Vasireddy Venkatadri Institute of Technology

This study explores the development of polyurethane foam concrete (PUFC) enhanced with nano-silica, nano-alumina, and graphene oxide to address the material’s inherent limitations in strength and durability while improving its sustainability performance. PUFC is widely recognized for its lightweight and thermal insulation properties, yet its low compressive and flexural capacity restricts structural applications. To overcome these drawbacks, a series of mixes were prepared with varying nanoparticle dosages, and their mechanical, durability, and microstructural properties were systematically evaluated. The results revealed that the optimized ternary blend (M2) achieved the highest compressive strength of 7.0 MPa at 28 days, representing a 42.6

Sikkim, with more than two hundred Buddhist monasteries and an extensive range of heritage and natural tourist sites, faces a disjointed digital tourism environment, lacking a comprehensive and AI-driven system for helping tourists plan visits to the region. While some generic applications provide shallow information, the official government websites lack relevance, and currently, there is nothing available that combines digital monasteries with AI assistance for creating itineraries, conversation with AI, and identification of landmarks via photographs. In this paper, an AI - Powered Sikkim Tourism and Monastery Discovery Platform is described, which is a full - stack application built using Django 4.2 and leverages Google Gemini's multimodal API to provide three different kinds of AI functionalities: a culturally themed chatbot, a virtual monk, an AI-assisted day-wise itinerary planner, and AI Lens for identifying landmarks using photographs. The website uses a centrally maintained and managed database of tourist spots and monasteries of Sikkim, using metadata, geographical location, and images, which are showcased via the use of the Google Maps Javascript API service. The platform incorporates a PLACES_FOUND parsing algorithm for connecting the output of AI to the structured database of places in order to dynamically generate place cards during the response. It incorporates twelve functional modules including optional user authentication, saving of places, storage of itineraries, importing JSON files, and admin intelligence dashboard using Chart.js. All twenty test cases - encompassing functional testing, edge-case testing, and security testing - received a PASS result. The platform is the solution to the Smart India Hackathon Problem Statement number SIH25061 and shows the feasibility of a multi-feature AI platform inside the B.Tech scope

Rapid urbanization and industrial growth have significantly increased the frequency of sudden air quality deterioration events, posing serious risks to public health and environmental sustainability. Conventional air quality monitoring systems are largely reactive, issuing alerts only after pollutant concentrations exceed regulatory thresholds, thereby offering limited response time for mitigation. To address this limitation, this paper proposes an early warning system for sudden air quality deterioration using time series analytics. The framework leverages multivariate, high-resolution air quality sensor data obtained from the UCI Machine Learning Repository to model temporal dependencies and predict short-term pollutant concentration surges. A Long Short-Term Memory (LSTM)- based time series forecasting model is employed to capture nonlinear temporal patterns and abrupt fluctuations in pollutant levels. An early warning mechanism is integrated with the predictive model to generate alerts prior to actual threshold violations. Experimental evaluation demonstrates that the proposed system achieves high forecasting accuracy, with a mean absolute percentage error below 10% across major pollutants, and effectively detects deterioration events with a detection accuracy of 91.3% and an average lead time of approximately 2.4 hours. Comparative analysis with traditional ARIMA and support vector regression models further confirms the superiority of the proposed approach. The results indicate that time series–driven early warning systems can significantly enhance proactive air quality management and support timely public health interventions.

Growing environmental concerns and stricter emission regulations have intensified the need for cleaner combustion and sustainable energy solutions. In this pursuit, Palmyra Methyl Ester (POME) stands out as a promising biodiesel, offering renewable origin and desirable fuel characteristics for cleaner, more sustainable engine applications. This study presents an integrated experimental and computational investigation into the performance, combustion, and emission characteristics of a diesel engine operating on POME blends, with a focus on optimizing injection timing and exhaust gas recirculation (EGR). Using a desirability-based multi-objective optimization framework, engine tests were conducted under varied conditions, guided by Response Surface Methodology (RSM). The predictive capabilities of RSM were benchmarked against advanced machine learning models like Extreme Gradient Boosting (XGBoost) and Random Forest. The optimal setting, found as POME20 with 23°bTDC injection timing and EGR, further improved BTE and significantly lowered NO x emissions. Among the predictive models, XGBoost outperformed RSM and Random Forest, yielding the highest test R 2 and lowest MSE and MAPE, demonstrating superior accuracy in predicting engine responses. These results highlight the synergistic potential of renewable fuel utilization and data-driven modeling in optimizing diesel engine operation. The findings provide a viable pathway toward cleaner, high-efficiency combustion systems, contributing to the broader goals of sustainable transportation and global energy transition.