446494 research outputs found
Sort by
Optimized Battery Management: A Comparative Analysis of Controller Systems for Enhanced Energy Efficiency
No full textBattery management systems (BMS) are important feature and is compulsory unit for all battery-powered systems. It was known that that battery management system (BMS) has various strategies like centralized, distributed, and mixed architectures. The BMS is important to understand their working nature, performance, safety and effectiveness at various environmental conditions. This article is focused on essential operations of BMS like precise state of charge (SOC) and state of health (SOH) estimation, fault detection and battery balancing. For better understanding, it is important to compare control strategies interms of complexity, cost and scalability. Emerging technologies like IoT, cloud computing, machine learning, and artificial intelligence will help in improving BMS, so that the data can be shared and monitor batteries in real time. With this the life of battery can be predictable, faults can be detected, and safety and performance can be improved. This work highlights the importance of BMS in securing the safety, efficiency and to extend the battery life
Intelligent multi-class classification and diagnosis of citrus leaf diseases using deep convolutional network
No full textCitrus plants, worthwhile to global agriculture, have their productivity drastically reduced because of diseases such as citrus canker, black spot, and greening. The actual diagnosis of these diseases requires a lot of technical expertise and takes considerable time; therefore it is impractical for extensive monitoring. This work proposes achieving an automated detection system using deep learning techniques for citrus leaf disease classification with four categories in the dataset, namely, canker, black spots, greening, and healthy. The dataset was augmented, thus improving model robustness by generating images. The system was developed using EfficientNetB0, which gives a good balance between accuracy and speed of the computational process. There was training and validation using k-fold cross-validation to ensure generalization. The model achieved test accuracy of 93%, supported by good precision, recall, and F1-scores across all classes. This study revealed that, for farmers, deep learning can be a trustworthy tool as far as fast and accurate disease recognition is concerned as required in precision agriculture
Digital Twin-Driven Predictive Maintenance for Electric Vehicle Powertrains: Case Studies and Quantitative Performance Insights
No full textThe rapidly increasing EV adoption across the globe demands advanced service strategies to ensure high levels of reliability, long component life and low operations costs. The key components of the EV powertrain, which are highly vulnerable to different electrical and mechanical failures, include sensor malfunctions, which can lead to poor performance, unforeseen repeated issues, and safety issues. The typical maintenance techniques are unable to address this dynamic and complicated behavior at a systems level among EVs, therefore making the malfunctions unpredictable, and the resources are very unproductive. This paper provides an evaluation of the opportunities of DT technology to revolutionise predictive maintenance in drive systems of PMSMs. It dwells upon the key principles of the DT architecture that may enable the AI/ML-managed PdM and provides the descriptions of the case studies that lead to tremendous decreases in unplanned downtimes, depending on the correctness of RUL estimates. The opportunities for future research are also noted, such as explainable AI and augmented cybersecurity. Concisely, the review under analysis reveals that DTs have a significant contribution to making future EV powertrains more reliable, efficient, and sustainable
A Swastika-Shaped Structure Monopole Filtenna for mm-Wave Applications
No full textA Swastika monopole antenna, which is 20 x 20 x 0.8 mm3 and covers the frequency range of 23 GHz to 32 GHz & 37 GHz to 80 GHz, a multi-band filtenna for various applications. The proposed filtenna with Roger substrate has a monopole ground size of 5 mm. This filtenna approach will assist the antenna in realizing the desired operating bands by covering the bandwidth of 9 GHz (23 GHz-32 GHz) & 63 GHz (37 GHz- 80 GHz). It is suitable for some parts of K-Band & Ka-Band, V-Band and applicable for 5G communications, Radar Communications, Satellite Communications. In addition to having good impedance matching and current distribution over the ground and radiating patch, this antenna offers a maximum gain of 8.5 dBi. Utilizing HFSS-19.2 software, the simulation procedure was completed
Mediapipe-based keypoint extraction with optimized neural models for offline smart control
No full textHand gesture recognition is a natural way of interaction between humans and computers. Among the many areas where it could be applied, smart homes and assistive systems are the most interesting ones. Still, most methods currently in use require sophisticated systems and cloud computing, thus, the setup causes latency, real-time and battery-operated applications are thereby limited. The present work proposes a straightforward, keypoint-based gesture recognition framework that employs the MediaPipe library for the efficient extraction of landmarks and optimized neural network classifiers for decision-making. By concentrating on four main gestures—Palm (ON), Fist (OFF), Thumbs Up (Increase), and Thumbs Down (Decrease)—the system enables offline, reliable, and fast control of home appliances. The experimental results have proved that the method reaches high accuracy while maintaining low computational cost, hence it becomes a suitable technology for embedded and real-time applications
Model Reference Adaptive Control for Wing-Rock Suppression in Delta-Wing Aircraft
No full textWing rock, a self-excited rolling oscillation occurring at high angles of attack, significantly affects flight stability and maneuverability in highly swept configurations. To address this, a nonlinear dynamic model representing the roll angle and roll rate behavior is developed based on experimental aerodynamic data. A Model Reference Adaptive Controller (MRAC) is designed to adaptively modify control inputs in real time, ensuring the system’s response follows a desired reference model despite nonlinearities and parameter uncertainties. The adaptive law, derived using Lyapunov stability theory, guarantees asymptotic error convergence and robust performance. Simulation results at a 30° angle of attack demonstrate that the proposed controller effectively suppresses oscillations, stabilizes roll dynamics, and achieves precise trajectory tracking under varying initial conditions. The study confirms the potential of MRAC as a reliable and scalable control approach for enhancing flight stability and safety in delta-wing aircraft
Thermovision Based Cursor Control Using Infrared Imaging and Deep Learning
No full textTouchless human-computer interaction is gaining significant attention for enhancing accessibility and hygiene. This project presents a cursor control system using thermovision, integrating deep learning with infrared imaging to enable screen navigation via hand gestures. The framework consists of four core components: a real-time thermal image processing pipeline that tracks the hottest regions via contour analysis and adaptive thresholding; a gesture classifier built on a TensorFlow Lite model, trained on thermal data to recognize five static gestures (FIST, ONE, PALM, SUPER, OPEN); a cursor engine that maps tracked hand movements to on-screen coordinates; and a stabilization module employing exponential moving averages and majority voting for improved accuracy and smoothness. By leveraging thermal tracking with a lightweight neural network, the system robustly handles varying lighting conditions, a common limitation of conventional RGB-based gesture systems. This thermographic approach provides reliable, contactless interaction without requiring visible light, making it highly suitable for assistive technology, industrial automation, and other touch-free applications. The entire system operates in real-time with low latency and is readily adaptable for edge device deployment
Design and Implementation of a YOLO-Based Screen Time Monitoring System Using PyQt and MySQL
No full textIn this paper, a YOLO is designed and developed. based screen time monitoring system that combines data logging and real-time face detection for precise computer tracking duration of use. The suggested system offers automated monitoring devoid of human involvement, addressing the growing issue of excessive screen time. The system is based on a specially trained YOLO object detection model that uses facial recognition to detect the presence of a particular user and guarantees accurate detection in a range of backgrounds and lighting conditions. There are four major subsystems that make up the core architecture: an OpenCV-based video processing pipeline for frame acquisition and visualization, a YOLOv8-based real- time detection engine tuned for webcam input, a MySQL-backed data storage system for recording cumulative screen time and presence intervals, and a PyQt-based graphical user interface with session control, usage analytics, and real-time monitoring
Analysis of Tooth Caries using the Deep-learning Model with Fused Features: A Study
No full textIn humans, the disease occurrence is happens in several ways and to manage it, it is necessary to implement appropriate diagnosis and treatment. Maintaining the oral health is a prime task and any abnormality will lead to various other health issues. This work considered the tooth enamel caries for the examination, which needs early diagnosis and treatment. This research proposed deep-learning scheme with EfficientNet (EN) model based examination of the tooth enamel caries. When a digital photograph of the tooth region is available, it is easy for evaluation the severity. This research presented a work to identify he mild and harsh enamel caries from the tooth based on the digital images. The different phases of this work includes the following sections; tooth-image collection and modifying its dimension, deep- features extraction with En-model, Softmax-based classification and identification of best two DL-model, reducing its features to 50% and serially integrating these values to generate a fused-feature vector, and verifying the merit using the machine-learning classifiers and 3-fold cross validation. The result of this work confirms that the developed system works well on the image database and provides>98% with the considered image examination task
AI-Based Secure SDN Framework for Smart City IoT Networks
No full textThe purpose of this project is to design an AI-Based Secure Software Defined Networking (SDN) Framework for Smart City IoT Networks to provide intelligent traffic management, real-time threat detection, and improved security for IoT devices that can communicate with each other. This project suggests a methodology that utilizes AI methods into SDN to improve the network's dynamic ability for identifying and mitigating cyberattacks. A hybrid security solution is utilized that uses both Rule-Based Detection mechanisms and Machine Learning (ML) Based Detection techniques. The Rule-Based Detection mechanisms make use of predefined rules and thresholds to recognize malicious activity, and a ML algorithm employs trained models to recognize sophisticated and unknown threats with high precision. The framework itself is realized in an SDN setup and also emulated through MATLAB software to analyze performance in various network attack situations. The outcomes show that the Rule-Based Detection registered an accuracy of 98.285% for well-known attack patterns, and the ML Based Detection was realized at a perfect degree of accuracy (100%) with the aim of efficient identification and classification of malicious network behavior. Overall, the AI based SDN framework integrates