Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Multi-feature fusion framework for enhanced image deduplication accuracy using adaptive deep learning
Get PDFImage deduplication is a critical task in domains such as digital asset management, content-based image retrieval (CBIR), and data storage optimization. This paper presents a novel method for improving deduplication accuracy by integrating multiple feature types. A comprehensive framework is proposed that combines visual, semantic, and structural image elements. The system employs deep learning architectures, including convolutional neural networks (CNNs) and transformers, to extract high-level features, which are fused through an adaptive weighting mechanism that dynamically adjusts based on image content. Experimental results across diverse datasets demonstrate that the proposed multi-feature fusion approach significantly outperforms traditional single-feature methods, achieving an average improvement of 15% in deduplication accuracy. By overcoming limitations in handling complex visual similarities, this study introduces a more robust and efficient solution for image deduplication
Multi-attribute based optimal location and sizing of solar power plant in radial distribution system
Get PDFAdvancements in renewable energy sources (RES) have significantly increased power generation and reduced emissions. Optimally integrating RES into distribution systems can minimize power losses, emissions, and enhance voltage profile and stability. Therefore, determining the optimal location and size of RES is crucial for their effective integration. This paper presents a novel approach for identifying the optimal location and size of a solar power plant (SPP) in a distribution system, considering system power losses, voltage profile, voltage stability, and emissions simultaneously. A simple yet effective methodology combining repeated load flow and fuzzy systems is proposed. Repeated load flow is used to calculate the relevant attributes, while fuzzy decision-making is employed to determine the optimal solution. The effectiveness of the proposed method is demonstrated through its application to the IEEE-33 bus system. The results illustrate that integrating a SPP at the optimal location and size can significantly reduce power losses and emissions while improving voltage profile and stability
Few-shot brain tumor classification: meta- vs metric-learning comparison
Get PDFMedical imaging requires accurate brain tumor recognition because precise classification is essential for early diagnosis and effective treatment planning. A major challenge in medical applications is that deep learning models typically require extensive amounts of labeled data to perform well. To address this, this research evaluates three few-shot learning (FSL) approaches-prototypical networks, Siamese networks, and model-agnostic meta-learning (MAML)-for brain tumor classification using the Figshare brain tumor dataset. The results show that prototypical networks consistently outperform the other approaches, achieving 89.07% accuracy (95% CI: 88.12–89.96%), 88.73% precision, and 88.67% recall, making them the optimal solution for this task. Siamese networks achieve 83.73% accuracy (95% CI: 82.64–84.76%), while MAML demonstrates significantly reduced performance, with 43.70% accuracy (95% CI: 42.10–45.22%). This study demonstrates that FSL can be applied effectively for medical image classification, with prototypical networks achieving the best performance in brain tumor detection. The inclusion of confidence intervals further validates the robustness and reliability of the results. Future research will focus on improving feature representation and exploring hybrid approaches to better handle rare tumor classes, thereby enhancing the clinical applicability of FSL models
Research on optimal design of surface permanent magnet synchronous generator
Get PDFThe more fossil energy is used, the less this energy source will become because it is not an infinite energy and it pollutes the environment, so there is a need for solutions with new and infinite energy sources such as wind energy. This paper designs and focuses on optimizing a floating magnet synchronous generator (SG) for a wind power generation system using finite element analysis (FEA) with ANSYS Maxwell software. This generator is compared with other types of generators such as squirrel cage induction generator (SCIG), wound rotor induction generator (WRIG), SG, doubly-fed induction generator (DFIG), and switched reluctance generator (SRG). Throughout the analysis and design process, the paper emphasizes the significant benefits of surface-mounted permanent magnet (SPM) motors in increasing efficiency and reliability while reducing supply costs. The research results of the paper aim to demonstrate that SPM can meet the needs of high efficiency and low cost in the industrial and civil fields. The results of this study by the authors will provide new contributions to serve as a basis for the design, manufacture, calculation and control of Halbach permanent magnet (Halbach PM) electric machines based on optimization techniques such as genetic algorithms (artificial intelligence) and sustainable optimization (for electrical equipment)
Design of internet of things-integrated programmable logic controller for demonstrating automated sorting systems
Get PDFThis project presents an automated workpiece sorting demonstration system controlled by a programmable logic controller (PLC) and a touch screen interface. The system integrates an internet of things (IoT) gateway that communicates with the PLC via Modbus remote terminal unit (RTU) over RS-485, allowing the transfer of digital data. This data is processed using JavaScript within the Node-RED platform to manage machine operations and display the operational status. The system supports both manual control and IoT-based management, enabling the sorting of cylindrical workpieces to designated areas. Metal and non-metal detection is achieved using capacitive and inductive sensors, respectively, which inform a stepper motor to manipulate the workpieces via a gripper pneumatic to the specified locations. Test results indicate a high detection capability of the sensors: the capacitive sensors achieved a 95% detection rate over 100 trials, while the inductive sensors recorded a 97% detection rate. Furthermore, the precision of placing workpieces at the target locations was 92% across 100 attempts. This system showcases an effective combination of automation and IoT technologies, improving efficiency in workpiece sorting processes
Route splitting and adaptive mutation in genetic algorithms for the capacitated vehicle routing problem
Get PDFThe capacitated vehicle routing problem (CVRP), where vehicle capacity constraints limit the load carried per route for multiple vehicles, is addressed using an optimized genetic algorithm (GA) framework. This work focuses on finding the best configuration of GA by systematically evaluating 12 distinct GA variants, differing in adaptive mutation rates and route-splitting strategies. The framework integrates adaptive mutation rates and novel route-splitting approaches—greedy, dynamic programming (DP), and heuristic—to enhance computational efficiency and solution quality. Experiments on six CVRP instances of varying complexity, encompassing differences in problem size, vehicle capacity, and geographical distribution, demonstrate the heuristic approach’s effectiveness. It achieves solutions within 2%–5% of the optimal cost of DP while being 3–4 times faster. Adaptive techniques reduce costs by up to 20% compared to standard GAs and heuristics. The framework’s scalability is evident in large-scale instances such as the 200-customer case, where the heuristic method balances cost (414.17) and computation time (0.003 seconds). The developed software is openly available at GitHub, providing a robust tool for addressing practical logistics challenges
Performance analysis of a proximity-coupled triangular slot microstrip patch antenna for ship radar applications
Get PDFMicrostrip patch antennas are extensively utilized in modern communication systems because of their small size and simple fabrication process. Among the different patch geometries, triangular patches offer size reduction compared to their rectangular and circular counterparts, making them suitable for space-constrained applications. This study focuses on the design and analysis of an equilateral triangular microstrip antenna (ETMSA) using proximity coupled feed with a triangular slot, targeting optimal performance at 2.2 GHz. The antenna is constructed using two FR4 substrates of identical permittivity but different thicknesses (h1 and h2), with a 50-ohm microstrip line feed positioned between them. The aim is to determine the optimal values of patch surface area, slot dimensions, and upper substrate thickness to achieve maximum bandwidth, minimal return loss, and ideal voltage standing wave ratio (VSWR). Simulations and measurements confirm that the antenna achieves a 120 MHz bandwidth achieving a return loss of –42 dB and a VSWR of 1.03, demonstrating excellent agreement. These results confirm the antenna's effectiveness for fixed-beam applications in wireless communication systems, highlighting its potential for efficient and compact antenna solutions
CODE NET: COVID-19 segmentation and detection via deep learning based networks
Get PDFHumans with COVID-19 have an infectious condition that affects the respiratory system. In addition to more serious conditions, headaches may be fatal for those who have the disease. Our difficulty with COVID-19 detection stems from the unreliability of computed tomography (CT) and magnetic resonance imaging (MRI) scans in identifying lung abnormalities. COVID-19 detection is a time-consuming process. In this research, a novel CODE NET model is proposed for the detection of COVID-19 virus from the gathered lung chest X-ray (CXR) images. The images are pre-processed utilizing an adaptive trilateral filter to improve the quality of the images. A reverse edge attention network (RE-Net) uses enhanced images to segment the CXR images for accurate virus detection. The segmented images are fed into a Link Net to extract relevant features and classify the COVID-19 cases. The classified cases are fed into the Grad-CAM model to generate heat maps for accurately detecting the virus. According to the result, the proposed model attains 99.75% of accuracy rate for the COVID-19 detection. The proposed CODE NET enhances the overall accuracy by 1.78%, 1.51%, and 2.20% over combined domain features-random forest (CDF-RF), Bayes-SqueezeNet, and bidirectional long short-term memory (Bi-LSTM) respectively
Advanced trajectory tracking control for wheeled mobile robots under actuator faults and slippage
Get PDFTrajectory tracking control for wheeled mobile robots (WMRs) faces significant challenges in real-world applications due to actuator faults, longitudinal and lateral slippage. This study proposes an innovative dual-loop control structure combining adaptive sliding mode control (ASMC) and backstepping control (BC), supported by robust fault observers, to address these challenges. The dynamic loop employs ASMC to handle model uncertainties and disturbances, while the kinematic loop integrates BC with fault information provided by the observers, enabling real-time error compensation. Simulation results show that the proposed method significantly reduces tracking errors and improves stabilization time compared to traditional SMC and ASMC controllers. The system exhibits enhanced fault tolerance and disturbance rejection, maintaining stability under both normal and faulty conditions. The effectiveness of this approach is demonstrated through simulations and theoretical analysis, ensuring system stability using Lyapunov stability theory. The proposed method enhances robustness, adaptability, and stability of WMRs, contributing significantly to the field of mobile robotics under adverse conditions
Multiword target-independent transformer-based model for financial sentiment analysis in colloquial Cantonese
Get PDFTokenization process decomposes a multi-word-span instrument name into several tokens and the transformer attention mechanism handles each token individually, thus hindering the treatment of the related tokens as a single entity. The existence of multiple instruments in a single message further exaggerates the complications and results in low predictive performance. This study proposed the use of sequentially tagged target-independent sentinel tokens to encapsulate multiword instrument aspects for natural language inference model fine-tuning. The encapsulation not only facilitated the attention mechanism to handle an instrument name as a single entity but also enabled the model to handle unseen instruments effectively. Our empirical analysis was based on 5,178 manually annotated instrument–sentiment pairs originated from finance discussion board messages that addressed sentiments of one to four instruments in a single post. The proposed approach consistently outperformed the direct bidirectional encoder representations from transformers (BERT) based approach in terms of recall, precision, and F1-score when handling financial commentaries written in colloquial Cantonese. This study demonstrated the potential benefits of target-independent sentinel token encapsulation for natural language inference. The underlying logic of multiword target-independent encapsulation was expected to hold for other languages, including Chinese, Japanese, and Thai