Mestrado em Engenharia Elétrica e da Computação (EMC)

URI Permanente para esta coleçãohttp://200.137.215.59/tede/handle/tde/263

Navegar

Agora exibindo 1 - 20 de 204

Gerenciamento e otimização de doses de radiação em equipamentos de hemodinâmica
(Universidade Federal de Goiás, 2025-09-05) Assunção, Murilo Felisberto Morais de; Silva, Jonas Oliveira da; http://lattes.cnpq.br/6007563566142317; Itikawa, Emerson Nobuyuki; http://lattes.cnpq.br/5014788656693729; Itikawa, Emerson Nobuyuki; Silva, Jonas Oliveira da; Freitas, Marcelo Baptista de
This study evaluated strategies for managing and optimizing radiation doses in hemodynamic procedures, covering aspects from imaging system characterization to the impact of optimization on exposure levels. Initially, equipment parameters were analyzed, including the accuracy of dose indicators and the behavior of AEC, as well as the influence of image processing on noise reduction without compromising diagnostic quality. Dosimetric parameters, including time, PKA, and dose, were systematically recorded for angioplasty, catheterization, vascular exams, and other high-exposure or more frequent procedures, expanding the reference base in the literature. The optimization of clinical protocols resulted in significant reductions in exposure levels. In catheterization, the median PKA was reduced by 16.3%, while in angioplasty, the reduction was 18,6%. The parameters were compared with 17 other national and international studies. For catheterization, the facility’s PKA remained above 63,16% of the analyzed studies, whereas for angioplasty, by the end of the study, it was above only 24%. Occupational analysis indicated a decrease of 64,6% in recording events and 64,8% in investigation events, reflecting greater adherence to radioprotection practices. The high-exposure study revealed that, despite the association with complex anatomical characteristics, no statistical correlation was established with the Complexity Index. The findings highlight the importance of continuous monitoring, automated recording, and active dose management, providing a replicable methodology for optimization in different centers.
Ampliação da Resolução de Sensores LiDAR Utilizando Redes Neurais Artificiais
(Universidade Federal de Goiás, 2025-09-12) Santos, Marlon Franco; Cardoso, Alisson Assis; http://lattes.cnpq.br/8216536516894987; Corrêa, Henrique Pires; Oliveira, Bruno Quirino de; Ferreira, Mateus de Paula
Environment perception with adequate resolution is essential for the safe navigation of autonomous robots. The LiDAR sensor is frequently used for precise distance measurements; however, the number of points in its scan is limited by its hardware design, which can compromise obstacle detection and mapping. This work addresses the problem of low point density in 2D LiDAR sensor, developing and evaluating low-complexity Artificial Neural Networks for resolution upsampling. The main objective is to double the angular resolution of a LiDAR sensor’s scan, with data acquired in a simulated environment using the Gazebo simulator and the TurtleBot3 robot. The measurements were preprocessed and split to train two architectures: a Multilayer Perceptron, using a windowing technique, and a One-Dimensional Convolutional Neural Network. The models were trained on a subset of data, simulating a lower-resolution sensor, with quantitative evaluation performed through the analysis of the Cumulative Distribution Function and the Kolmogorov-Smirnov statistical test. For the qualitative evaluation, a visual analysis of the reconstructed signal was conducted by plotting the results in polar coordinates. The results demonstrated that both models learned the spatial patters and were capable of reconstructing the missing measurements while maintaining the statistical characteristics of the original data. The Multilayer Perceptron architecture showed slightly superior performance compared to the One-Dimensional Convolutional Neural Network, with more stable training losses and less differences in the Cumulative Distribution Function analysis. We concluded that the use of low-complexity Artificial Neural Networks is a viable and effective approach for upsampling 2D LiDAR sensor data, offering a new method to enhance the perception of mobile robots with limited resources.
Seleção adaptativa de proxies com amostragem de Thompson e métodos Bayesianos
(Universidade Federal de Goiás, 2025-09-18) Souza, Paulo Henrique Cardoso de; Marques, Thyago Carvalho; http://lattes.cnpq.br/1763926064124591; Brito, Leonardo da Cunha; http://lattes.cnpq.br/6660680440182900; Brito, Leonardo da Cunha; Marques, Thyago Carvalho; Ramos, Jhonata Emerick; Pires, Sandrerley Ramos
This study investigated strategies for proxy selection in automated data capture systems, comparing traditional approaches with adaptive Bayesian strategies. The main goal was to evaluate the operational efficiency, stability, and adaptive capacity of different selection algorithms in both controlled and real environments. The methodology involved controlled simulations in four distinct scenarios (intermittent proxies, blocked proxies, permanently failed proxies, and heterogeneous proxies) and experimental validation in a real operational environment with 10 different robots performing public data capture from various domains over one week, processing 549,114 requests. Seven strategies were evaluated: four Bayesian (Beta, Gamma, Normal, Chi-Square), one deterministic (Exponential Backoff), and two basic (Round Robin and Random). The simulation results demonstrated the consistent superiority of Bayesian strategies, with the Beta distribution achieving success rates above 99% in critical scenarios and maintaining leadership in the real environment with an average rate of 76.00%. The stability analysis revealed significantly lower coefficients of variation for Bayesian strategies (0.191–0.334) compared to the basic ones (0.498–0.668). The temporal analysis showed that Bayesian strategies wasted 2.5 times fewer resources than basic approaches, demonstrating superior operational efficiency. The Beta distribution stood out for its exceptional ability to differentiate between resources and adapt over time, as evidenced by the detailed analysis of probability distributions. Beyond direct applications in data capture, the developed techniques show significant potential for adaptive anti-scraping systems, where the ability to identify suspicious behavioral patterns and dynamically adapt to evasion techniques can enhance protection mechanisms against automated activities that violate web resource usage policies. It is concluded that Bayesian strategies, particularly the Beta distribution, provide significant operational advantages for data capture systems and transformative potential for the development of adaptive countermeasures in web protection.
Inteligência artificial aplicada na quantificação de nódulos pulmonares em imagens de tomografia computadorizada
(Universidade Federal de Goiás, 2025-06-11) Camargo, Thiago Fellipe Ortiz de; Martins, Marcella Scoczynski Ribeiro; http://lattes.cnpq.br/5212122361603572; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Martins, Marcella Scoczynski Ribeiro; http://lattes.cnpq.br/5212122361603572; Siqueira, Hugo Valadares; http://lattes.cnpq.br/6904980376005290; Ribeiro, Guilherme Alberto Souza; Cruz Junior , Gelson da; http://lattes.cnpq.br/4370555454162131
Accurate quantification of pulmonary nodules in computed tomography remains challenging due to interobserver variability and the lack of scalable methods capable of generalizing across heterogeneous datasets. This study proposes an automated solution that integrates deep learning to generate nodule segmentation masks and compute volumetric measurements using spatial information from the images. The methodology is structured into three main stages: data preparation and pre-processing, model training and validation, and performance evaluation. The use of the nnU-Net architecture, which automates pre-processing, segmentation, and post-processing, provides scalability and dynamic adaptation to the workflow, enhancing the clinical applicability of the solution. The results indicate consistent volume and diameter measurements across successive scans and strong agreement with consensus masks, even for anatomically complex nodules. The 3D U-Net architecture achieved a mean Dice coefficient of DC = 0.7846 with a standard deviation of σ = 0.18, outperforming the interobserver Dice index of DC = 0.5218 and exhibiting a low volumetric deviation between acquisitions. The proposed methodology advances automated quantification of pulmonary nodules, offering a resilient and adaptable solution to support medical diagnosis in real-world clinical scenarios.
Protocolo orientado por inteligência artificial para controle e padronização de manobras em subestações elétricas
(Universidade Federal de Goiás, 2025-06-04) Campos, Gustavo Havilá de Freitas; Reis, Márcio Rodrigues da Cunha; http://lattes.cnpq.br/1167385371830496; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Bulhões, Júnio Santos; http://lattes.cnpq.br/8977468528143965; Araújo, Wanderson Rainer Hilário de; http://lattes.cnpq.br/8514981358543699; Silva, Saulo Rodrigues; http://lattes.cnpq.br/4246717998455436; Júnior, Gélson da Cruz; http://lattes.cnpq.br/4370555454162131
Notwithstanding recent advances in substation automation, there are still no protocols integrates human–machine interaction, intelligent interlocking, operational autonomy, and artificial intelligence analysis in sequential maneuvering contexts. This study proposes an automated interface to optimize and control switching operations in electrical substations by integrating operational protocols, automated documentation generation, and artificial intelligence techniques with interactive graphical visualization. The developed solution enables sequential command execution, classification of operational events, and automatic generation of auditable reports, enhancing accuracy and traceability in operations. A total of 108 real files, corresponding to 54 events with documented failures, were analyzed and used to train and validate the recurrent convolutional neural network model. The system achieved an accuracy of 82.92% in error detection, along with reductions of 42.7% in the average operational response time and 38.5% in failure frequency. In addition to standardizing procedures, the interface demonstrated adaptability to different substation topologies and configurations, establishing itself as a scalable, secure, and efficient alternative for assisted operation environments. The results suggest that the proposed solution contributes to reducing inconsistencies, increasing decision-making autonomy, and strengthening operational safety in the power sector.
Predição do não-comparecimento de pacientes em uma clínica de diagnóstico por imagem usando aprendizado de máquina
(Universidade Federal de Goiás, 2025-07-04) Oliveira, Carlos Eduardo Gonçalves de; Itikawa, Emerson Nobuyuki; http://lattes.cnpq.br/5014788656693729; Itikawa, Emerson Nobuyuki; Cardoso, Alisson Assis; Salazar, Aldo André Diaz
The objective of this work is to apply and analyze the performance of Machine Learning models for predicting patient no-shows at a diagnostic imaging clinic, using data from 2015 to 2023 from two units of Clínica Radiológica de Anápolis (CRA), in Anápolis, Goiás, Brazil. The relevance of this study is based on the possibility of building a final application and on the recurrence and negative impact of patient no-shows in health centers, requiring methods to optimize the use of clinical resources and reduce financial and efficiency losses. The procedure modalities considered in this work were Magnetic Resonance Imaging, Computed Tomography, consultations, and Ultrasound. The collected data included patient age, patient gender, patient no-show history, scheduling details (date and time), procedure type, distance from the patient’s registered address to the clinic, among others. The tested models, Logistic Regression, Multilayer Perceptron, XGBoost, LightGBM, and CatBoost, underwent hyperparameter tuning and probability threshold adjustment based on the Precision-Recall curve area and a customized "Cost" metric. The SHAP framework was used for interpreting the predictions. Comparisons with the literature indicated the agreement of the obtained results and the potential of the methods in this work to serve as a no-show prediction solution for optimizing tasks such as overbooking. The analysis using the SHAP framework, in turn, was able to highlight the most influential variables in the probability of no-show for different procedure modalities, reinforcing the utility of this method for identifying actionable variables.
HibridNet: Rede Neural Convolucional (CNN) Híbrida para classificação de doenças em folhas de bananeira
(Universidade Federal de Goiás, 2025-06-17) Silva, Vinicios Matheus Oliveira da; Nogueira, Tiago do Carmo; http://lattes.cnpq.br/3522572013466053; Cruz Junior, Gelson da; http://lattes.cnpq.br/4370555454162131; Cruz Junior, Gelson da; http://lattes.cnpq.br/4370555454162131; Nogueira, Tiago do Carmo; http://lattes.cnpq.br/3522572013466053; Vinhal, Cássio Dener Noronha; http://lattes.cnpq.br/9791117638583664; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867
Banana cultivation faces significant challenges due to foliar diseases such as black sigatoka, yellow sigatoka, Panama disease, and cordona, which reduce productivity and increase production costs. Traditional disease detection methods are often limited in accuracy and scalability, highlighting the need for automated solutions. This study proposes the implementation and evaluation of convolutional neural networks (CNNs) based on LeNet and Vision Transformer (ViT) architectures. Additionally, a novel hybrid model, named HibridNet, is introduced by combining the strengths of both architectures. Experimental results show that HibridNet achieves higher accuracy compared to individual ViT and LeNet models. The proposed hybrid approach demonstrates significant potential to support disease management in banana cultivation, improving productivity and reducing operational costs
Determinação da velocidade de impacto a partir da análise da energia de deformação acumulada
(Universidade Federal de Goiás, 2025-04-01) Arantes Neto, Antonio Pereira; ; Cruz Junior, Gelson da; http://lattes.cnpq.br/4370555454162131; Cruz Junior, Gelson da; http://lattes.cnpq.br/4370555454162131; Calixto, Wesley Pacheco; http://lattes.cnpq.br/9073478192027867; Nogueira, Tiago do Carmo; http://lattes.cnpq.br/3522572013466053
Vehicle accident reconstruction is an essential discipline aimed at under- standing critical details of collisions, such as vehicle speed at the moment of impact. The primary goal of this study is to determine vehicle collision speed through the analysis of frontal deformation observed post-impact. The methodology integrates modern techniques, including photogrammetry to create detailed three-dimensional damage models, LiDAR-based 3D scanning, and numerical simulations through finite element modeling (FEM). The obtained results demonstrated reduced error margins in speed estimation, confirming the accuracy and effectiveness of the adopted methods. Qualitatively, the proposed methodology proved robust and reliable across different real-world collision scenarios, highlighting its significance in forensic analysis and automotive safety. The study concludes that the developed method represents a reliable technical tool for forensic investigations, significantly advancing automotive safety and the scientific understanding of vehicle collision phenomena.
Modelagem e Identificação de Sistemas Dinâmicos com Redes Recorrentes Profundas aplicados em Pedais de Distorção e Robôs Móveis
(Universidade Federal de Goiás, 2025-02-21) Correia, Murilo Guimarães; Cardoso, Alisson Assis; http://lattes.cnpq.br/8216536516894987; Cardoso, Alisson Assis; Sousa, Marcos Antonio de; Cruz Júnior, Gélson da
This work addresses the Identification and Modeling of Dynamic Systems Using Deep Neural Networks, applying Multilayer Perceptron (MLP), Recurrent Neural Networks (LSTM and GRU), and the extended xLSTM (Extended Long Short-Term Memory) architecture in two distinct projects: the modeling of electric guitar signal distortion and the dynamic behavior of the TurtleBot 3 mobile robot. In the first project, MLP and LSTM were used to model the distortion applied to a guitar audio signal, simulating the effect of different resistances. The results were analyzed using the Cumulative Distribution Function (CDF), the Kolmogorov-Smirnov (KS) test, and the Mean Squared Error (MSE). LSTM demonstrated a strong ability to capture temporal dependencies in the audio signal, while MLP effectively modeled the relationship between inputs and outputs. The results showed low MSE values and a good match in the KS test, demonstrating that both architectures are effective in modeling audio distortions. In the second project, the modeling of the dynamic behavior of the TurtleBot 3 was carried out using four models: MLP, GRU, LSTM, and xLSTM. The goal was to predict its trajectories and velocities based on simulated data. The evaluation metrics included the Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Median Absolute Error (MdAE), and the coefficient of determination R2 . Among the evaluated models, xLSTM achieved the best results across almost all metrics, standing out as the most accurate architecture for modeling the robot’s dynamic system. However, MLP also demonstrated excellent performance, surpassing the GRU and LSTM models despite being a simpler approach. This finding reinforces the efficiency of MLP in capturing nonlinear relationships between inputs and outputs, making it a competitive alternative for dynamic system modeling. In conclusion, the results obtained in both projects highlight the capability of deep neural networks, including MLP, LSTM, GRU, and xLSTM, in modeling complex dynamic systems. The metric analysis confirms the robustness of the proposed methodology, making it a promising approach for Digital Twin applications, with potential for real-time monitoring and control.
Aplicação de algoritmos evolutivos na alocação de recursos energéticos distribuídos
(Universidade Federal de Goiás, 2025-01-24) Mendes, Saymon Fonseca Santos; Negrete, Lina Paola Garcés; http://lattes.cnpq.br/3707701912481754; Negrete, Lina Paola Garcés; Mello, Igoor Morro; Kopcak, Igor; Brigatto, Gelson Antônio Andrêa
This study investigates the integration of distributed generation and energy storage systems in medium-voltage electrical networks. The proposed approach involves the development of evolutionary algorithms to optimally determine connection points for various loads and provide the optimal active power injections from multiple generating units at different network locations. The allocation of Distributed Energy Resources (DERs) in Power Systems presents an intrinsic balance between benefits and challenges. Among the benefits, significant reductions in electrical losses, attraction of investments in the energy sector, diversification and expansion of the energy matrix with a focus on renewable sources, and positive socioeconomic impacts, such as job creation, stand out. However, challenges include difficulties related to system usage tariffs, the need for revisions in operational procedures, uncontrolled voltage profile elevation in cases of high penetration of arbitrarily allocated DERs, increased short-circuit levels, and compromised performance of protection systems. The optimization model formulation adopted the minimization of total active power losses in the system as its objective function. The study considered solar-based distributed generation and battery storage systems for simulations. Network operating conditions were analyzed using power flow studies of the base case. The developed algorithms were applied to two test systems: the IEEE 34-bus system and the Real Feeder System PD004. The results highlight the efficiency, robustness, and rapid convergence of the Hybrid Genetic Algorithm (HGA) compared to other implemented approaches, demonstrating its superiority in solving the problem. The developed computational tool shows significant practical potential for application in electric utility companies, serving as a strategic tool for distribution network planning. Utilizing the algorithm’s results enables the optimized connection of generating units and loads while respecting established generation limits.
Avaliação do desempenho de funções de proteção aplicadas em Reatores Shunt
(Universidade Federal de Goiás, 2024-11-21) Ferreira, Diogo Guilherme; Ferreira, Diogo Guilherme; http://lattes.cnpq.br/0347498831010752; Almeida, Maria Leonor Silva de; http://lattes.cnpq.br/7955955842189669; Almeida, Maria Leonor Silva de; Peres, LarissaMarques; Manassero Junior, Giovanni
The increase in the length of transport lines with high voltage levels has been distributed in countries with large territorial extensions, resulting in the demand for resources destined to control these levels. One of the possible adjustments attributed to voltage regulation is the use of shunt reactors in the lines. This equipment makes voltage regulation possible because, by absorbing excess capacitive reactives in the line, they maintain voltage within acceptable ranges for operation. Therefore, given the relevance of shunt reactors for the proper functioning of transmission lines and the electrical system, it is essential to ensure the correct functioning of the protection functions applied to shunt reactors. Therefore, this work presents an investigation into the application of different types of protection functions traditionally employed by manufacturers in reactors, such as restricted earth fault protection (REF), directional protection of negative sequence (32QF and 32QR), differential protection (87) and distance protection (21), which is normally used as a backup. The logic of each specific function aims to guarantee the operation and selectivity of protection in different short circuit situations. Therefore, in order to evaluate the protections, simulations were carried out in the software Alternative Transient Program (ATP), through which a 500 kV/60 Hz transmission line 400 km long and 60% shunt compensation was modeled. In the different simulations, different scenarios of internal short circuits in the reactor were evaluated, covering both the type of turn-to-ground faults and turn faults. Furthermore, we will understand the best functioning of protection schemes, external faults were applied to the equipment. Furthermore, we will understand the best functioning of protection schemes, external faults were applied to the equipment. Furthermore, a comparative evaluation was carried out between the performance of the different protection functions investigated, varying several criteria such as the type of fault, number of coils involved in the short circuit and the values of the leakage factors. Furthermore, transient analyzes and parametric sensitivity analyzes were performed. Based on all the results obtained, and comparing different protection schemes implemented in available commercial relays, it appears that the 21 and 32Q functions stand out, as similar and better performances were highlighted in relation to the other functions evaluated. In general, the 21 and 32Q functions work for all turn-to-ground faults and turn faults, regardless of the number of turns involved and the value of the leakage factor. Comparatively, the distance function proved to be faster, with an operating time of less than one cycle. Thus, it is stated that the distance function can also be used as main protection, together with other traditionally used functions, in order to increase the performance of the reactor protection scheme.
Alocação espacial de geração distribuída em redes de distribuição de energia elétrica utilizando um algoritmo genético híbrido
(Universidade Federal de Goiás, 2024-07-08) Silva, Carlos Henrique dos Santos; Trujillo, Joel David Melo; http://lattes.cnpq.br/4396532007704107; Negrete, Lina Paola Garces; http://lattes.cnpq.br/3707701912481754; Negrete, Lina Paola Garces; Trujillo, Joel David Melo; Flórez, Hugo Andrés Ruiz; Corrêa, Henrique Pires
This work presents a methodology for optimizing the allocation of Distributed Generation (DG) resources in electrical distribution systems, utilizing evolutionary strategies and considering georeferenced spatial aspects, with an exclusive focus on photovoltaic generation. The methodology is divided into three main stages: definition of the original system, determination of network operating conditions, and application of optimization strategies based on technical criteria, with an emphasis on cost and available area limitations, which are the differentiating factors of this study. In Stage I, the target Electrical Distribution System and its main spatial, technical, and budgetary limitations are defined, and simulations are performed to evaluate the system's performance without the allocation of DG using OpenDSS software. The spatial limitations refer to areas made available by the utility for the construction of photovoltaic plants. In Stage II, reference generation values for each bus are determined, considering the generation value that best reduces the system's total losses when only one DG is allocated, as well as the available area and budget restrictions. In Stage III, two specific optimization strategies are applied and discussed - Hybrid Evolutionary Strategy and Hybrid Genetic Algorithm - to determine the DG allocations that best meet the specified objectives. To apply the proposed methodology, case studies were conducted on two different systems. The first case study used the IEEE 34-Bus System, a reference feeder representing larger and more complex real scenarios. The second case study used a real feeder, with information extracted from the Geographic Database of the Distribution Company (BDGD) provided by ANEEL with data from 2023 and geographically represented using QGIS software. In both studies, the results analysis demonstrated the effectiveness of the proposed optimization strategies, showing significant reductions in system losses and improvements in voltage profile. The results of the case studies show that the Hybrid Genetic Algorithm performed slightly better than the Hybrid Evolutionary Strategy. In the first case study, voltage gains of 7.39%, loss reduction of 14.48%, and load decrease of 9.55% were observed. In the second case study, the allocation of five DG systems resulted in a voltage gain of 1.75%, loss reduction of 14.08%, and load decrease of 8.90%, using 98.47% of the available budget and respecting area limitations. The work concludes that the application of optimization strategies, considering spatial and cost aspects, allows for a satisfactory and efficient solution to the DG allocation problem. The step-structured methodology, together with the adopted optimization strategies, ensures a systematic and rigorous approach, facilitating the replication and validation of results. This methodology contributes to the understanding and implementation of optimal DG allocation strategies, supporting energy utilities in the efficient and sustainable management of their distribution systems.
Sistema de comunicação alternativa para pessoas com distúrbios neuromotores severos usando redes neurais artificiais
(Universidade Federal de Goiás, 2024-12-15) Floriano, Carolina de Souza; Silva, Adson Rocha; Brito, Leonardo da Cunha; http://lattes.cnpq.br/6660680440182900; Brito, Leonardo da Cunha; Rocha, Adson Silva; Gomide, Renato de Sousa
Communication difficulties are frequent for many people with severe motor disabilities, making it difficult for them to interact with their families, caregivers and society in general. Augmentative and Alternative Communication (AAC) then aims to compensate for the communication deficit of these people, providing the individual with a better quality of life. However, these individuals with severe neuromotor disorders who have severe movement restrictions find great challenges in the use of several current assistive technologies. In this context, the objective of this research is to present an Alternative Communication System based on Artificial Neural Networks with a user-centered approach and their needs, for use by this public. The input and signal processing are carried out by reading facial landmark points, using the MediaPipe FaceMesh library. The development of the gesture/facial expression classifier is performed through the implementation and comparison of two different models: a Convolutional Neural Network (CNN) model and a Recurrent Neural Network model using Long Short-Term Memory (LSTM) units and dense layers. Dynamic challenges were implemented to conduct a more in-depth analysis of the models’ performance in various contexts, varying parameters such as the quantity of samples and the inclusion of similar gestures. Real-time overall results indicate a consistent performance of the proposed system, suggesting that, in both approaches, the Convolutional Neural Network (CNN) stands out significantly compared to the Long Short-Term Memory Recurrent Neural Network (LSTM) in gesture recognition.
Sistema de comunicação alternativa para pessoas com distúrbios neuromotores severos usando redes neurais artificiais
(Universidade Federal de Goiás, 2023-12-15) Floriano, Carolina Souza; Silva, Adson Rocha; http://lattes.cnpq.br/4116708456419800; Brito, Leonardo da Cunha; http://lattes.cnpq.br/6660680440182900; Brito, Leonardo da Cunha; Gomide, Renato de Sousa; Rocha, Adson Silva
Communication difficulties are frequent for many people with severe motor disabilities, making it difficult for them to interact with their families, caregivers and society in general. Augmentative and Alternative Communication (AAC) then aims to compensate for the communication deficit of these people, providing the individual with a better quality of life. However, these individuals with severe neuromotor disorders who have severe movement restrictions find great challenges in the use of several current assistive technologies. In this context, the objective of this research is to present an Alternative Communication System based on Artificial Neural Networks with a user-centered approach and their needs, for use by this public. The input and signal processing are carried out by reading facial landmark points, using the MediaPipe FaceMesh library. The development of the gesture/facial expression classifier is performed through the implementation and comparison of two different models: a Convolutional Neural Network (CNN) model and a Recurrent Neural Network model using Long Short-Term Memory (LSTM) units and dense layers. Dynamic challenges were implemented to conduct a more in-depth analysis of the models’ performance in various contexts, varying parameters such as the quantity of samples and the inclusion of similar gestures. Real-time overall results indicate a consistent performance of the proposed system, suggesting that, in both approaches, the Convolutional Neural Network (CNN) stands out significantly compared to the Long Short-Term Memory Recurrent Neural Network (LSTM) in gesture recognition.
Minimização da ondulação de torque em motores a relutância variável por meio de correntes de fase de referência otimizadas por algoritmo genético
(Universidade Federal de Goiás, 2023-12-18) Soares, Israel Rodrigues; Paula, Geyverson Teixeira de; http://lattes.cnpq.br/0140145167826333; Paula, Geyverson Teixeira de; Oliveira, Eduardo Sylvestre Lopes de; Almeida, Thales Eugenio Portes de
This work proposes an innovative control strategy for the Switched Reluctance Motor with the aim of minimizing torque ripple. The strategy is based on an algorithm for generating current profiles that prioritize the smooth commutation mode of the asymmetric half-bridge converter. This algorithm employs genetic algorithms to calculate these profiles through simulations in a finite element model developed based on a 6x4 Switched Reluctance Motor from the Laboratório de Ensaios de Pequenos Motores at the Universidade Federal de Goiás. To enhance the adaptability of the proposed control, the addition of a compensation derived from torque error to these profiles has been suggested. Simulations compared the Proposed Control with Direct Instantaneous Torque Control and the Proposed Control without the addition of compensation under various operating conditions. The results highlight significant average reductions in metrics used to evaluate torque ripple. In the Torque Ripple metric, there was an average reduction of 16.02% compared to Direct Instantaneous Torque Control and 13.14% compared to the Proposed Control without compensation. As for the Torque Ripple Factor metric, this reduction was 15.34% and 15.96%, respectively. The study concludes by affirming the good performance of the generated current profiles, demonstrating that the inclusion of compensation derived from torque error in these profiles was crucial for the low levels of torque ripple achieved by the proposed control technique.
Algoritmos bioinspirados aplicados ao problema de alocação de geração distribuída
(Universidade Federal de Goiás, 2023-02-02) Santos, Josephy Dias; Garcés Negrete, Lina Paola; http://lattes.cnpq.br/3707701912481754; Garcés Negrete, Lina Paola; Belati, Edmarcio Antonio; López Lezama, Jesus Maria; Brito, Leonardo da Cunha
This work presents the performance comparison of different meta-heuristics, two classic and one modern. The implemented optimization algorithms aim to solve the distributed generation allocation problem in electricity distribution networks widely known in the literature. The study confronts the following computational techniques applied in algorithms classified as bioinspired: the Chu-Beasley Genetic Algorithm (AGCB), the Symbiotic Organisms Search (SOS) and the Coronavirus Optimization Algorithm (CVOA). The allocation of DG units in the Electric Power System gives the system advantages and disadvantages. Among the advantages we can mention: reduction of power losses, expansion of investments in the electrical sector, expansion and diversification of the electrical matrix, mostly, use of clean energy and indirect benefits such as job creation. Among the disadvantages are difficulties in charging for the use of the electrical system, possible incidence of undue taxes, need to change operating procedures, indiscriminate elevation of the voltage profile if the penetration factor is high and the allocation of DG is random, increase in short circuit levels, failures in the protection operation, among others. The network operating conditions are verified through the forward and reverse sweep method, specifically using the Power Sum Method. The objective function, in the optimization model for the allocation of distributed generation, aims to minimize the total losses of active power in the system. For the implementations, the allocation of modules (100, 200 and 500kW) of distributed generation is considered, with the number of these modules limited by the penetration factor of each network. The specialized algorithms are tested on four electrical systems: 10, 34, 70 and 126 buses. The results obtained show the rapid convergence and robustness of the AGCB of the implemented algorithms, the same cannot be said about SOS, which had an intermediate performance. The CVOA, as an unprecedented contribution in this work, presented a lower performance than expected, largely due to its nature and architecture of the proposed modeling.
Algoritmos de inteligência computacional aplicados à otimização de sistemas de controle em acionamentos elétricos
(Universidade Federal de Goiás, 2023-03-29) Santos, Guilherme Fernandes dos; Silva, Wander Gonçalves da; http://lattes.cnpq.br/4669127331497967; Cruz Junior, Gelson da; http://lattes.cnpq.br/4370555454162131; Cruz Junior, Gelson da; Silva, Wander Gonçalves da; Pickert, Volker; Oliveira, Marco Antonio Assfalk de; Cardoso, Alisson Assis
This work presents the use of different computational intelligence methods applied to the tuning of a set of PI controllers for a DC motor drive with speed and position control. For position control, three closed control loops are used: armature current, speed and position. For speed control, only the armature current and speed loops are considered. In both cases, the outputs of the PI armature current and speed regulators are limited to the rated armature voltage and current, respectively. Then, it is possible to use higher gains for the controllers, what makes the system to respond faster. However, the windup phenomenon can arise. To avoid it, anti-windup circuits are also used and therefore, the system becomes non-linear. Because of this, an optimum tuning of the controllers may become a difficult task. In order to explore different possibilities, firstly, the speed and position control problems are formulated so that only one objective is minimised. Within this single-objective optimisation context, the PSO and SA algorithms are used to tune the controller parameters, them, the capability of each one is investigated when compared to each other. Multiobjective formulations are also explored to address three objectives simultaneously. In this part of the work, the multi-objective evolutionary algorithms NSGA-II and SPEA2 are used. All algorithms were implemented in MATLAB and the electric drive models were developed in the SIMULINK environment. Simulation results are presented showing that for the single-objective formulation, for both, the for speed and position control problems, the PSO algorithm outperformed the SA. For the multi-objective formulation, the SPEA2 algorithm presented better characteristics with respect to the Spread quality indicator in the only, when compared to NSGA-II. Furthermore, it’s shown to outperform the NSGA-II with respect to the Hypervolume indicator within the position control problem. A series of tests were carried out by varying the values of the main parameters setting for each algorithm. However, in most cases, no statistically significant advantage was observed. In general, the results presented demonstrate the ability of the algorithms to find optimal tuning for the controllers, either for the single-objective or the multiple and conflicting objectives problem.
Desempenho de métodos clássicos e meta-heurísticas aplicadas à sintonia de controladores PID sob o aspecto do fator de incontrolabilidade
(Universidade Federal de Goiás, 2023-03-30) Andrade Filho, Alfredo de Paulo; Cardoso, Alisson Assis; http://lattes.cnpq.br/8216536516894987; Cruz Júnior, Gelson da; http://lattes.cnpq.br/4370555454162131; Cruz Júnior, Gelson da; Avelar, Henrique José; Teixeira, Edilberto Pereira
The PID (Proportional Integral Derivative) is one of the most used in industry, and has been employed around the world for decades in industrial control systems. Its popularity can be attributed in part, to its robust performance, in a wide range of operating conditions, and their functional simplicity, which allows engineers to operate them in a simple and direct way. However, tuning such controllers effectively, represents a challenge to control engineers, since it directly affects the efficiency of the system. In the context of modern industry, the setback is even greater, due to the complexity of the processes involved, in which the presence of the Dead Time and the Time Constant of the system, have a great impact on its response. Tuning of PID controllers aims to satisfy some specifications imposed on the transient and steadystate characteristics of the response of the system being controlled. Generally, these specifications fall on the values of Rise Time (Tr), Settlement Time (Ts), Maximum Value overshoot (Mp) and a maximum acceptable Stationary Error (Ess) value. Within the scope of control problems that offer great complexity, either because of the impossibility of previously knowing the transfer function of the system, or due to its multivariable characteristics, several researches have been carried out within the scope of optimizing the tuning of PID controllers using meta-heuristics. In many situations, the system PID cannot be accurately modeled, and still malfunctions in the presence of noise or external disturbances due to the adjustment problem of its parameters that are challenging in practice, as they depend on characteristics of the environment. In this work, we sought to evaluate the performance of the PID controller under the aspect of Uncontrollability Factor when tuned by Genetic Algorithm heuristics (GA) and Particle Swarm (PSO), which have been shown to be more effective in compared to classical tuning methods. Six case studies were evaluated, with varied Uncontrollability Factors.
Influência da taxa de subida da corrente na regularidade do processo MIG/MAG por curto-circuito
(Universidade Federal de Goiás, 2023-02-28) Niz, Weslei Rodrigues; Ferreira Filho, Demostenes; http://lattes.cnpq.br/2814935331164390; Souza, Daniel; http://lattes.cnpq.br/2132578584168482; Souza, Daniel; Ferreira Filho, Demostenes; Figueiredo, Kléber Mendes de; Rossi, Marcelo Lemos
The present work investigates the effects of the current rise rate (power source inductance adjustment) on the regularity of the MIG/MAG welding process operating in short-circuit mode. For this, initially the adjustment of the inductive effect of the power source was varied, keeping the regulation voltage and the supply speed constant. Using a regularity index calculated from the current voltage signals, the best fit of the inductive effect was determined. Then, the value of the inductive effect (average rate of current rise) was fixed and the welding voltage was varied with the feed speed kept constant. All tests were performed with 3 shielding gases with different percentages of CO2 (100% Ar, Ar+25%CO2, 100% CO2). As a result of the data analysis, it was possible to find the range of values for the rate of rise of welding current and voltage where the process operates with greater regularity. Preliminary results show that shielding gas influences the correlation between welding voltage, current rise rate and process regularity. After finding the range of values where the process is more regular, experiments were carried out to try to relate regularity with the generation of spatter, the generation of spatter had a tendency contrary to that of regularity, the best performance was with 100% CO2, followed by 100% Ar and the worst mass yield were Ar + 25% CO2.
Caracterização do modelo ZIP na análise de sistemas de distribuição de energia elétrica
(Universidade Federal de Goiás, 2022-10-27) Silva, Arthur Oliveira; Garcés Negrete, Lina Paola; http://lattes.cnpq.br/3707701912481754; Garcés Negrete, Lina Paola; Sousa, Thales; Oliveira, Marcelo Escobar de; Brigatto, Gelson Antônio Andrêa
This paper presents a proposal for optimizing the ZIP model coefficients for modeling the loads present in electrical power distribution systems. Initially, the basic power flow theory is presented and, through an example using the theoretical IEEE 118-bus network, it is shown that the technical losses and voltage profiles obtained from the power flow solution are impacted when three ZIP model scenarios are considered. From the results of the example, it can be seen that this approach should not be neglected in system planning and operation studies, since the non-consideration of the ZIP model can bring impacts in the network diagnosis and, consequently, lead to the direction of imprudent investments under the view of the minimum global cost, as recommended by the National Electric Energy Agency (ANEEL). Thus, this paper proposes a mathematical optimization model that allows obtaining the characteristic coefficients of the ZIP model of the loads, which has as an objective the minimization of the difference between the measured actual current and the calculated one through the solution of the load flow problem. This model will be solved using a genetic algorithm, where each consumer class has its own ZIP model coefficients, separated in active and reactive power portions. Two case studies in real power distribution networks are performed to demonstrate the impact of the optimal ZIP model coefficients on the technical losses (MWh) of the feeders. Finally, the results of the work allow utilities to perform network diagnostics in a more assertive way, since each network can have a ZIP model that best fits the behavior of its loads.

Navegar

Submissões Recentes