Authors: R.Renuka, Dr.S.Mohana

Abstract: Typically, sleep disorders like insomnia, sleep apnea, and narcolepsy may not receive appropriate diagnosis until serious physical and mental health issues develop. Traditional techniques, though effective, involve polysomnography, which is not only labor-intensive and time-consuming but also demands special clinical conditions. Hence, this study aims to develop a framework that relies on AI techniques to utilize a hybrid model of Deep Learning techniques, including Convolutional Neural Networks (CNN) and Long Short- Term Memory (LSTM), to process EEG signals to identify sleep disorders. The CNN model can automatically identify spatial features in the raw signals, and the LSTM model can identify temporal dependencies in the signals to correctly classify Awake, REM, and NREM stages. Preprocessing techniques have been employed to clean and normalize the signals. The system, trained and validated using standardized data sets like PhysioNet, exhibits robustness and generalization in dealing with different patterns of sleep. It can also be used to analyze new EEG signals in real-time, detect abnormal sleep patterns, and predict the occurrence of sleep disorders. This intelligent system can greatly improve the efficiency of diagnosis and reduce the need to rely on manual diagnosis. It can also prove to be a cost-effective solution.

DOI: https://doi.org/10.5281/zenodo.20132129

Published by: vikaspatanker

Authors: Prof. G. R. Padule, Shweta Anil Bhosale, Dnyaneshwari Ravikant Patil, Vrushali Vishal Zambare

Abstract: Induction motors are vital components in industrial and commercial systems, where unexpected failures can lead to costly downtime and reduced productivity. Traditional maintenance strategies such as corrective and preventive maintenance are often inefficient, either reacting too late or performing unnecessary servicing. Predictive maintenance, powered by machine learning (ML) techniques, offers a smarter approach by forecasting motor health conditions based on real-time data analysis. This review paper presents an overview of recent advancements in predictive maintenance for induction motors using ML algorithms. Various techniques such as support vector machines (SVM), artificial neural networks (ANN), random forests, and deep learning models are discussed for fault detection, diagnosis, and remaining useful life (RUL) estimation. The paper also highlights the importance of feature extraction from vibration, current, and temperature signals, as well as the integration of Internet of Things (IoT) and cloud computing for real-time monitoring. Comparative analysis of different ML approaches is provided to identify their strengths, limitations, and potential for industrial application. Finally, the review outlines current challenges and future research directions for developing efficient, scalable, and interpretable predictive maintenance frameworks for induction motors.

Published by: vikaspatanker

Authors: Mr. Prajwal Narayan Chaudhary, Mr. Pranav Prasad Kulkarni, Mr. Chetan Ashok Bhalekar, Mr. Aditya Ganesh Gunjal, Professor Kalyani Zirpe

Abstract: Corrosion is a significant cause of damage in industrial infrastructure, transportation systems, marine equipment, pipelines, and metal parts. Traditional methods for inspecting corrosion mainly rely on manual observation and regular maintenance. These processes are time-consuming, labor intensive, and are subjective, which can lead to human error. Delays in spotting corrosion can lead to serious structural failures, higher maintenance costs, operational downtime, and safety risks. To address these issues, this paper introduces a real-time AI-based Corrosion Detection and Monitoring System. This system uses the YOLOv5 deep learning framework along with a modern web-based structure. The new system combines computer vision, deep learning, and web technologies to automate the detection of corrosion and assess its severity. It uses the YOLOv5s object detection model to find corrosion areas in uploaded images and live camera feeds. A React.js frontend offers an engaging and responsive user interface. Meanwhile, a FastAPI backend handles image processing, runs the necessary calculations, and communicates results. The system evaluates detected corrosion areas using bounding box calculations to estimate the amount of corrosion and categorize its severity as mild, moderate, or severe. It also features graphical visualizations, historical tracking, and repair suggestions to support preventive maintenance. This framework provides nearly real-time detection with higher accuracy and less reliance on manual inspection. Its modular and scalable design allows it to be used in various industries, including maritime, civil infrastructure, manufacturing, automotive, and aviation. Tests show that the system successfully identifies corrosion under different environmental conditions while maintaining good computational performance. This solution represents a cost- effective and smart way to monitor structural health and perform predictive maintenance.

DOI: https://zenodo.org/records/20121779

Published by: vikaspatanker

Authors: Tanu Yadav, Neelam Sahu, Deepak Sahu

Abstract: The rapid expansion of the pre-owned automobile industry has increased the demand for reliable and intelligent digital platforms for vehicle trading. Traditional used-car marketplaces often face challenges such as lack of transparency, inefficient search mechanisms, inconsistent pricing, and fraudulent listings, which reduce user trust and overall customer satisfaction. This research proposes an AI- powered car marketplace designed to improve the process of buying, selling, and exchanging second- hand vehicles through intelligent automation and secure digital infrastructure. The proposed system integrates advanced technologies including intelligent search optimization, personalized recommendation systems, automated listing moderation, and secure authentication mechanisms to enhance platform reliability and usability. The platform provides users with detailed vehicle listings, filtering and comparison features, responsive communication channels, and mobile-friendly accessibility to simplify customer interaction and decision-making. The backend architecture is developed to support scalable data management and efficient transaction handling using modern web technologies. Artificial Intelligence modules are incorporated to improve recommendation accuracy, optimize search relevance, and identify suspicious or duplicate listings. Experimental evaluation indicates that the proposed system improves search efficiency, recommendation precision, and operational transparency compared to conventional online used-car trading systems. The research demonstrates how AI-driven digital marketplaces can enhance trust, user engagement, and efficiency within the pre-owned vehicle industry while providing a scalable solution suitable for modern automotive e-commerce applications.

Published by: vikaspatanker

Authors: M.R.Manas, Dr. Umakanta Choudhury

Abstract: This study provides an in-depth analysis of the electromagnetic comparative assessment of the unskewed and skewed rotors for a 1 kW, three-phase, inner-rotor permanent magnet synchronous generator intended for small-scale direct-drive wind power applications. The generator has 36 stator slots and 12 rotor poles, with a 220 mm outer diameter of the stator and a stack length of 60 mm. The unskewed generator uses a ring magnet rotor design and features a gap size of 2.0 mm, while the skewed rotor design uses a block magnet rotor with a linear step of 10 degrees in three stages, with the air gap size of 1.5 mm. Performance criteria used for the finite-element-based simulations using Altair FluxMotor include the following: cogging torque, back-EMF waveform quality, losses, torque ripple, voltage, and efficiency, combined with thermal analysis. The reduction of the peak-to-peak cogging torque of the skewed rotor reaches 84.5phase back-EMF decreases by 67requirements of IEEE 519 regarding harmonics. Both the unskewed and skewed rotors show comparable efficiency at the same operating point (19 N·m and 500 rpm): 95.34full-load efficiency of the unskewed rotor (92.87the corresponding efficiency of the skewed rotor (91.96.

DOI: https://doi.org/10.5281/zenodo.20120373

Published by: vikaspatanker

Authors: Abdullahi Idris, Aminu A. Abdullahi, Jamilu Awwalu, Abdullahi Uwaisu Muhammad

Abstract: Clinical time-series data are inherently complex, characterized by temporal dependences, irregular sampling and missing observations making accurate longitudinal risk prediction a challenging task. The study presents a novel hybrid Transformer framework for temporal representation learning and longitudinal risk prediction in clinical time-series that integrates the strengths of self-attention mechanism of Transformers to capture long-range interactions across time steps with the LSTM networks in modeling short-term temporal dependencies. A fusion module is introduced to adaptively combine representations from both components, enabling robust learning from irregular and partially observed clinical data. The experimental results demonstrate that the hybrid transformer framework effectively categorized patients into high-risk and low-risk categories based on their attributes. The training results indicate that the model performed well, with an accuracy of 98.6%, a sensitivity of 96.2% and a specificity of 97.8%. The model correctly identified 11 out of 18 high-risk patients and 16 out of 22 low-risk patients, with apparent errors of 38.9% and 27.3% respectively. These findings indicate that the hybrid Transformer framework can successfully learn patterns associated with cardiovascular risk from training data. Similarly, the test results confirm the model’s ability to predict previously unseen data. The model correctly categorized 9 out of 12 high-risk cases and 6 out of 8 low-risk cases, resulting an overall accuracy of 91.2%, sensitivity of 89.3% and specificity of 92.0% with a 25% apparent error in both cases.

DOI: https://doi.org/10.5281/zenodo.20118703

Published by: vikaspatanker

Authors: Priyanka N Godiyal, Rutuja Amrale, Revati Ma’am, Archana Ma’am.

Abstract: Road traffic accidents are a leading cause of mortality worldwide, with India recording over 1.5 lakh fatalities annually. Identifying 'black spots' — specific road segments with disproportionately high accident frequency — is critical for targeted infrastructure intervention. Traditional methods of black spot identification rely on statistical thresholds applied to historical data, which are often reactive and location-agnostic. This paper proposes an integrated framework combining Machine Learning (ML) and Geographic Information Systems (GIS) for predictive black spot detection. We review and compare ML algorithms including Random Forest, XGBoost, Support Vector Machines (SVM), and Deep Neural Networks applied to multi-source data comprising accident records, road geometry, traffic volume, and environmental factors. Spatial analysis techniques such as Kernel Density Estimation (KDE) and spatial autocorrelation are used for feature engineering. Results show that ensemble methods achieve accuracy above 90%, with XGBoost yielding the highest AUC-ROC of 0.94. GIS-integrated output maps provide actionable, zone- specific risk rankings to support road safety planning.

DOI: https://doi.org/10.5281/zenodo.20118287

Published by: vikaspatanker

Authors: Assistant Professor Biju George.G

Abstract: Earthquakes often create hazardous environments that delay rescue operations and increase risks to emergency responders. This paper presents a low-cost Internet of Things (IoT)-enabled autonomous rescue robot designed to assist in earthquake search and rescue missions through real-time monitoring and multi-sensor integration. The proposed system employs an Arduino Uno as the central controller and integrates a Passive Infrared (PIR) sensor for human detection, an ultrasonic sensor for obstacle avoidance, an MQ-series gas sensor for hazardous gas detection, a GPS module for victim localization, and an ESP32-CAM module for live video surveillance. Sensor data are transmitted through an ESP8266 NodeMCU to the Blynk IoT platform, enabling remote monitoring of environmental conditions, survivor alerts, and robot location. The developed prototype was experimentally evaluated under laboratory conditions to validate its sensing, navigation, communication, and monitoring capabilities. Experimental results demonstrate reliable human detection, obstacle avoidance, hazardous gas monitoring, GPS-based localization, and real-time data transmission with low implementation cost and reduced computational complexity. The proposed system offers a practical and scalable solution for improving rescue efficiency while minimizing the exposure of rescue personnel to hazardous environments. The modular architecture also provides flexibility for future enhancements using artificial intelligence and advanced sensing technologies.

DOI: https://doi.org/10.5281/zenodo.21884018

Published by: vikaspatanker

Authors: Harini . K, Assistant Professor S.Janani

Abstract: The rapid growth of Over-The-Top (OTT) platforms has fundamentally changed the way users consume multimedia content on the internet. Streaming services like Netflix have set a highstandard for user experience, interface design and content discovery. This project, titled “Netflix Clone – Web Application Using HTML, CSS, JavaScript, React and Vite” . The backend is built using Node.js, providing RESTful APIs to handle user interactions and data management. PostgreSQL is used as the relational database for storing user data, movie details, and other application information., aims to design and implement a responsive and interactive web interface that emulates the core look and feel of the Netflix platform using modern front-end technologies. The primary objective of this project is to build a single-page application (SPA) that allows users to browse a catalogue of movies and series, view categorized lists such as “Popular”, “Trending” and “Top Rated”, and navigate to detailed information pages for each title. The user interface isimplemented using HTML5 for structure, CSS3 for styling, and JavaScript for dynamic behavior. React is used as the front-end library to efficiently manage UI components and state, whileserves as a fast and optimized build tool and development server. The system implements CRUD (Create, Read, Update, Delete) operations for managing users, movies, and preferences. The application is developed using Visual Studio Code and follows modern web development practices. Together, these tools provide a modern, modular and scalable environment for building the Netflix clone. The project follows a structured approach including requirement analysis, interface design, component design, implementation, and basic testing. Emphasis is placed on responsive design to ensure that the application works across desktops, laptops and mobile devices. Features such as reusable React components, props and state, routing (if used), and API-like data retrieval from static JSON or mock data are incorporated to simulate real-world application behaviours

DOI: http://doi.org/

Published by: vikaspatanker

Authors: Nimisa Bhagchandani

Abstract: The rapid growth of digital commerce and the increasing use of artificial intelligence have significantly changed the way consumers make purchase decisions online. One of the most common applications of AI in this space is the use of personalization algorithms, which provide users with tailored product recommendations based on their preferences and past behaviour. While these systems are designed to improve convenience and enhance user experience, they may also create unintended challenges for consumers. This study examines the impact of personalization algorithms on consumer decision fatigue and purchase decision-making in digital commerce contexts. The research focuses on understanding whether personalized recommendations simplify the decision-making process or contribute to cognitive overload. Decision fatigue is considered as a key factor that may influence how consumers respond to multiple product options and recommendations. The findings of the study are expected to provide insights into how personalization influences consumer behaviour beyond its intended benefits. It highlights the need for digital platforms to balance personalization with user comfort and cognitive ease. The study contributes to a better understanding of the psychological effects of personalization and its role in shaping online consumer decision-making.

DOI: https://zenodo.org/records/20099925

Published by: vikaspatanker