Computer Science, Software, Information Systems, Human-Computer Interaction
19
Scopus Publications
Scopus Publications
Intelligent Crack Detection in Building Structures using Coupled Ultrasonic Guided Wave and Acoustic Emission Sensing Surajit Mohanty, Subhendu Kumar Pani Ssrg International Journal of Civil Engineering, 2026 Building structural integrity evaluation is important in the long-term safety and resilience of buildings. Visual inspection techniques that have been in place do not have the capability of detecting beneath surface cracks, or cracks that may develop at an early stage, that would compromise the structural performance. This paper introduces a smart crack sensor with a combination of Ultrasonic Guided Wave (UGW) and Acoustic Emission (AE) as a guide to the complicated Structural Health Monitoring (SHM) of building structures. The system proposed is based on the UGW-based wave propagation analysis along with the AE signal monitoring, which will detect, localize, and characterize surface and internal cracks with the highest accuracy. Algorithms of machine learning are used to comprehend complicated acoustic signals and distinguish between crack initiation, crack propagation, and the noise in the environment. It is experimentally verified on reinforced concrete specimens that the coupled UGW-AE methodology is more sensitive and accurate than the uniaxial methodologies. These are possible through the combination of real-time data acquisition, fusion of signals, and smart pattern recognition that allows early detection of damage and provides the ability to monitor the damage continuously. The study will help in the emergence of an intelligent, non-destructive, and scalable SHM system capable of improving structural dependability and maintenance effectiveness in contemporary infrastructure systems.
An integrated sensor technologies-based slope failure detection in mining operation using Xavier initialisation-based convolutional neural network Surajit Mohanty, Subhendu Kumar Pani Mining Technology Transactions of the Institutions of Mining and Metallurgy, 2026 The primary objective of modern mining endeavours in the twenty-first century is to safely and efficiently extract as much or as possible. Unstable slopes can result in fatalities and property damage, maintaining the stability of rock for economic sustainability and safety. This work innovative Xavier initialisation-based convolutional neural network (XI-CNN)-based model for detecting slope failures in mining operations. At first, the slope data is pre-processed in outlier removal using z-score, nominalisation, and normalisation using Min-Max. Then, the up-sampling is performed to improve the minority classes in the pre-processed data using adaptive synthetic sampling. After that, the slope features are extracted. Following this, important features are selected using good the bad and the ugly optimization to improve the classifier by reducing the dimensionality of the features. Lastly, the trained XI-CNN is fed the chosen features to classify the slope's stability. The proposed model is compared and analysed ##with the prevailing models and this demonstrates the higher detection accuracy (0.95) of the slope stability.
2025 International Conference on Digital Innovations for Sustainable Solutions (ICDISS) Design and Development of an AI-Augmented Web Analytics and Heatmap System for Real-Time User Behavior Insights Aarat Batra, Seema Verma, Madhumita Mahapatra, Surajit Mohanty Proceedings of International Conference on Digital Innovations for Sustainable Solutions Icdiss 2025, 2025 In the evolving landscape of digital user experience, real-time behavioral analytics plays a crucial role in data-driven decision-making. This paper presents the design and development of an AI-augmented web analytics system that captures and visualizes user interactions in real-time using heatmaps [2]. The system comprises three primary components: a React-based clientside SDK embedded into host websites to collect interaction data (clicks, hovers, scrolls), a Node.js backend server with WebSocket support for real-time data transfer and aggregation, and a Next.jspowered dashboard for visualization and analytics. Redis is employed for fast data caching and aggregation, while MongoDB serves as the persistent store. The dashboard displays dynamic heatmaps, session analytics, scroll depth, and interaction timelines, along with AI-powered tools such as feedback widgets and a natural language chatbot for interpretability. The architecture ensures minimal latency and scalability for hightraffic environments. This system enables website administrators to gain actionable insights into user behavior and optimize UI/UX in real-time. The integration of AI components further enhances user feedback collection and dashboard accessibility. The project countered the problems of real-time data streaming and collection effectively and provided the practical use cases of AI in the domain of web analytics. This project achieves a unique place among other web analytics solutions and thus, solves some of the major problems faced by businesses in terms of web-driven sales & marketing.
A Deep Learning Approach for Fake News Detection using CNN Sanjana Dash, Sutapa Susovita Bhuyan, Sanjit Kumar Dash, Rajeev Agarwal, Surajit Mohanty 2nd International Conference on Cognitive Green and Ubiquitous Computing IC Cgu 2025, 2025
Real time face mask detection in nuclear power plants: A deep learning framework using hybrid CNN-mobileNetV2 architecture Satya Ranjan Panda, Anuradha Rani Choudhury, Ashis Kumar Mishra, Surajit Mohanty, Satyajit Mishra Intelligent Computing Techniques and Applications, 2025 The paper involves the development of a real-time face mask detection system in nuclear power plants using Python, Keras, OpenCV, CNN and MobileNetV2. Its goal is to automate mask compliance detection in risky environments for the safety of workers. The dataset has 11,386 randomly gathered images divided between “with mask” and “without mask.” Adoption of lightweight MobileNetV2 architecture along with convolutional neural networks entails the highly efficient computation with the implementation of various libraries in Python including TensorFlow, Keras, OpenCV, numpy, imutils, matplotlib and scipy. The model was trained for over 40 epochs, reaching near-perfect accuracy, respectively, of 99.48% for mask detection and 99.9% for no-mask detection during daytime tests. Night-time performance slightly dropped to about 89–92% for masks and 93–94% for no masks. Post-training integration with OpenCV allows for the analysis of real-time video streams and provides immediate feedback towards compliance in safety. This system is robust and adapts to all lighting conditions, rendering reliable and efficient monitoring in nuclear facilities.
Asthma risk prediction through stacking-based ensemble learning Kumar Janardan Patra, Sanjit Kumar Dash, Jibitesh Mishra, Rajendra Prasad Panigrahi, Surajit Mohanty, Subhasis Mohapatra Intelligent Computing Techniques and Applications, 2025 Asthma is a widespread respiratory condition, making early and accurate prediction essential for effective treatment and management. This study investigates the use of a stacking-based ensemble learning approach to predict asthma risk. A dataset from Kaggle, containing information from 2,390 patients, was used to train and evaluate the model. A stacking ensemble method is implemented, combining the strengths of multiple classifiers to improve prediction accuracy. The results of the study are highly promising, with the stacking model achieving an Accuracy of 99.33%, F1-score of 98.48%, precision of 98.44% and recall of 99.33%. These results highlight the model’s effectiveness in identifying patients at risk for asthma with minimal errors. The findings demonstrate the potential of stacking-based ensemble methods to significantly improve asthma prediction accuracy. This approach could be a valuable tool for healthcare professionals in risk assessment and early intervention, paving the way for better patient outcomes and personalized treatment plans. Keywords: Ensemble technique, machine learning models, stacking method, t-SNE
Enhancing COVID-19 CT Image Classification with Attention-Augmented Pretrained Models Rohit Gupta, Raghunath Dey, Samrat Karna, Sweekar Koirala, Subhash Mehta, Sumit Gupta, Jayashree Piri, Surajit Mohanty 2025 2nd International Conference on Circuits Power and Intelligent Systems Ccpis 2025, 2025 To support overburdened healthcare facilities, the COVID-19 pandemic has spurred the development of quick and precise diagnostic techniques. To help radiologists with early detection and clinical decision-making, a deep learning classification algorithm for detecting COVID-19 infections from lung CT images is proposed in this study. We took four advanced pre-trained convolutional neural networks, Xception, VGG19, and ResNet50, and made improvements to them using transfer learning on a publicly available COVID-CT dataset, then tested how accurate they were. We used various data enhancement methods like rotating, flipping, and changing the contrast of the images to make the models stronger and prevent them from learning too much from the limited medical image data. We used key performance metrics, including accuracy, precision, recall, and F1-score, to compare the models. To separate COVID-positive cases from normal and other pneumonia cases, the Xception model enhanced with an attention head outperformed the other designs in terms of accuracy and sensitivity. The findings demonstrate the effectiveness of transfer learning in medical imaging tasks, especially when there is a lack of data, and point to a viable direction for developing reliable and scalable CT-based COVID-19 classification systems for environments with limited resources.
Medicare: A telemedicine healthcare website Payal Payal, Raghunath Dey, Rohan Senapati, Jayashree Piri, Surajit Mohanty Proceedings of International Conference on Circuit Power and Computing Technologies Iccpct 2024, 2024 Despite the advancement of technology, the modern healthcare system still has many limitations. In order to address many unresolved issues, the world is moving toward digitalized systems, including the healthcare system. A telemedicine system would greatly benefit society if developed to be inexpensive, easily accessible, and equipped with substantial, dependable medical resources. This paper explores existing research on telemedicine within healthcare management systems (HMS) and the services it offers. Traditional healthcare has limitations in accessibility and keeps evolving with technology. Telemedicine addresses these limitations by providing convenient, affordable, and dependable medical services. Like other communication technologies, telemedicine is intended for wider adoption and use. Our web application "Medicare", incorporates features related to HMS that will significantly improve the usability and reliability of this technology management system. This research identified similar existing applications, and additional functionalities are implemented to make the telemedicine website more impactful and relevant for future needs.
Economic Order Quantity Models with Exponential Demand Rate and Single Level Trade Credit Sweta Patro, Milu Acharya, Surajit Mohanty, Pranati Satapathy, Debabrata Singh Procedia Computer Science, 2024 In case of formulating more realistic inventory models, two factors of the problem have been raising attention of the researchers, one being the change in the demand rate and the other being the different types of trade credits. Keeping in view of these, this study includes an exponential demand rate, where demand is proportional to the derivative of inventory level, under the single level trade credit, which attracts more customers, leading to minimization of total cost. Optimum result for the time cycle and minimum total cost corresponding to the inventory model problems in crisp as well as cloudy and intuitionistic fuzzy situations are obtained by taking a suitable numerical example and the obtained results are presented in graphical form. Sensitivity analysis is carried out corresponding to some key parameters involved in the model problem using the Matlab-15 software and finally it ends with conclusion and future scope for research.
Approach for Preprocessing in offline Optical Character Recognition (OCR) Raghunath Dey, Rakesh Chandra Balabantaray, Surajit Mohanty, Debabrata Singh, Marimuthu Karuppiah, Debabrata Samanta 2022 International Conference on Interdisciplinary Research in Technology and Management Irtm 2022 Proceedings, 2022
