Institutional-Repository, University of Moratuwa
Welcome to the University of Moratuwa Digital Repository, which houses postgraduate theses and dissertations, research articles presented at conferences by faculties and departments, university-published journal articles and research publications authored by academic staff. This online repository stores, preserves and distributes the University's scholarly work. This service allows University members to share their research with a larger audience.
 Research Publications |  Thesis & Dissertation |  E- Books |
 UoM Journal Publications |  UoM Conference Proceedings |  Articles published in Scimago's Q1 journals |  UoM Research Reports |  Other Articles authored by UoM staff |
Recent Submissions
item: Conference-Full-text
An AI-driven smart mirror application for social event-based fashion recommendation and facial skin health analysis
(IEEE, 2025) Rathnasekara, N; Damith, N; Herath, D; Liyanaarachchi, SH; Seneviratne, OV; Wijenayake, U
The rapid development of Artificial Intelligence (AI) has opened the doors for many innovative applications in everyday use. Among them, a smart mirror is a conventional mirror with integrated features to provide real-time personalized recommendations and health advice. Modern lifestyles often prevent individuals from prioritizing skincare, outfit selection, or day-to-day planning. This paper proposes AI integrated smart mirror system to address these past difficulties. It consists of face verification for user authentication, skin health analysis, social event-based fashion recommendation system, displaying calendar events and weather updates with a proposed hardware device. The skin health analysis module has acne detection feature with localizing acne within “danger triangle” of the face, an important clinical aspect. After evaluating and comparing multiple stateof-the-art models, the most effective models were selected for the application. The system achieves separate performance metrics: 98% accuracy and 24 fps on a Central Processing Unit (CPU) for face verification (MediaPipe & FaceNet), 64% mean Average Precision (mAP) for acne detection (YOLOV8), and 98% mAP for fashion object detection (YOLOV8), demonstrating its suitability for real-time environments. By leveraging these features, the proposed smart mirror system achieves a significant step toward smart computing by seamlessly integrating with digital intelligence.
item: Conference-Full-text
Detection of multiple structural damage through modal analysis: an experimental and FE Approach
(IEEE, 2025) Rathnayake, CM; Hidallana-Gamage, HD; Lewangamage, CS
This study investigates the effectiveness of modalbased damage detection techniques, specifically the Difference in Mode Shape Slope (DMSS) and Difference in Mode Shape Curvature (DMSC), in identifying and localizing both single and multiple damage in beam and framed structures. Experimental tests were performed on a scaled 7-storey steel framed structure using a shake table, while Finite Element (FE) models were developed in ABAQUS and validated against experimental results, showing minimal frequency deviations. Damage scenarios were simulated, and the resulting mode shapes were analyzed. The DMSS and DMSC indices were computed and normalized to evaluate their sensitivity to damage. Results demonstrated that while both DMSS and DMSC successfully localized single and multiple damage scenarios in beam model and single damage scenario in framed structure, only the DMSC method reliably identified multiple damage in the complex frame structure. This highlights the DMSC method’s superiority for advanced damage detection. The study confirms the feasibility of combining experimental and numerical modal analysis for reliable structural health monitoring and sets the foundation for future applications in irregular and asymmetric structures.
item: Conference-Full-text
A Hybrid VLC/RF indoor localization with supervised learning
(IEEE, 2025) Maduranga, MWP; Saengudomlert, P; Jayaweera, VL; Dharmaweera, N
Indoor localization has become a key technology for a wide range of applications, from smart buildings to asset tracking, but conventional RF-based systems are limited by multipath fading and interference between signals. In this paper, a hybrid Visible Light Communication (VLC) and Radio Frequency (RF) localization system is proposed, using supervised learning to improve its performance against these limitations. By integrating the high precision of VLC and the extensive coverage of RF, our system achieves stable and accurate indoor positioning. Machine learning methods—Support Vector Regression (SVR), Random Forest Regression (RFR), and Gaussian Process Regression (GPR)on hybrid VLC/RF signal features for inferring device locations. Experimental results demonstrate that the hybrid system reduces root mean square error (RMSE) by 27% compared to RF-only and 4% compared to VLC-only approaches, with Random Forest Regression (RFR) emerging as the top-performing model. The system achieves sub-0.5m accuracy for 90% of test points, validated in a controlled 5m×5m×3m indoor environment. This work demonstrates the potential of hybrid VLC/RF systems, augmented by supervised learning, to revolutionize indoor localization for IoT and smart infrastructure use cases.
item: Conference-Full-text
Optimisation of washing cycles to enhance pre-processing of coal fly ash as a precursor for wastewater treatment using response surface methodology
(IEEE, 2025) Galappaththi, RY; Arunapriya, P; Hissalle, D; Chamith, S; Jayawardena, C; Fernando, A; Wickrama, MG
Coal, the most abundant and cost-effective fossil fuel critical for global electricity production, especially in developing economies, generates significant coal fly ash (CFA) waste, presenting environmental concerns. This study aims to enhance CFA's suitability as a zeolite precursor for wastewater treatment applications. We selected washing cycles for CFA pre-processing over mechanical activation, chemical modification, and leaching due to its simplicity, costeffectiveness, and efficiency. Accordingly, the CFA was washed multiple times using deionised water within a controlled environment. The activities were carefully monitored to ensure accuracy throughout the experiment. Detailed measurements were continuously taken of key factors such as the pH, solution conductivity, and CFA particle size across a range of conditions, including variations in temperature, number of washing cycles, and stirring duration. These observations were essential for method optimisation through response surface methodology (RSM). Findings of this research suggest that the optimal conditions are achieved at 6 washing cycles with a 3-minute stirring time and a temperature of 30 0C. These parameters enhance efficiency while reducing costs, demonstrating how waste material can become a valuable resource for sustainability. Further research on stirring rate and solid-toliquid ratio is recommended to optimise CFA pre-processing for high-yield zeolite production in wastewater treatment.
item: Conference-Full-text
Segment-driven corporate report summarization using positional-aware clustering and hybrid summarization
(IEEE, 2025) Weerasinghe, CDRM; Gunasinghe, MRAAK; Siriwardana, HBKS; Perera, I
Summarizing lengthy corporate reports is challenging due to their complex structure and varied formatting. Existing summarization methods often fail to preserve coherence or extract relevant content in these large-scale documents. This study introduces a novel divide-and-conquer hybrid summarization framework that combines semantically-aware text segmentation, extractive summarization using GUSUM with positional encoded embeddings, and transformer-based abstractive summarization. The proposed framework achieves strong results on benchmark datasets, 87.74 BERTScore and 54.65 ROUGE-1 on the Gov-Report dataset, and 79.7 BERTScore on the FindSum dataset, while a qualitative question-answering-based evaluation further reveals that summaries generated using our framework with LLaMA-3.1 (8B) covered 53% of answerable content compared to GPT-4-generated reference summaries, an impressive milestone considering the latter’s scale and cost.