Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Anomaly detection in quadcopter flight: harnessing frequency domain analysis and barnacle mating optimization
Get PDFEnsuring the safety and efficiency of unmanned aerial vehicles (UAVs) requires effective fault detection and identification (FDI). Traditional multi-stage FDI methods, particularly those using residual detection layers, increase complexity and computational cost, limiting real-time applications. This study proposes a single-stage anomaly detection framework integrating barnacle mating optimization (BMO) with discrete cosine transform (DCT) for UAV fault detection. While prior research explored model-based and data-driven FDI, bio-inspired optimization techniques remain underexplored in frequency-domain analysis. This study develops a BMO-based fitness function analyzing 3rd, 5th, and 7th harmonic peaks to detect UAV anomalies. Software-in-the-Loop (SITL) simulations validate the method, achieving a 5-second optimal frame size, mean absolute percentage error (MAPE) of 0.05, and root mean square error (RMSE) of 195.52. The findings confirm that a single-stage detection framework via optimization method and frequency domain analysis is possible, making it viable for real-time UAV applications. This study bridges the gap in bio-inspired UAV fault detection, paving the way for safer and more efficient UAV operations
Recommender systems in real estate: a systematic review
Get PDFThe constant growth of online real estate information has emphasized the need for the creation and improvement of intelligent recommendation systems to help mitigate the difficulties associated with user decision-making. This systematic review, following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines and criteria, investigates current approaches and models used in real estate recommendation systems, with a focus on papers published in 2019 and 2024. The review identifies four main techniques: content-based filtering, collaborative filtering, knowledge-based systems, and hybrid approaches. Key findings indicate a preference for deep learning models, specifically convolutional neural network and long-short term memory (CNN-LSTM) architectures, and highlight the most used property characteristics: price, number of rooms, size, and location. The research addresses several important challenges, including the cold start problem, data sparsity, and the importance of adaptive learning in dynamic markets. Potential future research fields are outlined, with a focus on hybrid model architectures, attention mechanisms, and explainable artificial intelligence (AI). This review provides a comprehensive overview of the field, enabling scholars and practitioners to improve the accuracy and user experience of real estate recommendation systems
Analysis of voltage drop using transformer tap changer and placement of capacitor bank with genetic algorithm
Get PDFThe demand for electrical energy is increasing due to high economic growth and population. The impact is that electrical energy operates excessively to meet the required demand. Unbalanced loads, higher power losses on the line, and voltage drops that are higher than allowed are just a few of the issues that may result from this. Adding tap changers and capacitor banks is one method of improving the voltage profile and power losses. To conduct this study, tap changers and capacitor banks were added to the IEEE 33 bus network system. The value, capacity, and location of the tap changers and capacitor banks in the system were ascertained using the genetic algorithm (GA) approach. According to the simulation results, the voltage profile, which initially had 21 buses outside the IEEE standard limits, may be ideal by installing two tap changers and two capacitor banks. Additionally, reactive power losses decreased from 41.8 kVar to 93.3 kVar, and active power losses decreased from 202.7 kW to 130.7 kW, a decrease of 72 kW
Automated real-time cervical cancer diagnosis using NVIDIA Jetson Nano
Get PDFCervical cancer is a global health concern, making early detection critical for ensuring effective treatment outcomes. Screening technique, the Papanicolaou test (Pap test), has been adopted globally for timely detection. Nevertheless, the process of screening is subjective. The current study aims to advance the development of an automated real time framework for cervical cell analysis for early-stage diagnosis using supervised classification on NVIDIA Jetson Nano platform. Our approach, leveraging adaptive fuzzy k-means (AFKM) clustering and k-means clustering, extracts distinctive features from cervical cell images for accurate classification. Utilizing multilayer perceptron (MLP) and support vector machine (SVM) classifiers, we achieved a classification accuracy of 97%, highlighting the potential of our system for real-time applications in cervical cancer investigation. Validation by two expert pathologists further supports the system’s practical utility
Driver activity recognition using deep learning based on multi-step batch size up
Get PDFThe increasing popularity of electric motorbikes in Indonesia, while promoting sustainable mobility, also raises concerns regarding traffic safety. Given the high incidence of motorcycle-related accidents, there is a critical need for systems capable of monitoring and recognizing driver behavior. This study proposes a driver activity recognition system for electric motorbikes, utilizing an event data recorder (EDR) to capture seven key sensor signals: three-axis acceleration, voltage, current, power, and speed. A custom dataset was constructed using data collected from 10 subjects, each performing five driving activities including forward drive, brake, stop, left turn, and right turn for over three-minute intervals per activity. The classification model is based on a long short-term memory (LSTM) neural network. To optimize training efficiency, a multi-step batch size up (MSBU) strategy was introduced, which accelerates training time by 1.84× compared to a fixed batch size of 32. The best performance was achieved using a segment length of 75 time-steps, yielding an accuracy and macro F1-score of 0.9873. These results demonstrate the effectiveness of the proposed system for real-time driver behavior monitoring and activity recognition in electric motorbike applications
Prediction of asphalt performance based on plastic waste using machine learning
Get PDFThe incorporation of plastic waste into asphalt mixtures offers a promising solution to address the growing environmental concerns while enhancing the performance of road materials. Traditional methods, such as the Marshall test, are costly and time-consuming, thus highlighting the need for more efficient prediction techniques. Machine learning (ML) models, including random forest (RF), extreme gradient boosting (XGBoost), and artificial neural networks (ANN), have shown significant potential in predicting asphalt performance, optimizing material compositions, and reducing the dependence on labor-intensive laboratory tests. Key influencing factors such as bitumen content, plastic size, and temperature have been identified as crucial for improving asphalt properties. This systematic review emphasizes the potential of ML in streamlining the development of plastic-modified asphalt, offering a sustainable and cost-effective approach to road construction. Furthermore, it supports the advancement of green infrastructure and lays the foundation for future innovations in sustainable pavement engineering, contributing both to academic research and practical applications in the construction industry
Design of a machine learning model for predicting credit risk in microfinance using environmental data
Get PDFAgricultural microfinance is a sector that is significantly impacted by climate-related risks, such as temperature fluctuation, soil degradation, and irregular rainfall. These environmental factors have not only impact on crop yield but also results in influencing borrowers’ ability to repay agricultural loans. Traditional credit scoring models lack in predicting due to the complex interplay between environmental and borrower-specific variables. This research study proposes a new predictive machine learning model based on XGBoost for assessing the credit risks in agricultural microfinance. This model utilizes environmental indicators, borrower characteristics, and loan attributes for computing the continuous credit risk score. The model was trained utilizing a real-world dataset of 142,017 loan applications with a 70/30 split. When compared with other traditional models, the results of the model showcases an accuracy of 99%, a recall of 84%, a precision of 89%, and an F1-score of 86%, outperforming traditional algorithms such as logistic regression and decision tree. This model has substantial implications for microfinance organizations. With this model, borrowers can evaluate risk accurately during the loan application stage by utilizing environmental data, resulting in better loan targeting, enhanced financial inclusion, and better risk mitigation for vulnerable farming communities in climate-sensitive regions
Enhanced sensing of liquid levels: multipoint detection using twisted polymer optical fiber in cylindrical configurations
Get PDFRecent interest in water level detection is driven by the need for accurate monitoring in sectors such as agriculture, flood management, and environmental conservation. The measuring of liquid levels and solution concentrations is critical in a variety of industrial sectors. This research is to design, implement, and test the use of plastic optical fiber (POF) as an optical sensor for sensing multi-point liquid levels and densities. The developed device includes a multi-point liquid level sensor that uses refractive index modification of macro bend POFs selectively twisted around a cylindrical column. When the POFs were submerged in various mediums such as pure water, seawater, saltwater, and cooking oil, a set of U-shaped detecting heads was designed to detect changes in liquid levels. The experimental setup included an 850 nm optical light source and a power meter to measure the output power. A comparison of the performance of straight, bent, and twisted POF forms revealed that the twisted and bent POF topologies attained a sensitivity of 30%, exceeding bending-only 21% and twisting-only 12% configurations. In summary, changes in liquid levels resulted in an increase in output power for all liquid media, highlighting the potential of POF as a reliable sensor for sensing liquid levels
The extraction of a brief summary from scientific documents using machine learning methods
Get PDFThis study proposes a machine learning-based approach for automatic summarization of scientific documents using a fine-tuned DistilBART model a lightweight and efficient version of the bidirectional and auto-regressive transformers (BART) architecture. The model was trained on a large corpus of 12,540 scientific articles (2015–2023) collected from the arXiv repository, enabling it to effectively capture domain-specific terminology and structural patterns. The proposed pipeline integrates advanced text preprocessing techniques, including tokenization, stopword removal, and stemming, to enhance the quality of semantic representation. Experimental evaluation demonstrates that the fine-tuned DistilBART achieves high summarization performance, with ROUGE-2=0.472 and ROUGE-L=0.602, outperforming baseline transformer-based models. Unlike conventional approaches, the method shows strong applicability beyond academic research, including automated indexing of technical documentation, metadata extraction in digital libraries, and real-time text processing in embedded natural language processing (NLP) systems. The results highlight the potential of transformer-based summarization to accelerate scientific knowledge discovery and improve the efficiency of information retrieval across various domains
Evaluating random–Nyquist sampling ratios in combined compressed sensing magnetic resonance imaging
Get PDFCompressed sensing (CS) has been widely applied in magnetic resonance imaging (MRI) to accelerate the image acquisition without significantly reducing its image quality. In Cartesian MRI, acquisition time can be reduced by skipping phase-encoding steps for faster data acquisition. However, the balance between random under-sampling and Nyquist sampling at the k-space center strongly determines image quality. In this study, we systematically evaluate the impact of different random-to-Nyquist sampling ratios for both single-coil (CS-MRI) and multi-coil (CS-pMRI) reconstructions. Simulation results reveal that dense Nyquist sampling around the k-space center is essential for maintaining image fidelity, whereas reconstruction quality deteriorates sharply when random sampling exceeds approximately 60% of the total under-sampled data. Moreover, CS-pMRI consistently outperforms CS-MRI under equivalent under-sampling factors, benefiting from additional coil sensitivity information that improves resilience against aliasing and noise. These findings provide practical guidelines for hybrid under-sampling design, emphasizing that sufficient Nyquist sampling coverage of central k-space is crucial for achieving high-quality reconstructions while enabling high acceleration in CS-MRI