vital na economia brasileira, sendo uma das principais impulsionadoras da produção e exportação de aço — um insumo essencial para setores como construção civil, automotivo e infraestrutura. Dentro dessa cadeia produtiva, o lingotamento contínuo se destaca como uma etapa crítica, especialmente devido aos desafios associados a defeitos dimensionais e superficiais nas placas de aço. Essas falhas comprometem a qualidade do produto, reduzem o rendimento metálico e resultam em perdas financeiras significativas.

Este trabalho apresenta o desenvolvimento de um sistema de inspeção dimensional para placas de aço imediatamente após o lingotamento contínuo, com foco na medição da largura por meio de técnicas de visão computacional. O objetivo do sistema é verificar se as placas atendem à largura mínima exigida pelo processo e se a diferença entre as larguras das duas extremidades está dentro dos limites tolerados, possibilitando a detecção de formatos trapezoidais indesejados. Para isso, foi instalada estrategicamente uma câmera de espectro visível para capturar imagens da superfície superior das placas, permitindo a segmentação dos quadros e o processamento em tempo real.

A metodologia envolveu aquisição sistemática de imagens em ambiente industrial, aplicação de algoritmos de segmentação utilizando o modelo YOLOv11 e validação das medições com base em dados fornecidos pela usina siderúrgica parceira. Os testes demonstraram que o sistema atingiu uma precisão média de medição de 5,9 mm quando comparado às medições manuais realizadas pela equipe operacional da planta.

Keywords: Artificial Intelligence, Deep Learning and Machine Learning, Industry 4.0
Abstract: Este trabalho realiza uma avaliação comparando o desempenho das arquiteturas YOLOv11 e YOLOv12 para detecção automática de defeitos superficiais em chapas metálicas. Foram utilizados três conjuntos de dados industriais - NEU-DET (1.800 imagens, 6 classes), GC10-DET (3.570 imagens, 10 classes) e CR7-DET (4.140 imagens, 7 classes) - contendo diversas classes de defeitos, desde trincas e inclusões até manchas e arranhões. As versões Nano e Small de cada arquitetura foram avaliadas em 100 épocas de treinamento, considerando múltiplas métricas de desempenho: precisão média (mAP@0.5 e mAP@0.5:0.95), recall e velocidade de inferência (latência e FPS). Os resultados demonstram que a YOLOv12 apresenta vantagens em precisão para certos tipos de defeitos (mAP@0.5 de 0.8583 no CR7-DET), particularmente os de menor tamanho, enquanto a YOLOv11 mantém superioridade em eficiência computacional (latência de 109.05ms na versão Nano). Esta análise comparativa fornece bases técnicas para a implementação dessas arquiteturas em sistemas de inspeção, auxiliando na seleção da solução mais adequada conforme os requisitos específicos de cada aplicação industrial

Keywords: Artificial Intelligence, Large Scale and Network Control, Life Support Systems & Techniques
Abstract: A maturidade digital tornou-se elemento central para orientar a transformação digital nos sistemas de saúde. Avaliar esse nível de maturidade permite identificar lacunas, definir estratégias e alinhar investimentos. Este estudo teve como objetivo identificar e analisar os principais modelos e indicadores utilizados na avaliação da maturidade digital em organizações de saúde. Para isso, realizou-se uma revisão sistemática da literatura na base Scopus, resultando em 12 artigos selecionados conforme critérios de relevância e rigor metodológico. Os resultados apontam a existência de diferentes modelos com múltiplas dimensões como estratégia, governança, interoperabilidade, infraestrutura e cuidado centrado no paciente, mas com limitações quanto à validação prática, clareza de métricas e aplicabilidade em contextos diversos. Os níveis de maturidade observados nas organizações avaliadas variam entre baixos e moderados, evidenciando a necessidade de estratégias mais robustas de transformação digital. O estudo contribui ao consolidar evidências sobre o tema e oferecer subsídios para o desenvolvimento de um modelo de avaliação mais adequado à realidade da saúde pública brasileira.

Keywords: Deep Learning and Machine Learning, Industry Applications, Diagnosis, Prognosis and System Identification
Abstract: A passagem de escória da panela para o distribuidor durante o lingotamento contínuo compromete a qualidade do aço podendo reduzir a tenacidade e ocasionar retrabalho. Além disso, a entrada de escória no distribuidor pode levar à formação de inclusões não metálicas que se depositam em válvulas e canais, provocando entupimentos que interrompem o processo e acarretam perdas financeiras devido a paradas não planejadas. Métodos convencionais de detecção, como inspeção visual, análise do peso da panela, sensores ópticos e eletromagnéticos, ainda predominam nas usinas, apesar da subjetividade, alto custo ou baixa repetibilidade. Este trabalho propõe uma aplicação baseada em redes neurais recorrentes Long Short-Term Memory (LSTM) para prever automaticamente a passagem de escória por meio da análise de sinais de vibração. Os dados foram coletados com um acelerômetro triaxial operando a 6,4 kHz e rotulados com base em evidências visuais de borbulhamento. Após aplicação de filtro Butterworth, segmentação em janelas sobrepostas e extração de métricas estatísticas (raiz do valor quadrático médio, desvio-padrão, curtose, assimetria e taxa de cruzamento por zero), foram treinadas duas variantes da rede, uma com a métrica estatística mais representativa (RMS) e outra com coeficientes extraídos da CWT+CUSUM. Uma análise espectral prévia revelou atenuações entre 6–20 Hz em cerca de 35% das corridas, indicando possível assinatura vibracional associada à escória; por isso, o treinamento foi feito uma das vezes com essa faixa de frequência. Embora a acurácia global tenha sido alta, análises qualitativas indicam que o modelo nem sempre identifica adequadamente o evento de escória, especialmente em corridas com menor sinal vibracional ou baixa relação sinal-ruído.

Keywords: Robotics and Mechatronics, Internet of Things, Deep Learning and Machine Learning
Abstract: This research addresses industrial-scale data science challenges in precision agriculture through the development of an affordable embedded system for real-time weed detection. By applying advanced deep learning optimization techniques to YOLOv5 neural networks deployed on edge computing platforms, we demonstrate how industrial machine vision can be democratized for smaller operations. Our methodology combines post-training quantization (16-bit floating point, 8-bit integer) and ONNX format conversion with systematic performance benchmarking on the Raspberry Pi 4B platform, representing a generalizable approach to edge AI deployment in industrial settings. Comprehensive evaluation on agricultural image data revealed two optimal configurations for different industrial use cases: YOLOv5l with int8 quantization achieved highest detection accuracy (mAP@0.5: 0.815) for stationary applications, while YOLOv5n with ONNX conversion delivered real-time performance (2.66 FPS) with acceptable accuracy (mAP@0.5: 0.727), enabling tractor-mounted operations at speeds up to 10 km/h. Beyond agriculture, this quantization framework offers valuable insights for industrial IoT applications requiring efficient AI deployment in resource-constrained environments. The system represents a significant advancement in affordable industrial machine vision, potentially reducing agricultural chemical usage by 60-80% through site-specific application—demonstrating how edge computing and deep learning optimization can deliver substantial economic and environmental benefits across industrial sectors.

Keywords: Deep Learning and Machine Learning, Artificial Intelligence
Abstract: The present study proposes a computer vision-based model for the early detection of Brown Spot disease (Helminthosporium oryzae) in rice plantations, a pathology that has the potential to cause significant crop losses. The methodology integrates image segmentation techniques, including Watershed and GrabCut, with a customized Convolutional Neural Network (CNN) and a Deep Neural Network (DNN), implemented in TensorFlow. A public Kaggle dataset containing images of healthy and diseased rice leaves was utilized for training and testing purposes. Subsequent to the pre-processing stage, the images were segmented to isolate regions of interest. These regions were then subjected to feature extraction and classification using CNN-DNN. The results obtained were promising, demonstrating the model’s effectiveness in dentifying the disease and its ability to distinguish between healthy and diseased leaves. In comparison with previous studies, the model demonstrated superior performance, underscoring the potential of customized architectures for addressing agricultural challenges. This work contributes to the field of precision agriculture by offering an automated solution for detecting leaf diseases, with practical implications for reducing losses in rice crops. The subsequent trajectory of this field will entail the implementation of field tests and the integration of the technology with embedded devices, with the objective of facilitating real-time applications.

Keywords: Deep Learning and Machine Learning
Abstract: Technological advances have amplified medical image analysis, particularly for the detection of brain tumors. However, current methods for analyzing these images have limitations such as data security, accessibility, and collaboration. This study integrates blockchain technology with the swin transformer deep learning model to construct a decentralized system for secure storage, retrieval, and sharing medical images using encryption, access control, and distributed storage capabilities. Using Ethereum blockchain, InterPlanetary File System (IPFS) storage, smart contracts, and Advanced Encryption Standard (AES), this system ensures data integrity, immutability, and traceability. Moreover, this study investigates the benefits of combining blockchain technology with the swin transformer model, as current methods are also not ideal for accurately predicting brain tumors. Recognizing the relationship between blockchain technology and deep learning models, this study constructs a combined dataset (such as Figshare MRI brain tumor dataset, SARTAJ, Br35H) of 7023 MRI images from various sources. The framework achieves an accuracy of 98.93% in the prediction of brain tumors, while it improves the ways to use medical images and patient outcomes with the help of blockchain to secure healthcare data.

Keywords: Artificial Intelligence, Deep Learning and Machine Learning
Abstract: This study presents an evaluation of the effectiveness of the YOLO11s model for automated detection of colorectal polyps in colonoscopy images. A combined dataset of 3,224 images from the public Kvasir-SEG and CVC-ClinicDB datasets was used, including 1,128 negative images (without polyps) and post-resection cases, aiming to improve robustness to clinical variability and reduce false negatives.

Segmentation masks were converted into bounding boxes to match the YOLO architecture requirements. Images were standardized in resolution and organized according to the Ultralytics YOLOv11 structure. Data were split proportionally between positive and negative cases: 70% for training and 30% equally divided for validation and testing.

The model was trained for up to 200 epochs with automatic hyperparameter tuning and early stopping, using a patience of 20 epochs. Evaluation metrics included Precision, Recall, mAP@0.5, and mAP@0.5:0.95. On the original test set, YOLO11s achieved 96.8% mAP@0.5, 96.7% F1-score, 95.6% precision, and 97.6% recall. In subsets of the multicenter PolypGen2021 dataset, the model reached F1-scores above 84% in distinct clinical centers without retraining.

Despite computational constraints imposed by a low-performance NVIDIA GeForce MX130 GPU, YOLO11s demonstrated robust and generalizable performance, confirming its feasibility for real-time clinical applications even in resource-constrained environments.

Keywords: Electrical Vehicle and Energy Storage, Industry Applications, Industry 4.0
Abstract: Auxiliary lithium batteries play a key role in hybrid and mild-hybrid (HEV and MHEV) vehicles, ensuring power supply to safety-critical and electronic control systems during engine-off phases and transient events. Despite their growing relevance, standardized testing methodologies suited to their specific operational dynamics remain limited. This study proposes a certification-oriented experimental framework for the characterization of 12V and 48V lithium-ion auxiliary batteries in hybrid applications. The methodology integrates Design of Experiments (DOE), Hybrid Pulse Power Characterization (HPPC), and Electrochemical Impedance Spectroscopy (EIS), addressing stress factors such as temperature, depth of discharge, and cycling regimes. Aligned with international standards—particularly IEC 63118-1:2024—the framework aims to inform certification processes and regulatory evolution within Brazil’s Rota 2030 initiative. Importantly, this study focuses exclusively on low-voltage auxiliary batteries for hybrid and mild-hybrid vehicles, excluding applications in pure electric vehicle traction systems.

Keywords: Big Data in Industry Applications, Internet of Things, Industry Applications
Abstract: This paper presents the development of an automated system for temperature control and monitoring in industrial reservoirs, with a focus on food processing applications such as tomato sauce pasteurization. The proposed solution aims to ensure thermal stability and product safety through an architecture composed of a Siemens S7-200 Programmable Logic Controller (PLC) integrated with an ESP32 microcontroller responsible for data acquisition. PT100 temperature sensors were used, along with industrial communication protocols such as Modbus RTU over RS485, and the MQTT protocol for remote supervision. Temperature control was implemented using an on-off logic strategy, seeking to reduce energy consumption without compromising process quality. A graphical interface was developed in Python to allow real-time monitoring and data logging for later analysis. Experimental results demonstrated that the system effectively maintained the reservoir temperature within the ideal range (60 °C to 70 °C), ensuring product integrity. The results indicate that the solution is technically effective, economically feasible, and has potential for application in small and medium-sized industries.

Keywords: Cyber Physical Systems, Digital Twins and Knowledge Systems, Industry 4.0, Resource Efficienty & Circular Economy Tracking
Abstract: The pursuit of sustainability in industry requires assessing environmental impacts throughout the product life cycle (PLC). Life cycle assessment (LCA) and circular economy (CE) are essential tools for this purpose. However, monitoring and evaluating assets across all life cycle phases remains a challenge. Digital twins have emerged as a promising approach to support sustainability goals, yet there is still no consolidated architecture that integrates digital twins with sustainability metrics in a structured and scalable way. This paper presents a systematic literature review to investigate definitions of sustainability, how it can be measured, and how digital twins have been applied in this context. The findings confirm a growing trend toward leveraging digital technologies — particularly digital twins, the internet of things, and artificial intelligence — to enhance sustainability metrics such as energy consumption, emissions, and material usage. Although relevant studies propose simulations and predictive actions based on real-time data, most contributions remain fragmented or theoretical, with limited application across the full PLC. This highlights a gap in comprehensive and integrative frameworks. The study concludes by outlining the need for a robust digital twin architecture that operationalizes sustainability assessment across all PLC stages. Limitations related to database coverage and conceptual heterogeneity were recognized, and the importance of future work that includes the development and validation of a digital twin architecture was reinforced.

Keywords: Robotics and Mechatronics, Artificial Intelligence, Industry 4.0
Abstract: This work presents a comprehensive literature review and comparative analysis of state-of-the-art approaches to solve the inverse kinematics problem in industrial robots, including analytical, numerical, and artificial intelligence-based methods. Analytical solutions, most commonly used by robot manufacturers, although efficient and accurate, struggle with singularities and lack generalization. Numerical methods offer greater adaptability, albeit at higher computational cost. AI based techniques, particularly hybrid and metaheuristic algorithms, have shown promise in addressing singularities and improving generalization, although they require large training datasets and longer development times. This study systematically reviews and compares these methods, highlighting their advantages and limitations in different scenarios. The findings assist in selecting the most suitable inverse kinematics technique based on accuracy, computational performance, and real-time applicability.

Keywords: Cyber Physical Systems, Digital Twins and Knowledge Systems, Industry 4.0, Smart Grids
Abstract: A demanda por fontes renováveis e a necessidade de sistemas de armazenamento de energia impulsionam a adoção de tecnologias de aquisição e supervisão de dados capazes de acompanhar e gerenciar o desempenho desses sistemas. Este trabalho apresenta o desenvolvimento de um Supervisory Control and Data Acquisition (SCADA) voltado ao acompanhamento do funcionamento de baterias em sistemas de armazenamento de energia. A metodologia empregada neste trabalho apresenta o desenvolvimento de um sistema SCADA integrado a um Digital Twin (DT) voltado para o gerenciamento térmico de baterias. O sistema proposto foi implementado com tecnologias de código aberto, incluindo Flask, Docker, Dojot e tecnologias web, operando em uma arquitetura Cliente-Servidor. A aplicação possibilita o monitoramento online de grandezas elétricas e térmicas, como temperatura da bateria, Estado de Carga (SoC, sigla do termo em inglês), Estado de Saúde (SoH, sigla do termo em inglês) e tempo de operação. Dessa forma, o sistema pode ser utilizado pelos gestores e operadores na tomada de decisões, permitindo a identificação de anomalias térmicas e auxiliando na manutenção preventiva. Os resultados demonstram que o sistema desenvolvido possibilita o monitoramento do comportamento térmico das baterias, contribuindo para o aumento da vida útil e segurança dos sistemas de armazenamento de energia elétrica. O sistema desenvolvido foi originado no âmbito de um Projeto de Pesquisa e Desenvolvimento da Agência Nacional de Energia Elétrica (ANEEL), através de uma parceria entre a Concessionária de Energia Norte Energia S.A, a Universidade Federal do Pará e outras empresas privadas.

Keywords: Artificial Intelligence, Electrical Vehicle and Energy Storage, Machine Learning
Abstract: Approximately one fifth of global CO_2 emissions are due to road transportation. To cope with that, Battery Electric Buses (BEBs), among other low emission transportation technologies, are being adopted worldwide. Given that, properly managing BEB fleets is essential to achieve economic and technical viability. This management depends on real time actual data obtained from buses, where monitoring energy-related data is crucial for a efficient and safety operation. In this context, detecting anomalies is essential and particularly challenging due to the lack of labeled data for supervised approaches. This work addresses this challenge through a comprehensive evaluation of unsupervised learning algorithms for energy anomaly detection in electric buses. We analyze four key techniques: the statistical Z-score method, density-based KDE and LOF, and distance-based KNN. Using real-world Bus Controller Area Network (CAN) data, we apply the Agreement Score to evaluate the consistency and effectiveness of these methods without ground-truth labels. The study presents a critical comparison and provides the trade-off and applicability of unsupervised techniques in the context of energy consumption anomaly detection.

Keywords: Artificial Intelligence, Industry 4.0, Automation and Process Control
Abstract: Desalinhamentos de correias são um problema constante nas indústrias, levando a tempo de inatividade operacional e gastos de recursos. Este artigo apresenta um sistema automatizado eficaz para detectar desalinhamentos de correias transportadoras usando o algoritmo YOLO e técnicas de visão computacional. O método demonstrou precisão em testes realizados em ambiente controlado, com erro médio inferior a 1,5 mm. Esta abordagem mostra potencial para integração em sistemas de monitoramento contínuo industrial, garantindo maior eficiência operacional e confiabilidade.

Keywords: Artificial Intelligence, Industry 4.0, Industry Applications
Abstract: Este trabalho apresenta o desenvolvimento de um chatbot inteligente voltado para auxiliar a tomada de decisão operacional em uma concessionária de energia elétrica da região Norte do Brasil. A aplicação utiliza Modelos de Linguagem de Larga Escala (LLMs), embeddings semânticos e a técnica de Recuperação Aumentada por Geração (RAG) para oferecer respostas precisas, contextualizadas e multimodais, a partir de uma base de dados textual e visual indexada. A arquitetura do sistema foi estruturada com o uso da biblioteca LangGraph, que permite o controle dinâmico do fluxo conversacional, e do Milvus, um sistema de gerenciamento de dados vetoriais. A interface desenvolvida com HTML, CSS e JavaScript proporciona uma experiência interativa e acessível ao usuário. Testes demonstraram precisão nas respostas e tempo médio de resposta inferior a dois segundos, evidenciando o potencial do chatbot como ferramenta estratégica para aumento da eficiência operacional no setor elétrico. O estudo destaca a viabilidade técnica e o impacto positivo da aplicação de IA generativa em ambientes industriais críticos.

Keywords: Artificial Intelligence, Deep Learning and Machine Learning, Industry 4.0
Abstract: The growing amount of fines and residues in the steel industry, as well as the constant environmental concern with the emission of gases (CO2) and the large volume of waste generated in iron ore processing plants and the need to dispose of them properly, have motivated the study of technological alternatives that allow the reuse and reprocessing of these materials. A viable alternative is to reintroduce it into the iron and steel manufacturing process itself, through agglomeration techniques, consolidating briquetting as one of the most suitable technologies for using fine ore, coal and waste from the sector in an economical and environmentally friendly way. This work aimed to develop a computer vision system using artificial intelligence for the detection and classification of briquettes in a mining company in the state of Espírito Santo, addressing fundamental concepts of the process and presenting a detailed analysis of the main articles and related works in the area of computer vision applied to the steel industry. YOLOv11 was used as a model for the detection and classification of briquettes and obtained an mAP of 0.84, demonstrating its efficiency and accuracy for this process.

Keywords: Artificial Intelligence, Deep Learning and Machine Learning, Human Machine Symbiosis
Abstract: This study analyzes the performance of three generative artificial intelligence models—Grok (xAI), Gemini (Google), and ChatGPT (OpenAI)—in creating and responding to a questionnaire consisting of 80 questions about the eight production phases of the ASD 2312 tugboat in a shipyard. The phases covered include: Commercial Proposal and Contract; Engineering and Planning; Materials Procurement; Fabrication and Assembly; Electrical and Mechanical Systems; Painting and Finishing; Commissioning and Delivery; and finally, After-Sales. Two naval engineers with over 10 years of experience evaluated the questions and answers based on three criteria—Usefulness, Originality, and Clarity—assigning scores from 1 to 5. The answers were also rated using the same scale, and the variability of the results was examined through standard deviation. The results indicate that the ChatGPT model stood out positively overall, demonstrating greater consistency in answers applicable to shipbuilding. In a criterion-specific analysis, Grok achieved the highest score in Usefulness, Gemini excelled in Originality, and ChatGPT was superior in Clarity.

Keywords: Power Electronics, Renewable Energy
Abstract: This paper presents the development of techniques based on Particle Swarm Optimization (PSO) for Maximum Power Point Tracking (MPPT) in shaded photovoltaic systems. The study analyzes the P&O method and variations of PSO, including DPSO (Deterministic Particle Swarm Optimization), PSO modified by linear and exponential variation of the inertia factor, and PSO modified by neighborhood improvement. All optimization algorithms were applied under partial shading conditions. Simulations were performed using a Cúk DC-DC converter, allowing the evaluation of each algorithm's performance in terms of convergence time, stability and efficiency in tracking the global maximum power point. It is highlighted, through some of the approaches, that modifications to the traditional PSO technique prevent it from getting stuck under partial shading conditions, as well as allowing the algorithm to operate at the global maximum power point, promoting greater efficiency and reliability in photovoltaic systems.

Keywords: Power Electronics, Renewable Energy
Abstract: Photovoltaic solar energy has emerged as a sustainable alternative for electricity generation. However, one of the challenges of implementing these systems in large urban centers is the space required for installation. In this sense, the approach of Building-Integrated Photovoltaic (BIPV) stands out, which unlike the traditional Building-Applied Photovoltaic (BAPV) systems, aims to integrate the modules into the building structure, making them part of the architectural design. The proposed study investigated the impact of pigmented PVA optical filters on the efficiency of photovoltaic modules, with the purpose of allowing traditional modules to also be seen as architectural elements, without losing their functionality of producing clean and renewable energy. Filters of different colors (blue, green, yellow, red and black) were tested on 5 W polycrystalline modules from Komaes Solar. The results indicated that the color of the filter directly influenced the short-circuit current and the maximum power of the module, with the yellow filter presenting the smallest efficiency losses, with values below 8%.

Keywords: Smart Grids, Power and Eneergy Systems, Industry Applications
Abstract: In recent years, the frequency and intensity of heatwaves (HWs) have increased globally, posing substantial risks to human health and the reliable operation of power systems. Among the most vulnerable components of these systems are distribution transformers, whose insulation lifespan is highly sensitive to temperature fluctuations. This study examines the impact of HWs on the operational performance and insulation aging of distribution transformers within a smart campus distribution grid in Brazil. By analyzing electrical load and ambient temperature data collected from 2019 to 2024, 17 HW events were identified. During these events, the hottest-spot temperature of one transformer unit exceeded 140(^{circ})C, potentially damaging its insulation. Moreover, a transformer sizing methodology based on the equivalent aging factor is proposed, integrating daily demand fluctuations and temperature variations. This approach ensures adequate transformer capacity under extreme conditions while minimizing unnecessary technical losses due to oversizing. These findings underscore the urgent need for proactive, climate-resilient planning and management strategies to enhance the reliability of distribution grids in the face of extreme weather events.

Keywords: Electrical Vehicle and Energy Storage, Artificial Intelligence, Diagnosis, Prognosis and System Identification
Abstract: The growing demand for sustainable energy solutions has accelerated the adoption of lithium-ion batteries, particularly in electric vehicles (EVs). One of the main challenges related to this technology is accurately estimating the state of charge (SoC), which is essential to guarantee the system's safety, efficiency, and longevity. This work proposes an approach to SoC estimation by combining machine learning (ML) models — long short-term memory (LSTM), support vector regression (SVR), and random forest (RF) — with statistical replication using the maximum entropy bootstrap (MEB), followed by post-processing with the unscented Kalman filter (UKF). The precision of the proposed model was confirmed using experimental data collected from the Center for Advanced Life Cycle Engineering (CALCE) battery group, conducted on the A123 battery cell under three dynamic profiles: DST, US06, and FUDS. Results demonstrate that the LSTM model achieved the best performance, with approximately 77% lower mean absolute error compared to SVR and 69% lower than RF on the DST profile, and also excelled on the US06 profile, while SVR performed better for the FUDS profile.

Keywords: Power and Eneergy Systems, Large Scale and Network Control, Virtualization, Simulation Techniques and Augmented Reality
Abstract: This work presents a dynamic assessment of load increment variations and their effects on the nonlinear and time-delayed behavior of electric power systems, focusing on long-term voltage stability. Unlike most long-term voltage stability studies that rely on static or steady-state analyses—restricted to equilibrium points—this work highlights the importance of analyzing system dynamics that may not have reached a steady state when subsequent load increments occur. Such dynamics include nonlinear responses and temporized phenomena such as operational limits, dead zones, discrete tap-changer steps, and both fixed- and inverse-time delays. By explicitly considering these non–steady-state interactions, the proposed approach provides a more realistic characterization of how locally driven mechanisms propagate through the network and affect overall voltage stability. The results are obtained through simulations of the IEEE Nordic Test System, a widely used benchmark for voltage stability studies, whose detailed component models enable a comprehensive evaluation of the impact of nonlinear and time-delayed dynamics on system voltage response.

Keywords: Power and Eneergy Systems, Optimization Heuristics and Methods, Large Scale and Network Control
Abstract: The coordination of protection relays in electrical transmission systems is a challenge that requires a balance between efficient protection and adequate coordination and selectivity. In this work a new approach is proposed for the optimized coordination of distance relays (DR) and directional overcurrent relays (DOCR) using a dedicated genetic algorithm. In the objective function it is considered multiple relay adjustment parameters, including nominal and setting currents, time multiplier setting, relay type, actuation time of second zone of distance relays, and operating time of reverse zone. To validate the effectiveness of the proposed approach, an 8-bus test system was utilized, consisting of the adjustment of 28 relays (14 DR and 14 DOCR). The system data was obtained from reliable sources, and simulations were performed using ATP® software, with the genetic algorithm implemented in Python. Simulation results demonstrate that the genetic algorithm can find optimal solutions for relay coordination, respecting all selectivity and coordination constraints and minimizing operation times. The proposed approach proved to be flexible and efficient, capable of handling different types of relays and time settings. Additionally, the inclusion of overcurrent relays with different characteristic curves allowed the evaluation of various coordination possibilities.

Keywords: Power and Eneergy Systems
Abstract: The world is increasingly dependent on electrical energy. With the growing complexity of electrical power systems, reducing the computational time of power flow calculations — the main tool for network analysis — becomes essential. Traditionally, these calculations are based on bus-oriented equations; however, the recently proposed Line-wise Power Flow (LWPF) method has emerged as a more computationally efficient alternative by structuring the problem along transmission lines and transformers, among others. In this context, this work contributes by analyzing the LWPF methodology in depth and by proposing an innovative model for representing voltage control in LTC (Load Tap Changing) transformers within the LWPF framework. This unified approach not only demonstrates computational advantages and structural simplifications but also introduces a novel strategy for coordinated voltage regulation, highlighting its potential for future large-scale applications in power system studies.

Keywords: Power and Eneergy Systems, Industry Applications, Resource Efficienty & Circular Economy Tracking
Abstract: Mineral oil has long been the standard insulating fluid in power transformers due to its excellent dielectric and thermal properties. However, growing environmental and safety concerns have sparked interest in alternative, eco-friendly insulating fluids. Esters have emerged as promising candidates due to their high biodegradability, flame retardance, and lower ecological impact. This paper compares two such insulating fluids—a natural ester (Envirotemp FR3) and a synthetic ester (Midel 7131)—under the influence of lightning impulse voltages, representing a critical stress condition for transformer insulation. High voltage tests, including dielectric loss factor (delta tangent) measurements, were performed before and after applying standardized impulse sequences. Results indicate that both esters maintained dielectric performance within acceptable limits, with the synthetic ester demonstrating superior stability under impulse stress. The findings confirm the technical feasibility of ester-based insulating oils as viable and sustainable alternatives to mineral oil in power transformers, supporting broader environmental and operational safety goals in modern power systems.

Keywords: Power Electronics, Internet of Things, Power and Eneergy Systems
Abstract: This study presents a comparative analysis between Simpson’s Rule and the Trapezoidal Rule for numerical integration applied to the differential protection of transmission lines integrated with renewable energy sources. With the increasing penetration of solar and wind power, traditional protection systems are facing challenges related to frequency deviations, harmonic distortions, and electrical noise. A simulation framework developed in Python was used to model the differential current signals, including artificial pulses and Gaussian noise to emulate real-world disturbances. Integrals were calculated using both methods and compared against a predefined protection threshold. The results show that Simpson’s rule achieves superior numerical accuracy and robustness, effectively distinguishing internal faults from non-critical anomalies and reducing false positives. In contrast, the Trapezoidal Rule showed a higher sensitivity, triggering incorrect trips in borderline cases. To validate the proposed approach in a real-time embedded context, Simpson’s rule algorithm was implemented on an ESP32 microcontroller. The experimental results confirmed consistency with the simulations, demonstrating the feasibility of using Simpson’s rule for real-time protection in smart grids with high renewable energy penetration.

Keywords: Power and Eneergy Systems, Optimization Heuristics and Methods, Renewable Energy
Abstract: A crescente inserção de geração distribuída (GD) nos sistemas de distribuição de energia elétrica evidencia a necessidade de estratégias avançadas de planejamento operacional. Sistemas de Armazenamento Baseados em bateria (SABB) surgem como ativos estratégicos para mitigar desafios técnicos, como sobrecarga de linhas e violações de tensão, além de ampliar a flexibilidade do sistema. Este trabalho apresenta uma abordagem de otimização para a operação de SABB com o objetivo conjunto de minimizar as perdas técnicas e maximizar a capacidade de hospedagem de geração renovável. Adota-se uma metodologia híbrida que combina Algoritmo Genético (AG) para decisões discretas sobre a operação diária dos SABB e um modelo de programação não linear para o cálculo do fluxo de potência e das perdas do sistema. A formulação multiobjetivo é analisa sob diferentes combinações de pesos entre os critérios. Os resultados de simulação demonstram que a otimização multiobjetivo aponta para melhores soluções visando o custo e benefício da operação, sendo priorizado de acordo com a necessidade da distribuidora.

Keywords: Internet of Things, Industry Applications, Life Support Systems & Techniques
Abstract: This paper presents a hybrid Thread network that integrates commercially available devices and open-source Border Routers (BRs) that we developed for Matter-based Internet-of-Things (IoT) deployments. In particular, the developed solution demonstrated the interoperability between Google Nest WiFi Pro and Raspberry Pi-based OpenThread BRs, enabling flexible mesh networks. We showcase a successful Matter communication scenario using ESP32-C6 devices, including LED control actions, emphasizing the value of open ecosystems with multiple manufacturers.

Keywords: Resource Efficienty & Circular Economy Tracking, Bioprocessess Applied to Industry, Personalized Products
Abstract: Expanded polystyrene (EPS) is recognized in the civil construction industry for its thermo-acoustic packaging properties. In Brazil, it is estimated that around 30 per cent of EPS is recycled, being the second most recycled post-consumer plastic material. The search for sustainable products in the construction industry, combining alternative materials and reduced energy consumption, is necessary. The inclusion of EPS in wood panels for thermal insulation is a challenge, given the high pressing temperatures used in their manufacture, which can melt the EPS and adversely affect characteristics such as low thermal conductivity. The present study analyzed the technical feasibility of using post-consumer EPS and residual wood in the production of panels for use as insulating linings. For the panels, residual Pinus sp. wood, castor oil-based polyurethane resin and post-consumer EPS were used, prepared under two press conditions (4 and 4.5MPa) at a ratio of 30%. The properties of the panels were evaluated considering: apparent density, moisture content, 24-hour thickness swelling, static bending strength and modulus of elasticity (MOE). The EPS panel sample identified as having superior physicalmechanical performance was selected for the production of the lining. The internal surface temperature of the roof under analysis was tested with and without the installation of the lining, using infrared thermography. The results indicated that the addition of EPS to the lining contributes significantly to thermal attenuation within buildings (a reduction of up to 17.4°C at certain times of day), as well as contributing to lowering environmental impact, both in terms of manufacturing inputs and thermal amenity and energy consumption within buildings.

Keywords: Industry Data Science Applications, Artificial Intelligence, Deep Learning and Machine Learning
Abstract: A significant portion of the data generated daily by industries worldwide is lost if not properly stored or processed. Many companies rely on computational resources such as local computers or virtual instances on cloud computing platforms to handle this data. On the other hand, companies that operate production processes using Programmable Logic Controllers (PLCs) often experience scheduled downtime during the day—such as shift changes or personnel breaks—when it is not feasible to shut down production machinery. By leveraging multiple idle machines with limited computing capacity, such as PLCs, it becomes possible to distribute and defer machine learning-related tasks across these devices. This strategy enables parallelized execution, resulting in a linear reduction in computation time per connected PLC, and allows for efficient processing even of large datasets. Some limitations were observed during testing, such as memory allocation constraints inherent to the PLC and occasional failures due to high processing intensity. Nevertheless, a reduction in processing time was achieved as additional cluster nodes were added to the data processing system.

Keywords: Industry Data Science Applications, Resource Efficienty & Circular Economy Tracking, Industry Applications
Abstract: A crescente competitividade no mercado de alimentos congelados, impulsionada pela busca por praticidade e qualidade, leva as empresas a buscarem maior eficiência e redução de custos. Nesse cenário, a metodologia DMAIC, do programa Seis Sigma, surge como solução eficaz ao promover a eliminação de desperdícios e a melhoria contínua dos processos. Este trabalho aplicou o DMAIC em uma indústria de alimentos congelados, visando melhorar a qualidade, reduzir desperdícios e aumentar a produtividade. Foram realizadas entrevistas e coletas de dados, identificando problemas na variabilidade da gramatura dos salgados. Após definir objetivos e plano de ação, coletaram-se e analisaram-se dados com ferramentas estatísticas. A análise apontou causas principais, investigadas com o Diagrama de Ishikawa e o Teste dos 5 Porquês. Na etapa de melhoria, implementaram-se ações como padronização de processos e treinamento de funcionários, com monitoramento contínuo. Os resultados mostraram redução da variabilidade, melhoria na qualidade e menor desperdício, aumentando a competitividade da empresa. A metodologia demonstrou ser uma ferramenta valiosa para a gestão da qualidade na indústria de alimentos congelados.

Keywords: Internet of Things, Industry 4.0, Industry Applications
Abstract: The use of acoustic emission sensors has intensified in recent years, especially in industrial environments. The growing demand from Industry 4.0 for real-time monitoring and control has driven this advancement. However, the adoption of smart sensors, such as acoustic emission sensors, is still limited by their high costs, which restricts access primarily for small and medium-sized enterprises. This work aims to investigate the feasibility of using low-cost piezoelectric ceramics for application in acoustic emission sensors, with the goal of reducing overall costs. To this end, the ceramics were encapsulated, and a pencil lead break (PLB) test was conducted to characterize the sensor element, using a commercial sensor from Physical Acoustics, model WSa–AE11, as a comparative reference. Time and frequency domain analyses indicate that certain low-cost ceramics show potential for application, either as a complement to commercial sensors in specific frequency ranges or as a viable alternative in ranges with similar spectral responses. The results obtained demonstrated a correlation coefficient R² of up to 0.889 between the spectra of the commercial and encapsulated sensors, highlighting the achieved spectral equivalence.

Keywords: Automation and Process Control, Robotics and Mechatronics, Virtualization, Simulation Techniques and Augmented Reality
Abstract: The objective of this document is to raise awareness among engineering students and industry professionals about the importance of feedback control. The proposal consists in comparing different strategies in a laboratory session with an Arduino board and Matlab/Simulink. First, students attempt to make a nonlinear pendulum stop at the unstable equilibrium by entering the torque value with a joystick, where different friction scenarios are considered. Then, as a second task, the user enters the position reference and different control strategies are implemented, such as Bang-Bang, PID (proportional-integral-derivative) and FPID (fractional PID) controllers, analyzing the system response by inspecting the signals on an oscilloscope and on a 3D animated model. The control methods are explained from an intuitive point of view, focusing on the meaning and motivation of the control actions.

Keywords: Automation and Process Control, Industry Applications
Abstract: Power amplifiers (PAs) exhibit an intrinsic nonlinear behavior that can hinder their application to high-power radio frequency transmission. Such characteristic of PAs produces severe output signal distortion when the input signal achieves a certain amplitude, thereby limiting their efficiency to maintain a linear behavior. This problem becomes critical with modern technologies, such as 5G or 6G, which employ wide-bandwidth signals, thus implying drastic spectral regrowth phenomena and intermodulation distortion (IMD). In practical terms, the PA spectral efficiency is degraded and the input information becomes corrupted. This paper proposes a control system design based on reference tracking to address this problem. The input-output mapping of a PA is modeled through a state-space representation derived from a classical Volterra series. Then we employ fuzzy Takagi-Sugeno (T--S) representation to exactly describe the nonlinear behavior of the PA in terms of a convex combination of linear local models. This approach enables the design of a fuzzy dynamic output feedback controller using linear matrix inequalities (LMIs) and considering the minimization of the Hoo norm between the input reference and the PA output. Simulation results demonstrate that the proposed technique is able to mitigate the open-loop PA distortion effects, as evidenced by the attenuation of IMD components, especially within the input signal bandwidth.

Keywords: Automation and Process Control
Abstract: The present paper proposes an Advanced Process Control system for controlling in a flexible manner different Indoor Air Quality parameters through natural ventilation, i.e., carbon dioxide, formaldehyde, and Total Volatile Organic Compounds. Different control strategies are designed and compared, i.e., rule-based control strategy, Proportional-Integral-Derivative control strategy based on an override architecture and nonlinear Model Predictive Control strategy. A simulator based on an ad hoc process model is exploited and significant winter season scenarios are selected in order to test and compare the designed control strategies. Tailored Key Performance Indicators are defined and exploited to assess the performance of the designed controllers from a quantitative point of view.

Keywords: Automation and Process Control, Electrical Vehicle and Energy Storage, Industry Applications
Abstract: In the present paper the model of a gridconnected, battery-enabled DC Fast-Charging station for Electric Vehicles is studied. The plant consists of a parametric number of DC Fast-Charging units, a Battery Energy Storage System, a solar-PV unit, and a grid-interface Voltage-Sourced Converter. The Discrete Event System model of all devices and subsystems will be presented thoroughly. The desired operation is expressed in the form of regular languages. The realization of these regular languages by supervisor automata is developed. The physical realizability of the supervisors and the nonblocking property of the controlled automaton are satisfied. The effectiveness of the proposed supervisors is shown through the satisfied properties and the simulation results. Finally, the implementation of the supervisors in Structured Text, for PLC use, is developed in Codesys platform.

Keywords: Automation and Process Control, Industry Applications, Industry 4.0
Abstract: A conectividade entre dispositivos e sistemas industriais é um pilar essencial da Indústria 4.0, aumentando eficiência, flexibilidade e controle. Neste contexto, este artigo apresenta o desenvolvimento de um sistema de controle e monitoramento aplicado a um módulo didático de Sistemas de Manufatura da FESTO, projetado para reproduzir, em escala laboratorial, processos industriais por meio de unidades interconectadas. O projeto foi implementado com um Controlador Lógico Programável, configurado com para gerenciar as funcionalidades do sistema, e complementado por duas formas de interação, ambas voltadas para monitoramento e controle, sendo a primeira baseada em um sistema embarcado para controle local e a segunda em um painel de controle web para operação remota. A comunicação entre os dispositivos foi estruturada com base em dois protocolos MQTT e OPC UA. Foram conduzidos uma série de testes em diferentes cenários com o objetivo de avaliar a funcionalidade de ambos os sistemas. Os resultados demonstraram que a adoção dessas tecnologias é viável, sendo potencialmente escalável para sistemas SCADA em aplicações que não exigem tempos de resposta críticos.

Keywords: Diagnosis, Prognosis and System Identification, Deep Learning and Machine Learning, Machine Learning
Abstract: Microscopic analysis of blood smears remains a gold standard in hematological diagnosis, yet it is traditionally manual, slow, and prone to inter-observer variability. This work intro duces an automated blood cell classification model leveraging the ResNet50 architecture, combined with explainable artificial intelligence (XAI) via SHAP (SHapley Additive exPlanations) to enhance diagnostic transparency and reliability. The dataset used, the Blood Cells Image Dataset, consists of 17,092 high quality annotated microscopic images, covering six cell types: basophils, eosinophils, erythroblasts, lymphocytes, monocytes, and platelets. Data preprocessing included resizing, ImageNet normalization, and controlled augmentation to preserve mor phological characteristics. The model, adapted for multiclass classification via transfer learning, incorporated regularization strategies such as L1/L2 penalties, dropout, early stopping, and ReduceLROnPlateau to optimize training and prevent overfitting. Performance metrics, including global accuracy, class-specific precision, recall, F1-score, and confusion matrices, were utilized. The model achieved an impressive accuracy of 97.33%, with F1-scores exceeding 0.95 across all classes. SHAP visualizations aligned with classic hematological criteria, reinforcing the clinical reliability of the system.

Keywords: Diagnosis, Prognosis and System Identification, Personalized Products, Automation and Process Control
Abstract: Esta pesquisa busca identificar os principais desafios enfrentados pelos microempreendedores do setor varejista de vestuário e acessórios da cidade de São Paulo, com foco nas práticas de gestão financeira e no acesso à educação financeira. Trata-se de uma pesquisa aplicada, de caráter descritivo, realizada por meio de um levantamento (survey). Os dados foram coletados com um questionário estruturado, obtendo-se 31 respostas válidas, que foram submetidas à análise descritiva. Os resultados revelam que, embora a maioria reconheça a importância da educação financeira, muitos empreendedores enfrentam dificuldades na gestão de fluxo de caixa, precificação e planejamento financeiro. A análise qualitativa reforça a carência de conhecimento técnico e aponta o planejamento orçamentário e a separação entre finanças pessoais e empresariais como temas críticos. Além disso, 90,3% dos participantes demonstraram interesse em uma plataforma digital específica sobre educação financeira para o setor, proporcionando oportunidades para o desenvolvimento de soluções educacionais segmentadas. O estudo conclui que políticas e programas de capacitação devem ser adaptados ao perfil dos empreendedores, levando em conta suas preferências de aprendizagem, nível de escolaridade e maturidade empresarial.

Keywords: Diagnosis, Prognosis and System Identification, Industrial Ultrasound Theory and Application
Abstract: Monitoring transformer failures is crucial to ensure the quality of the electrical power system. In this context, this paper evaluated the influence of noise on fault characterization in power transformers using acoustic emission technique. Two types of faults were considered in acoustic measurements under low-frequency noise and white noise: partial discharges in bushings and electric arcs. After analyzing how noise can interfere in the measurements, self-organizing maps were proposed based on statistics such as skewness, kurtosis, effective bandwidth, and average band. In addition, the influence of a 20 kHz high-pass filter was investigated. The results indicate that filtering at 20 kHz can improve the characterization and differentiation of the two faults, and the combination of average band and skewness showed promising results for failure classification. The combination of skewness and kurtosis without filtering also shows promising results.

Keywords: Machine Learning, Artificial Intelligence, Self Configuration & Self Diagnosis
Abstract: The integration of Machine Learning (ML) in healthcare has driven significant advancements in medical imaging analysis and automated diagnosis. One of the ongoing challenges in this area is the detection and classification of multiple abnormalities in a single radiographic examination. This study focuses on evaluating the effectiveness of the YOLO12x (You Only Look Once, version 12x) model in detecting and classifying seven distinct spinal pathologies: disc space narrowing, foraminal stenosis, osteophytes, other lesions, spondylolisthesis, surgical implant, and vertebral collapse.

Trained and evaluated on 5,006 radiographs from the VinDr-SpineXR dataset, the model without preprocessing achieved average values of precision 0.9677, recall 0.8671, F1-score 0.9127, mAP@0.5 0.9271, and mAP@0.5:0.95 0.8501. Class-wise, the strongest performance was observed in spondylolisthesis, surgical implant, and vertebral collapse (F1 geq 0.94; mAP@0.5 geq 0.94). The normalized confusion matrix indicated high accuracy for foraminal stenosis (90%), osteophytes (89%), surgical implant (85%), and vertebral collapse (82%), with lower accuracy for other lesions (72%). A preprocessing pipeline based on CLAHE, Sobel filtering, and intensity normalization did not yield improvements (precision 0.9339; recall 0.7747; F1 0.8454; mAP@0.5 0.8722; mAP@0.5:0.95 0.6917), although it slightly increased precision for foraminal stenosis. These results indicate that YOLO12x reliably detects multiple spinal pathologies in radiographs without preprocessing, supporting real-time clinical workflows; future work should explore class balancing and data augmentation techniques to improve recall in more challenging categories.

Keywords: Diagnosis, Prognosis and System Identification, Deep Learning and Machine Learning, Artificial Intelligence
Abstract: In this paper, we propose a novel edge-AI architecture for real-time monitoring of dementia patients using fine-tuned Large Language Models (LLMs) to detect and classify high-risk scenarios in a low-bandwidth environment. We use LLaMA 3.8B with 4-bit quantization and Low-Rank Adaptation (LoRA) to fine-tune the pre-trained model effectively on a specialized "Dementia Risk Dataset" of 50,000 labeled text samples. The optimized LLM is built to execute on the network edge (e.g., on border routers), with real-time transcriptions of patient speech analyzed for identifying key events such as disorientation, wandering, refusal of care, and medication management errors, and short, structured alerts sent to the cloud. Experimental results provide quick convergence at training, accurate classification of different risk categories, and reduction in message length compared to lengthy raw transcripts. Our approach ensures patient confidentiality by avoiding continuous audio streaming, reduces communication overhead in low-bandwidth networks, and enables timely intervention by clinical professionals and caregivers. The proposed system offers an extensible solution to cognitive dementia care and demonstrates the viability of LLM deployment on low-resource edge devices for domain-specific use cases.

Keywords: Power and Eneergy Systems, Power Quality, Smart Grids
Abstract: It is proposed an analysis on the protective superconductor devices in direct current transmission systems, more specifically in multiterminal HVDC power transmission systems with Voltage Source Converters VSC. The use of Superconductor Current Fault Limiters SCFLs can reduce the transient overcurrent and overvoltage during several fault occurrences, increasing the stability and reliability of both HVDC and HVAC grids

Keywords: Power and Eneergy Systems, Power Electronics, Power Quality
Abstract: This paper presents the modeling and analysis of a hybrid HVDC transmission system that combines Line Commutated Converter (LCC) and Voltage Source Converter (VSC) technologies in a bipolar configuration with ground return. The ±250 kV, 1000 MW, 1000 km system was implemented in ATP software to investigate electromagnetic transients caused by pole-to-ground faults. Three scenarios were considered: (i) energization with firing angle control, (ii) temporary faults at the line entrance and midpoint, and (iii) a permanent fault leading to a bipolar-to-monopolar transition. Simulation results show that entrance faults generate peak currents up to 2.5 p.u., while midpoint faults cause overvoltages reaching 2.2 p.u. Firing angle control of the LCC partially mitigated fault currents but was insufficient to keep them within equipment safety limits. In contrast, monopolar operation ensured system continuity under permanent fault conditions, albeit with higher electrode currents and reduced voltage levels. These findings underscore the asymmetric fault response of hybrid LCC–VSC systems and highlight the need for coordinated control and protection to ensure reliable operation.

Keywords: Power and Eneergy Systems, Renewable Energy, Power Electronics
Abstract: The dynamic response of the load is essential for stability studies of bulk power systems. This response is represented in stability studies by an equivalent, which should represent the dynamic response in different hours of the day. This response depends on the load composition and the level of Distributed Energy Resources (DERs) connected at the distribution level. Considering that the profile of the load composition and usage patterns changes constantly changes, it is important to have approaches able to estimate the best model to represent the best model to represent the load composition dynamic response for different periods of the day. This paper presents a cluster-based approach to identify the patterns of dynamic load response during a day. The approach is applied to the 33-bus distribution system to demonstrate its performance for a distribution system.

Keywords: Power and Eneergy Systems, Optimization Heuristics and Methods, Renewable Energy
Abstract: In this paper it is proposed the development of a short-term approach for electric power distribution system, considering the allocation of fixed and switched capacitor banks, photovoltaic panels, distributed generators, and energy storage systems. The model incorporates uncertainties related to conventional demand and solar irradiance through representative scenarios. Electric vehicles demand is modeled based on a probability distribution curve. The model is solved using a matheuristic based on neighborhood criteria, and the results are compared with those obtained from a mixed-integer second order conic programming model solved by commercial solvers, highlighting computational time and feasibility limitations. The test system used is a 136-node network, and the objective function minimized includes the costs of energy purchase at the substation, investment costs, and operation and maintenance costs. The results demonstrate the effectiveness of the proposed methodology, showing superior performance compared to commercial solvers.

Keywords: Power and Eneergy Systems, Renewable Energy, Machine Learning
Abstract: This paper proposes a methodology to improve fault detection selectivity in wind farm collector lines by integrating a threshold-based technique with an Artificial Neural Network (ANN). The method uses local current and voltage measurements processed through the Stockwell Transform, from which time-frequency features are extracted and used as inputs to a Multilayer Perceptron (MLP). The ANN is trained to identify whether the detected fault is internal to the measurement point or located externally—either at an adjacent collector line or at another transformer. The proposed strategy was evaluated under 55,800 simulated fault cases, covering variations in fault resistance, inception angle, location, generation level, and noise. Results showed 100% accuracy under noiseless conditions, but reduced selectivity in noisy environments. To address this, a retraining strategy using both noiseless and 40~dB data significantly improved robustness, achieving accuracy above 89% across all scenarios. The technique complements conventional detection methods, such as the Current Slope-Base (CSB) index, by ensuring reliable fault region identification even under degraded signal conditions. This work lays the foundation for integrating intelligent classifiers into protection schemes, supporting more selective and adaptive fault identification in renewable power systems.

Keywords: Artificial Intelligence, Smart Grids, Renewable Energy
Abstract: As microrredes vem ganhando grande destaque devido às pressões crescentes de redução de emissões de gases poluentes, utilização de energias renováveis e diversificação da matriz energética. Contudo, a correta gestão de equipamentos se torna um desafio para garantir uma resposta adequada com controles que visam equilibrar a tensão dos barramentos, uma vez que o tempo entre a identificação da necessidade de correção e a atuação do sistema precisa ser rápida. Para isto, foi desenvolvido um algoritmo que controla as cargas e fontes de potência de uma microrrede composta por armazenadores de energia (baterias e supercapacitores), conversor de energia conectado à rede concessionária e cargas controláveis com uso de uma meta heurística bio-inspirada (algoritmo Differential Evolution) que apresentou resultados coerentes com as necessidades do problema.

Keywords: Artificial Intelligence, Industry Applications, Deep Learning and Machine Learning
Abstract: Os transportadores de correia são amplamente utilizados na indústria e estão sujeitos a furos e desalinhamentos que podem comprometer a operação e aumentar os custos de manutenção. Inspeção tradicional como a observação visual e a utilização de sensores, são limitados e geralmente detectam falhas somente depois que elas ocorrem. Este trabalho apresenta parte de um sistema, baseado em Computação Visão e Inteligência Artificial, para detecção automática de defeitos em correias transportadoras. Usando o YOLOv11 arquitetura, o sistema realiza segmentação e falha identificação com alta precisão. Os ensaios realizados numa A planta piloto demonstrou a eficiência da abordagem, reduzindo a necessidade de inspeções manuais e permitindo manutenção mais rápida e eficaz.

Keywords: Artificial Intelligence, Deep Learning and Machine Learning, Human Machine Symbiosis
Abstract: O artigo apresenta o desenvolvimento de uma ferramenta baseada em inteligência artificial (IA) e termografia para o diagnóstico não invasivo da saúde de colônias de abelhas nativas sem ferrão no Brasil. A meliponicultura, atividade de relevância crescente tanto econômica quanto ecológica, enfrenta desafios no monitoramento preciso das colmeias sem a necessidade de interferência direta. No estudo, empregou-se uma câmera termográfica de baixo custo para registrar imagens térmicas do interior das colmeias, as quais foram analisadas por uma Rede Neural Convolucional (RNC) treinada com um conjunto de 251 imagens de colmeias classificadas como ativas, hipoativas ou inativas. O modelo foi implementado na biblioteca de aprendizado profundo de código aberto PyTorch e, após 50 épocas de treinamento, atingiu 93,33% de acurácia na classificação de imagens inéditas. A métrica F1-Score também apresentou desempenho elevado (95,15%), evidenciando a robustez do modelo, por equilibrar a precisão (redução de falsos positivos) e a revocação (redução de falsos negativos). Assim, a proposta mostrou-se promissora, acessível e de baixo custo, configurando-se como uma alternativa viável para práticas sustentáveis na meliponicultura.

Keywords: Artificial Intelligence, Deep Learning and Machine Learning, Machine Learning
Abstract: A evasão escolar representa um desafio crítico nas instituições educacionais, impactando diretamente a eficiência acadêmica e acentuando problemas de desigualdade social. Um dos principais obstáculos para a predição é o desbalanceamento de classes, que consiste na distribuição desigual entre estudantes evadidos e não evadidos. Este estudo utilizou dados acadêmicos, demográficos e socioeconômicos de 4.424 estudantes do Instituto Politécnico de Portalegre (Portugal) para desenvolver modelos preditivos. Foram aplicadas técnicas de reamostragem supervisionada — SMOTE (Synthetic Minority Over-sampling Technique), SMOTE Tomek (SMOTE with Tomek), SMOTE ENN (SMOTE with Edited Nearest Neighbor) e Undersampling — combinadas aos algoritmos Random Forest e XGBoost. A análise comparativa entre os dez modelos foi baseada nas métricas F1-score, AUC-ROC (Area Under the Receiver Operating Characteristic Curve) e precisão/recall. Para interpretar as decisões, foi aplicada a técnica LIME (Local Interpretable Model-Agnostic Explanations). O modelo Random Forest com SMOTE Tomek obteve o melhor desempenho: F1-score de 82,3%, precisão de 84,4%, recall de 80,3% e AUC-ROC de 93,5%. A análise LIME revelou que o desempenho acadêmico nos primeiros semestres é o principal determinante para a predição da evasão. O estudo contribui para o entendimento do impacto das técnicas de balanceamento em contextos educacionais e fornece insights interpretáveis para estratégias de retenção estudantil.

Keywords: Artificial Intelligence, Industry 4.0, Internet of Things
Abstract: This article presents the robotic platform Robô-INOVEE, developed by the Center for Innovation in Energy Efficiency (INOVEE), aiming to support the teaching of robotics and programming logic at different levels of education, from basic education to higher education. Robô-INOVEE can be programmed with the help of free software tools, such as Autodesk TinkercadTM, using coding or visual programming languages, and also supports artificial intelligence (AI)-assisted programming, for example, with the use of Gemini or ChatGPT tools. Based on the Arduino Nano board, Robô-INOVEE offers a robust and versatile architecture equipped with several sensors, actuators and communication modules, allowing its programming for various applications. Integration with AI expands the possibilities for experimentation, making Robô-INOVEE a rich tool for the development of computational thinking, interdisciplinarity and skills formation.

Keywords: Internet of Things, Virtualization, Simulation Techniques and Augmented Reality, Smart Grids
Abstract: Este artigo apresenta a implantação e observabilidade de arquiteturas IoT baseada em middlewares conteinerizados, com base nas plataformas Dojot, FIWARE e Node-RED, implantadas em contêineres Docker executados em máquinas virtuais (VM) gerenciadas pelo hipervisor Proxmox VE, compondo uma infraestrutura flexível para criação e gerenciamento de dispositivos IoT. Para assegurar a visibilidade operacional, uma stack baseada em Grafana, Prometheus e cAdvisor coleta e monitora métricas de desempenho das VMs, incluindo uso de CPU, memória e outros recursos essenciais. A proposta demonstra como a combinação de middlewares conteinerizados e monitoramento avançado pode facilitar a implantação, o gerenciamento de recursos e o planejamento da escalabilidade em sistemas baseados ou integrados a tecnologias IoT.

Keywords: Industry 4.0, Cyber Physical Systems, Digital Twins and Knowledge Systems, Big Data in Industry Applications
Abstract: This paper presents a modular software architecture using open-source technologies for online monitoring of <15 kW electric motors to support condition-based maintenance (CBM). The proposed solution combines a hardware and a distributed backend for data processing and visualization. The system was validated in a testbench with induced motor faults, demonstrating its ability to help the detection of motor anomalies using vibration data analysis

Keywords: Internet of Things, Industry Applications, Industry 4.0
Abstract: The Fine Time Measurement (FTM) protocol, defined by the IEEE 802.11-2016 standard, introduces a bidirectional ranging approach to improve Wi-Fi positioning accuracy. Nevertheless, real-time distance measurements using FTM are susceptible to several sources of error. In this study, we conducted exploratory tests in both indoor and outdoor settings using ESP32-C3 microcontroller-based devices. A dataset comprising 240,000 FTM samples was collected to calibrate and validate optimal protocol configurations. Subsequently, the protocol’s performance was assessed in an electronics manufacturing facility located in the Manaus Industrial Hub. The results demonstrated a Mean Absolute Percentage Error (MAPE) as low as 6.23% under favorable conditions and a maximum Root Mean Squared Error (RMSE) of 5.38 m in complex environments. Additionally, power consumption analysis indicated average currents of 95 mA in Access Point (AP) mode and 72.4 mA in Station (STA) mode during 1.8 ms transmissions. These findings underscore the FTM protocol’s efficacy for realtime distance estimation in environments equipped with Wi-Fi infrastructure.

Keywords: Internet of Things, Industry 4.0, Industry Applications
Abstract: LoRa is a low-power wireless modulation developed by Semtech and is widely used in various Internet of Things (IoT) applications. It typically operates in sub-GHz frequency bands ranging from 863 to 968 MHz, depending on regional regulations. In recent years, Semtech introduced SX1280, a new model of radio that operates in the globally available 2.4 GHz ISM band, which is shared by well-known IoT protocols such as Wi-Fi, Bluetooth, and ZigBee. Since the 2.4 GHz ISM band offers availability worldwide without regional restrictions, this opens the possibility for LoRa 2.4 GHz to be used globally in Industrial Internet of Things (IIoT) applications. This article presents practical throughput tests that compar two LoRa radios and two ZigBee radios measured at various distances in an outdoor environment under line-of-sight conditions. The goal is to evaluate the performance of LoRa 2.4 GHz relative to an protocol IEEE 802.15.4 based, under similar conditions, as both are categorized as Low Power Wide Area Networks (LPWANs) and offer comparable data rates, providing insights into their respective suitability for IIoT applications.

Keywords: Internet of Things, Industry 4.0, Industry Applications
Abstract: Este trabalho apresenta as etapas de caracterização de uma rede mesh composta por microcontroladores ESP32-WROOM-32. A biblioteca ESP-WIFI-MESH, desenvolvida pela empresa Espressif Systems, foi utilizada para implementar o protocolo de comunicação mesh utilizado neste trabalho. Uma interface gráfica, construída a partir da ferramenta de desenvolvimento Node-RED, foi utilizada para exibir, em tempo real, um conjunto de métricas de desempenho associadas aos microcontroladores conectados à rede, tais como latência, intensidade do sinal (RSSI) e número de saltos entre dispositivos, além dos parâmetros transmitidos por meio da rede sem fio. Os experimentos foram realizados em ambiente fechado, simulando condições de atenuação associadas a cenários críticos de comunicação sem fio, utilizando módulos com antenas internas e externas. Nos ensaios realizados, foram transferidos pacotes de até 1024 bytes. Para pacotes de até 64 bytes, os valores de latência medidos variaram entre 10 ms e 30 ms por salto. Esses valores aumentaram de forma progressiva quando pacotes maiores foram transmitidos, variando de 50 ms para pacotes de 128 bytes até 207 ms para pacotes de 1024 bytes. A rede apresentou taxa de entrega de 100% para pacotes de até 512 bytes, mas observou-se uma redução dessa taxa para 81% quando pacotes de 1024 bytes foram transmitidos.