SLIIT International Conference On Engineering and Technology,Industry Connect Papers Vol. 04 [SICET] 2025
Permanent URI for this collectionhttps://rda.sliit.lk/handle/123456789/5176
Browse
Item Open Access Comparison of Pile Driving Equations(Faculty of Engineering, 2025-09-09) Thilakasiri H. SIn Sri Lanka, remote highway projects, driven piles are installed in large quantities mainly as a ground improvement where the soft soil thickness is very high and closer to the bridges to decrease the settlements. In building sector as well, the installation of the driven piles are very common to decrease the foundation construction time and has seen the installation of driven piles as the settlement reduction means. Recently, silent pile installation technology has increased and the Sri Lankan investors has invested large sum of money for the silent hammers. Capacities of the piles can be estimated using the pile driving equations during installation of the piles. This is very convenient to determine the capacities of the piles without performing a pile load test. On the other hand, this can be reducing the site investigation requirements to a minimum. As the capacities of the piles are estimated during installation, it can take into account in-situ subsurface conditions. The rapidly varying ground conditions, especially in Sri Lanka, or any other location, can be captured by pile driving records. Recently, especially in Sri Lanka, the installation of the PHP piles with a hollow area at the middle of a pile section has increased. The design community multiplies the area of the pile cross section by the end bearing capacity to obtain the end bearing force whether it is a PHP pile or a solid pile. This research will compare the pile driving equations with each other and select the best pile driving equations which predict the capacity very close to the measured capacity as an extension of the earlier research done by Thilakasiri and Jayaweera (2009). In addition the mobilization of the end bearing capacities in PHP piles with a hollow area at the middle of a pile section is compared with a solid pile and a relationship is derived to predict the capacity of the hollow area as a function of the hollow area as a percentage of the total area of the pile cross section.Item Open Access Design, Manufacturing and Operation of a Down draft Gasifier for Coconut Shell Charcoal Generation in Dankotuwa, Sri Lanka.(2025-09-09) Nagahawatta, R; Perera, D; Ratnaweera, J; Nilanka, PThis study presents the design, fabrication, and operation of a downdraft gasifier specifically developed for coconut shell charcoal production in Dankotuwa, Sri Lanka. Traditional pit methods of charcoal production result in low yields, uncontrolled emissions, and limited process control. The proposed gasifier overcomes these issues by enabling batch feeding of green coconut shells from the top, continuous operation, and controlled partial oxidation to extract charcoal without reducing it to ash. The system incorporates a combustion cone with four strategically oriented air nozzles to ensure uniform material flow and high-temperature combustion, followed by a rotating grate in the reduction zone to regulate charcoal expulsion. Commissioning trials with a 1000 kg feedstock demonstrated a certain charcoal yield of at least 28% (w/w) while generating producer gas of sufficient calorific value to replace approximately three cubic yards of wood logs per 10-hour shift for industrial heating. The dual outputs of charcoal and producer gas provide both economic and environmental benefits, making the system a viable solution for small- to medium-scale enterprises seeking sustainable, low-cost energy.This study presents the design, fabrication, and operation of a downdraft gasifier specifically developed for coconut shell charcoal production in Dankotuwa, Sri Lanka. Traditional pit methods of charcoal production result in low yields, uncontrolled emissions, and limited process control. The proposed gasifier overcomes these issues by enabling batch feeding of green coconut shells from the top, continuous operation, and controlled partial oxidation to extract charcoal without reducing it to ash. The system incorporates a combustion cone with four strategically oriented air nozzles to ensure uniform material flow and high-temperature combustion, followed by a rotating grate in the reduction zone to regulate charcoal expulsion. Commissioning trials with a 1000 kg feedstock demonstrated a certain charcoal yield of at least 28% (w/w) while generating producer gas of sufficient calorific value to replace approximately three cubic yards of wood logs per 10-hour shift for industrial heating. The dual outputs of charcoal and producer gas provide both economic and environmental benefits, making the system a viable solution for small- to medium-scale enterprises seeking sustainable, low-cost energy.Item Open Access Enhancing Patient Safety with the MedAlert SYstem (MASY): A Low- Cost Timer for Medication Administration Alerts(Faculty of Engineering, 2025-09-09) Rodrego, P; Fernando, N; Wijekoon, N; Wanigasekara, D; Herath, JWhen it comes to patient safety, the timely administration of medication tends to be a critical factor in almost all the healthcare related environments, especially in hospital wards of any scale and clinics where multiple patients simultaneously require any form of medication. In many small to medium scale hospital wards and rural healthcare environments, where commercially available systems which are developed for patient management tend to be extremely costly to deploy due to many factors such as the scale of the facility and lack of staff with the necessary knowledge to run and maintain such a sophisticated system, staff always seem to rely on logbooks for the purpose of tracking the issued doses and the times at which these specific medications are issued. This process is prone to human error, inefficiency, data tampering, and lack of accountability. This paper therefore presents an economical solution addressing the issues mentioned above: the MedAlert System (MASY), which is a low-cost, standalone timer unit specifically designed and developed for healthcare settings. This system, based on the popular Arduino open-source platform, enables medical staff members to manage multiple medication timers simultaneously with no reliance on any logbooks or other external references. This system also offers both visual and auditory alerts to ensure that medication is always administered at the correct time. Unlike commercial patient monitoring systems implemented in large-scale hospitals, MASY can operate independently, with no reliance on servers or internet connections. Due to its simplicity, this system can be easily modified and implemented according to local workflows. This paper discusses the system’s software, design, human machine interaction, and the future scope of the system’s development. It is believed that implementing such a simple device has the potential to reduce missed or delayed medication doses and, in turn, improve patient safety in almost any setting.Item Open Access Faster Than the Teacher, Smarter Than the Student: Classifying with Wisdom via Knowledge Distillation in LLMs(Faculty of Engineering, 2025-09-09) Gobihanath B.; Abishethvarman V.; Prasanth S; Banujan K.; B.T.G.S KumaraLarge language models (LLMs) have achieved remarkable success across various natural language processing (NLP) tasks, driven by their ability to capture complex language patterns through large-scale pretraining. However, their substantial computational demands limit their deployment in resourceconstrained environments. To address this, this research introduced Knowledge distillation-based framework for text classification using a multiclass approach across three domains: entertainment, sports, and politics. We utilize both hard labels (ground-truth categories) and soft labels (logits from a teacher model) to train a student and a distilled model. The teacher model is accurate but computationally expensive. The student model is lightweight and fast, yet less accurate. Through distillation, we derive a task-specific distilled model that balances speed and accuracy. We also compare the performance of these models against traditional classifiers such as LSTM, SVM, and Naive Bayes. Traditional models excel comparing to the LLMs. Considering only task agnostic language models, evaluation shows that the distilled model performs significantly better than the student and competitively against the teacher, offering a practical trade-off. Our study demonstrates the value of soft label transfer and semantic alignment for improving classification performance in resource-constrained environments. The text classification code can be found at: https://github.com/Abishethvarman/KD-Text-ClassificationItem Open Access Impact of IoT on Personal Area Networks (PANs)(Faculty of Engineering, 2025-09-09) Thilakarathna D.R.T.D.; Gamage N.H.; Mendis B.P.U.; Kavindya P.M.S.; Sammani H.M.; Senanayake M. M. V.The Internet, as a revolutionary technology, continues to develop new technologies and software, making it accessible to all. Today, the most common forms of communication are either human-to-human or human-to-device communication; however, the Internet of Things (IoT) foresees a promising future for machine-to-machine (M2M) communication. Many novel wireless technologies, including ZigBee and Bluetooth, compete to provide the Internet of Things with low-power wireless communication solutions; however, in some IoT applications, the technological options are constrained by hardware resource limitations, low power consumption requirements, and overall device costs. Low power consumption is a basic prerequisite for enabling IoT expansion. Besides low power consumption, other requirements must be considered, such as technology cost, security, manageability, usability, wireless data rates, and communication ranges, among others. This paper discusses how the Internet of Things is transforming PANs, with particular focus on proximity communication protocols such as IEEE 802.15.4, Bluetooth, and ZigBee. The scope of this study extends beyond conventional device interconnection to cover new application areas like smart homes, healthcare, wellness, and wearables, succinctly presenting key trends and challenges from current literature, technology standards, and empirical evidence while analyzing critical factors in IoT-integrated PANs such as network scalability, privacy, interference control, and data security, which introduce new complexities to design and administration. This research offers details of IoT-PAN integration by conducting an analysis of wireless integrated personal area networks to identify research gaps and propose future directions. It defines the most important communication protocols and examines adequate levels of security and privacy, analyzing relevant literature to develop a robust framework that enables researchers and practitioners to address gaps in the literature regarding IoT-PAN integration. Therefore, this paper highlights the importance of PAN solutions that are secure, adaptable, and interoperable to enable next-generation IoT ecosystems by providing future insights.Item Open Access Integrating Large Language Models into Personalized Diabetes Care: A Systematic Review of Clinical Applications, Model Adaptation Strategies, and Ethical Implications(Faculty of Engineering, 2025-09-09) Jayakody,J.A.U.S.; Jayawardhana, S.M.M.S.; Thilakarathne, P.R.H.N.G.This systematic review examines the integration of Large Language Models (LLMs) into personalized diabetes care, focusing on clinical applications, adaptation strategies, and ethical considerations. As diabetes management demands increasingly personalized approaches, LLMs including GPT-3 and GPT-4 show promise for patient education, clinical decision support, and diagnostic assistance. This review synthesizes findings from studies published between 2018 and 2025 to evaluate LLM clinical applications and assess adaptation techniques. Key applications include conversational agents for patient education, personalized decision-making systems, and predictive modeling for diabetes-related complications. Model adaptation through domain-specific training and multimodal integration demonstrates enhanced performance in clinical settings. However, significant challenges persist, including data privacy concerns, model fairness issues, and limited real-world validation. Ethical considerations encompass training bias and data security, highlighting the need for privacy-preserving approaches. The review identifies critical gaps in current research and proposes future directions emphasizing explainable AI models to build trust among healthcare professionals and patients. While LLMs offer transformative potential for personalized diabetes care, their responsible integration requires addressing technical, ethical, and regulatory challenges. This synthesis provides a foundation for advancing LLM applications in diabetes management while ensuring patient safety and equitable care delivery.Item Open Access Karate Kata Scoring and Performance Evaluation Using Video Analysis and Deep Learning(Faculty of Engineering, 2025-09-09) Liyanage, H.L.S.S.; Deshpriya, H.M.S.D.; Kavindya,M.T.; Ranasinghe, R.M.L.D; Herath, H.M.D.P.; Aththanagoda, A.K.N.L.Karate kata, a fundamental element of traditional martial arts, consists of choreographed sequences of stances, strikes, and blocks performed against imaginary opponents, requiring precise posture, timing, and technique. However, self-practice often lacks objective feedback, leading to unnoticed errors and uneven skill development, and even competition scoring can be influenced by subjective judgment. To address this, we present a vision-based scoring system that evaluates kata from structured video recordings without wearable sensors, providing accurate, consistent scores to help practitioners track progress. The system processes videos through a multi-stage pipeline: frames are extracted at 10 frames per second using OpenCV and segmented with YOLOv8-seg to remove background clutter. Pose estimation is performed with MediaPipe, generating 99 normalized 3D keypoints, while hip-centered normalization and pelvic-width scaling ensure consistency across different body sizes. A modified ResNet50 classifies frames into nine fundamental stances with high accuracy and near real-time processing (~150 ms per frame at 1080p), and a custom Convolutional Neural Network evaluates the angular accuracy of 14 critical joints using trigonometric analysis against reference poses. Each kata is decomposed into 19 key positions, with a stance marked correct only if both classification and joint angles (≤ ±2.5° deviation) meet the threshold, and final scores are calculated on a 5.0–10.0 scale to align with traditional judging. Experimental results show that ResNet50 achieves 95.97% accuracy, while the CNN reaches 96.11%, demonstrating that this approach offers a low-cost, accessible, and consistent scoring tool. It supports remote training, reduces judging bias, and provides practitioners with reliable feedback to improve their kata performance.Item Open Access Smart Train-Elephant Collision Management System(Faculty of Engineering, 2025-09-09) Perera G.P.K.N; Nimsara P.P.; Perera P.R.D.N.; Weerasinghe T.G.J.N; Kularathna P.H.G.U.; Morapitiya S.SThis paper presents a novel solution for the Train-elephant collision issue. It is a national-level issue, and a higher number of elephants die yearly. Approximately 47 elephants died from 2021 to 2024 due to a trainelephant collision. Therefore, we introduce a novel technical management system to avoid train-elephant collisions. The primary objective of this work is to explore the details of the issue and implement the system to repel the elephant using a real-time warning system. Simulation and hardware implementation were both carried out for the final outputs. In addition, introduce a communication system to make the train driver and the two nearby stations aware. The study demonstrates the potential real-time implementation system for the train-elephant national-level issue.