Authors: Aditi Nandiraju, Hunar D, Ashutosh,, Somraj, Janaki Kandasamy

Abstract: As critical infrastructure in aviation and healthcare becomes increasingly complex, traditional reactive strategies for maintenance and security are proving insufficient for handling dynamic real-world environments. This research examines the integration of AI-enabled predictive monitoring and security frameworks to create resilient, self-sustaining systems that can manage uncertainty with minimal human intervention. Central to this transition is the application of AI and machine learning models—such as XGBoost, CNNs, and LSTMs—to move from scheduled to proactive maintenance by accurately predicting the Remaining Useful Life (RUL) of aircraft engines and providing early warnings for cardiac events in healthcare. Simultaneously, the study prioritizes security by developing defense mechanisms against cyber-physical threats, including GPS spoofing, ADS-B vulnerabilities, and unauthorized network intrusions across both aviation and smart airport infrastructures. Despite these advancements, significant barriers remain, including high computational overhead, a lack of model interpretability (the "black box" problem), and a gap between simulation and real-world deployment. This work concludes that the future of dependable infrastructure lies in unified, lightweight, and explainable frameworks that allow systems to autonomously detect threats, recover from faults, and maintain themselves in unpredictable conditions.

DOI: https://doi.org/10.5281/zenodo.20093181

Published by: vikaspatanker

Authors: Saanvi Anup K, Nandana D Nair, Shajahan Basheer, Suresha R

Abstract: In the digital age, social media platforms generate vast amounts of unstructured data that serve as a goldmine for businesses, marketers, and content creators. Identifying trending topics and understanding content engagement dynamics is critical for strategic decision- making. This report reviews 30 research papers focusing on social media analytics, ranging from big data architecture to advanced deep learning models. Based on this review, we propose a ‘Social Media Analyzer’ system designed to extract trending hashtags, perform sentiment analysis on user engagement, and provide actionable insights. We select "A Deep Learning Sentiment Analyser for Social Media Comments in Low-Resource Languages" (Paper #13) as our base paper for its robust handling of informal text. The proposed work integrates Topic Modelling (LDA) with a Hybrid Deep Learning Classifier to predict content virality and audience sentiment.

DOI: https://doi.org/10.5281/zenodo.20092205

Published by: vikaspatanker

Authors: Sukanth Tumu, Balasubbareddy Mallala, Sudhakar Babu Thanikanti, U.Nikhil tej, B.Murali, N.Suresh

Abstract: The Real-Time Grid Monitoring System is an IoT-based project designed to continuously monitor and control electrical parameters in a power distribution setup. The system utilizes a NodeMCU (ESP8266) microcontroller for real-time data acquisition, processing, and wireless communication. A potentiometer is used to simulate and monitor voltage variations, while an LM324 operational amplifier serves as a crucial component for detecting short-circuit and open-circuit faults in the grid. In the event of such abnormalities, or when undervoltage conditions occur, a buzzer is activated to provide an immediate alert.The system incorporates two relays, enabling remote switching of connected loads through an IoT-based web interface, allowing users to manually control devices from anywhere using a smartphone or computer. Additionally, a 16×2 LCD display presents real-time voltage status, load condition, and fault information locally. This integration of hardware monitoring and IoT control ensures improved reliability, safety, and user convenience. The proposed system provides a cost-effective and scalable approach to enhance smart grid management, offering real-time visibility and quick response to faults. It demonstrates the potential of IoT in modern electrical systems by bridging automation, monitoring, and fault detection into a unified platform. Keywords: NodeMCU, RELAYS, BUZZER, LOADS, LCD, VOLTAGE MONITOR, LM324, OC, SC.

DOI:

Published by: vikaspatanker

Authors: Nitin, Ishan Rana, Dr. Neha Gupta

Abstract: Food spoilage is a big concern affecting health, safety, and economy worldwide. This paper presents the design and implementation of a food spoilage detection system using an Arduino Uno microcontroller and MQ-135 gas sensor. The system detects gases such as ammonia, carbon dioxide, and volatile organic compounds (VOCs) released during food decomposition. Gas sensors provide a non-destructive and efficient way to monitor food quality by detecting chemical changes in the surrounding environment [1]. The MQ-135 sensor is mainly used due to its sensitivity to harmful gases associated with spoilage [2]. When the gas concentration exceeds a predefined threshold, the system alerts the user through an LED indicator in Arduino Board. The proposed system is cost-effective, portable, and easy to implement. It can be used in households, food storage facilities, and small-scale industries to ensure food safety and reduce wastage.

DOI: https://doi.org/10.5281/zenodo.20196415

Published by: vikaspatanker

Authors: Tejaswini Bagade, Preeti Wagh, Ms. Neeta Takawale

Abstract: This project focuses on the design and development of a Real-Time Traffic Flow Forecasting and Management System using machine learning and deep learning techniques. The system aims to predict traffic conditions accurately by analyzing real-time and historical traffic data collected from sensors, CCTV cameras, and GPS devices. Data preprocessing techniques are applied to remove noise and handle missing values for improved prediction accuracy. Advanced models such as LSTM, GRU, and CNN–LSTM are implemented to forecast traffic flow and support intelligent traffic management decisions. The proposed system helps reduce traffic congestion, improve road safety, optimize signal control, and enhance transportation efficiency through real-time monitoring and adaptive management strategies.

Published by: vikaspatanker

Authors: Mani G

Abstract: The fast growth and acceptance of cloud computing technology have completely changed the IT infrastructure of organizations, but along with that transformation, there have been several emerging security concerns. These security concerns have become hard to detect using conventional security approaches, due to the complexity and the evolution of new cyber attacks. In this paper, a complete deep learning cybersecurity framework will be proposed, to detect any threats in real-time within cloud computing environments. The cybersecurity framework consists of several deep learning models. They include the TCN with an autoencoder to detect anomalies at 99% accuracy with a false positive rate of 2.2% based on CSE-CIC-IDS2018 dataset, a transformer with CNN to detect network intrusions with 99.12% accuracy, and a federated learning method for detecting attacks in distributed environment without violating any user’s privacy at 98.3% accuracy in 300 communication rounds.

DOI: https://doi.org/10.5281/zenodo.20085675

Published by: vikaspatanker

Authors: M.Indumathi, Uddandam Vinodkumar

Abstract: Tumor segmentation in Magnetic Resonance Imaging (MRI) plays an important role in diagnosis, treatment planning, and disease surveillance. But still there are many hurdles in the process because of low contrast tissues, unclear boundaries and high morphology variations. In this paper, we propose Contour-Aware U-Net (CAU-Net), which uses explicit contour refinement techniques along with multi-level feature fusion. Our framework includes three main components that are as follows: (1) Contour-Aware Decoder with Attention Fusion blocks for contour enhancement, (2) adversarial learning constraint for anatomically plausible results, and (3) combined hybrid loss function using cross entropy loss, dice loss, and sub-differentiable Hausdorff loss. Extensive experiments on tumor datasets have proven that our proposed approach outperforms existing approaches in terms of accuracy by producing Dice Similarity Coefficient score of 0.92 and reducing Hausdorff Distance by 38%. Our model performs exceptionally well in terms of boundary delineation that was the crucial requirement in clinical practice.

DOI: https://doi.org/10.5281/zenodo.20085607

Published by: vikaspatanker

Authors: Ashish Shrimali, Dr. Priyanka Pandey

Abstract: Eco-friendly cement-based bricks incorporating stone mine waste and quarry dust have attracted considerable interest in the quest for environmentally sustainable building resources. The global research trends in sustainable cement-based bricks made with granite waste, sandstone waste, quarry dust and other mining waste products are reviewed from a performance and a bibliometric perspective in this study. The research identified 400 research articles from Lens.org database using a search query on the keywords: cement/concrete, stone waste products and mechanical and durability properties. Bibliometric analysis of articles was performed by VOSviewer for the mapping of the growth of publications, notable authors, research hot spots and evolutionary trends. Publication trends indicate steady increase after 2018 and an explosion of research between 2020 and 2023, which are the factors related to the increase of research and industrial interest handling eco-efficient masonry products. country wise analysis for published literature shows India leads the study as it is the most studied, followed by China and Malaysia, indicating good regional interest in valourisation of stone waste and promotion of material innovations. Key journals, such as Materials and Sustainability, are important in this field, highlighting the interdisciplinary nature. Keyword co-occurrence analysis revealed that "compressive strength", "durability", "quarry dust" and "recycled aggregates" are commonly used; new directions are circular economy principles and eco-efficient designing of materials. The insight gained from the case studies towards performance suggest that partial replacem However, there are problems on standardization, large-scale adoption and long-term testing which need to be resolved. This research provides an overview of existing research directions and possibilities for further research, e.g. hybrid forms of waste recycling and life cycle assessment. Our results encourage the role of stone waste as a resource sustainable brick-making as well as to guide further research and development of the industry in producing affordable and sustainable building materials.

DOI: https://doi.org/10.5281/zenodo.20281288

Published by: vikaspatanker

Authors: Aniket Mishra, Ajinkya Bagal, Jayesh Jadhav, Rushikesh Nath

Abstract: This study examines the incorporation of generative artificial intelligence (Gen-AI) into e-commerce recommendation systems. Traditional approaches, such as collaborative filtering and content-based filtering, face challenges like sparse data, cold-start issues, and changing user preferences. Gen-AI models, especially transformer-based frameworks like GPT and diffusion models, provide innovative solutions for understanding and creating personalized content. This paper reviews the progression of recommendation systems, introduces generative models, and proposes a framework that integrates Gen-AI with current recommendation strategies to enhance accuracy, diversity, and contextual relevance.

DOI: Name : aniket mishra Contact No : +919518352808

Published by: vikaspatanker

Authors: Anshika Singh, Sneha Chhabra, Rajat Takkar, Harshwardhan Singh Thakur

Abstract: This study investigates the rising global disease burden, emphasizing the need for early detection to minimize mortality and healthcare costs. This article proposes a machine learning model for predicting disease risk from a dataset of 4240 patient records. Each record is characterized by 15 clinical and demographic attributes. This research paper employed five classifiers—Logistic Regression, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and Naive Bayes—to identify disease presence. Using hold-out validation, the models were evaluated, and Logistic Regression achieved the highest accuracy of approximately 84%, followed by Random Forest (~83.7%), SVM (~83.3%), and KNN (~82–83%). These results show the potential for early disease detection, enabling timely interventions. By integrating such models into practice, clinicians can maximize patient outcomes and reduce the disease burden globally. Future development includes expanding the dataset and adding an accessible interface for real-time analysis of disease risk.

DOI: https://doi.org/10.5281/zenodo.20226307

Published by: vikaspatanker