Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Design and optimization of a linear fiber-reinforced soft actuator for improved linear motion performance
Get PDFThe demand for safe and flexible actuators has increased as traditional actuators pose safety risks due to their rigid materials, especially in applications requiring human-machine interaction. This study focuses on designing and optimizing a linear fiber-reinforced soft actuator to enhance linear motion performance while maintaining safety and flexibility. Finite element method (FEM) analysis was used to evaluate the effects of varying key design parameters, including core radius, actuator length, and core wall thickness. The analysis revealed that increasing the core radius leads to greater linear extension, while increasing the actuator’s length and wall thickness reduces extension. Among the tested designs, the R10 design exhibited the highest linear extension, with a 44.41% increase in length compared to the original design. However, the R10 design also showed undesirable bulging at the free end under pressure, which necessitated further optimization. By increasing the thickness of the sheath wall, the bulging was reduced, and the optimized design achieved a 34.53% increase in extension. This study highlights the significance of parameter optimization in fiber-reinforced soft actuators to achieve superior linear motion performance. Future work will explore further improvements in structural stability, sensor integration for precise control, and advanced fabrication techniques for better customization and durability
SeeAround: an offline mobile live support system for the visually impaired
Get PDFThe inability of blind or partially-sighted people to understand visual content and real-life situations reduces their standard of living, especially in a world mainly tailored for sighted individuals. Despite the progress made by certain devices to assist them in using touch, sound, or other senses, these solutions often fall short of bridging the comprehension gap. Our work proposes an intuitive, user-friendly mobile-based framework named "SeeAround" that is capable of automatically providing real-time audio descriptions of the user's immediate visual surroundings. Our solution addresses this challenge by leveraging key point detection, image captioning, text-to-speech (TTS), optical character recognition (OCR), and translation algorithms to offer comprehensive support for visually impaired individuals. Our system architecture relies on convolutional neural networks (CNNs) such as Inception-V3, Inception-V4, and ResNet152-V2 to extract detailed features from images and employs a multi-gated recurrent unit (GRU) decoder to generate word-by-word natural language descriptions. Our framework was integrated into mobile applications and optimized with TensorFlow lite pre-trained models for easy integration on the Android platform
Advance technique for online condition monitoring of surge arresters
Get PDFGapless surge arresters (GLSA) constructed with zinc oxide (ZnO) element is connected in electrical power system for protection against surge voltage. For condition monitoring of GLSA conventionally offline and online techniques are available. However, offline techniques are not much useful as it requires system shutdown and hence online techniques are more useful. Online surge arrester monitoring technique based on leakage current analysis is adopted by all stakeholders. However, in this method supply system harmonics plays major roles in measurement accuracy to determine health index. In this paper improved health monitoring indexes for GLSA diagnostic based on ratios of leakage current components has been proposed. The ageing process has been done with application of lightning Impulse current, surface contamination due to salt fog, temperature effect, and moisture ingress. Various experimental tests have been carried out on porcelain and polymer housed surge arresters to evaluate the ability of proposed method. Obtained results of 9 kV, 18 kV, and 30 kV healthy and degraded metal oxide GLSAs have been shown the viability of improved health indexes on surge arrester condition monitoring procedures. The experimental investigation and discussion of the obtained results reflects sufficient and effective trails to utility engineers to determine health of surge arresters and able to effectively schedule further maintenance plan
Optimized convolutional neural network enabled technique for sentiment analysis from social media data
Get PDFSentiment analysis is an area of computational linguistics that studies natural language processing. The most significant subtasks are gathering people's thoughts and organizing them into groups to determine how they feel. The primary purpose of sentiment analysis is to determine whether the individual who created a piece of material has a positive or negative opinion about a subject. It has been claimed that sentiment analysis and social media mining have contributed to the recent success of both private sector and the government. Emotional analysis has applications in practically every aspect of modern life, from individuals to corporations, telecommunications to medical, and economics to politics. This article describes an improved sentiment analysis model based on gray level co-occurrence matrix (GLCM) texture feature extraction and a convolutional neural network (CNN). This model was created using tweets. First, texture characteristics are extracted from the input data set using the GLCM technique. This feature extraction improves categorization accuracy. CNNs are used to classify objects. It outperforms both the support vector machine and the AdaBoost algorithms in terms of accuracy. CNN has achieved an accuracy of 98.5% for sentiment analysis task
Enhanced detection of android ransomware families using machine learning and network traffic analysis
Get PDFRansomware attacks on Android devices often go undetected until damage occurs, as prevention strategies are limited by inconsistent threat detection and classification. This paper presents a framework for evaluating machine learning models to detect and classify Android ransomware families through network behavioral analysis. The framework extracts discriminative features from network traffic data and segregates them into four optimal clusters using the k-means clustering method. A total of 84 critical network traffic features are identified, including source IP, destination IP, source port, destination port, traffic duration, and the total number of forward and reverse packets. These optimal features are effectively utilized to train well-known machine learning models, including decision trees (DT), random forest (RF), K-nearest neighbors (KNN), support vector machines (SVM), and bagging, to evaluate their accuracy in classifying ransomware families. Simulation results demonstrate that RF achieves the best performance with an accuracy of 95.18%, precision of 95.21%, recall of 95.27%, and F1-score of 95.19%. This framework, focused on network behavioral analysis rather than static or dynamic analysis, provides deeper insights into the behavior and characteristics of ransomware
Ambulance integration model in smart city for health services
Get PDFSmart city is a city built to create efficient, sustainable, fair and livable cities. Increasing the percentage of population survival in standard and emergency times or emergency medical services (EMS) is one of the main pillars of a smart city. This study develops previous research. The weakness of this model is that it needs to consider the dynamic travel time factor. The model only considers the fixed travel time between two locations, which may not represent the actual conditions on the highway. It is less flexible because it only finds one threshold value α for all λi. Researchers succeeded in developing a model to overcome the model weaknesses in (4) by adding new constraints that consider ambulance capacity and overcome the shortcomings of the model (5) by adding tighter limits on the value of α to minimize the difference between the values of λi and αλi. In addition, constraints can also be added to ensure that α is a realistic value and meets business requirements
Durian plant health and growth monitoring using image processing
Get PDFThe demand for durians has increased considerably, gaining significant popularity in the market. Under the Industrial Revolution 4.0, precision agriculture is expanding globally, utilizing a range of digital technologies to provide the farming industry with crucial information for enhancing farm productivity. For durians to produce high-quality fruit, it is essential that the plants receive sufficient nutrients. Therefore, it is crucial for farmers to monitor the growth rate of durian plants to ensure they receive suitable nutrients for optimum growth. Manual growth monitoring often yields inaccurate results and is prone to human error. Thus, automatic systems for plant image analysis could prove highly beneficial for practical and productive agriculture. This research utilizes the you only look once version 5 (YOLOv5) model alongside an image referencing method for growth monitoring. It begins with the detection of the durian tree, segmenting the leaf area and computing tree size through image referencing. This method achieves a precision of 96% in detecting durian trees from images. Through these images, the growth rate of the durian is assessed through comparisons of canopy growth, stem size, and tree height
Screening capabilities for the 3D dyscalculia identification game framework
Get PDFDyscalculia, a learning difficulty in mathematics, remains a concealed challenge affecting individuals of average intelligence or remarkable creativity. This inconspicuous disability often leads teacher to misinterpret students as lacking intellect. Regrettably, this condition can prompt students to disengage from routine activities, resulting in diminished performance and self-confidence. To address this issue, our research introduces a serious game framework, namely the “3D-dyscalculia identification game framework†(3D-DIG framework), integrating a screening feature aimed at detecting mathematical shortcomings in students. This paper focuses on detailing the screening feature, wherein a Petri net structure orchestrates its functionality within the 3D game environment. Specifically, our study highlights how this feature assesses and captures potential student deficiencies during work on game challenges. Employing game engine, and web server technologies, the dyscalculia screening feature captures students' responses, enabling an evaluation of their mathematical proficiency. Analysis of student data affirms that the screening feature's in identifying potential mathematics-related deficiencies. Moreover, the 3D game incorporates a distinctive element: it notifies teachers when a student surpasses a 60-second threshold while solving a problem, facilitating timely interventions. By offering actionable insights, the framework empowers teacher to identify student with the mathematics' deficiency and support the student with the appropriate intervention
Hybrid CNN-ViT integration into Siamese networks for robust iris biometric verification
Get PDFIris recognition has emerged as a critical biometric verification method, valued for its high accuracy and resistance to forgery. However, traditional convolutional neural network (CNN)-based models, despite their strength in extracting local iris features, struggle to capture global dependencies, which limits their generalization across different datasets. Additionally, conventional classification-based approaches struggle to accurately verify new individuals with limited training data. Thus, this study proposed a hybrid CNN-vision transformer (CNN-ViT) model within a Siamese network to enhance one-shot learning capability by combining CNN’s local feature extraction with vision transformers (ViT’s) global attention. To evaluate its performance, the hybrid model was compared with VGG16 and ResNet under the same training conditions for 20 epochs. VGG16 and ResNet rely on pre-trained models, whereas the hybrid CNN-ViT model is specifically designed to achieve this task with an increment to 98.9% training accuracy, surpassing the TinySiamese model's benchmark accuracy. It also attained a recall of 75%, demonstrating strong sensitivity in correctly identifying positive matches. The hybrid model maintained an excellent balance between learning and generalization by employing the binary cross entropy (BCE) loss function. These findings contribute to the development of efficient iris recognition systems, paving the way for advanced biometric applications in financial transactions, border control and mobile security
CLAHE-AlexNet optimized deep learning model for accurate detection of diabetic retinopathy
Get PDFDiabetic retinopathy (DR) is a disease that affects the blood vessels that are located in the retina. Loss of vision due to diabetes is a common consequence of the illness and a key factor in the progression of vision loss and blindness. Both ophthalmology and diabetes research have become more dependent on computer vision and image processing techniques in recent years. Fundus photography, also known as a fundus image, is a method that may be used to capture an image of the back of a person's eye. This article presents optimized deep learning model for diagnostic marking in retinal fundus images towards accurate detection of retinopathy. For experimental work, 500 images were selected from available open source Kaggle data set. 400 images were used to train deep learning model and remaining 100 images were used to validate the model. Images were enhanced using the contrast limited adaptive histogram equalization (CLAHE) algorithm. Pre trained convolutional neural network (CNN) models-AlexNet, VGG16, GoogleNet, and ResNet are used for classification and prediction of images. Accuracy, specificity, precision and F1-score of AlexNet is better than VGG16, ResNet-50, and GoogleNet. Sensitivity of ResNet-50 is higher than other pre trained CNN models