DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Atomic rivers. The (un)sustainability of nuclear power in an age of climate change
No full textThe sustainability of nuclear energy amidst climate change and environmental regulations poses critical challenges, particularly in European contexts where major rivers like the Rhine, the Danube, and the Rhône are experiencing declining water levels and rising temperatures. We scrutinise the operational difficulties nuclear power plants encounter, arising from insufficient cooling water and environmental mandates that prevent the discharge of overly warm cooling water into rivers. These conditions have led to partial or full shutdowns of nuclear facilities across France, Germany, Switzerland, Belgium, Spain, Romania, and other countries, emphasising the tension between nuclear energy as a low-carbon solution and its environmental impacts. We explore the concept of sustainability in the context of riverine nuclear energy from three angles: technical challenges posed by water scarcity, regulatory constraints on cooling water temperatures, and the ecological impacts of thermal discharges on riverine ecosystems. Our analysis reveals an emerging contradiction between ensuring electricity supply and adhering to environmental protection, highlighting the need for a reevaluation of nuclear energy's role in a future sustainable energy landscape.Published versionThe research underlying this publication, which is part of the NUCLEARWATERS project (www.nuclearwaters.eu), was made possible with generous support from the European Research Council (grant no. 771928)
Efficient compression in 3D gaussian splatting
No full textRecent advancements in 3D Gaussian Splatting (3DGS) have made real-time photore-
alistic novel view synthesis by explicitly representing scenes using Gaussian primitive.
However, the substantial memory footprint required to store full feature embeddings
across multiple resolution levels is a significant barrier for scalable applications. In
this work, we explore a memory-efficient enhancement to the Hash-grid Assisted Con-
text (HAC) framework by integrating a softened dynamic codebook, inspired by the
success of Variable Bitrate Neural Field (VBR-NF) with a Straight-Through Estimator
(STE) based differentiable indexing mechanism. Our proposed method replaces con-
ventional hash embedding storing full feature embeddings at each spatial coordinate
with a learnable feature indices matrix and a global codebook, enabling memory effi-
ciency. Experimental results on datasets such as Tanks&Temples, DeepBlending and
BungeeNRF reveal that our method achieves comparable rendering quality to HAC
while significantly reducing memory usage at the embedding level. Nonetheless, the
overall memory falls short of the HAC baseline introduced by additional anchor at-
tributes. This suggests the trade-off between embedding level compression and memory
overhead in other elements. We analyze these challenges and propose future directions
for spatially adaptive indexing strategies to better balance efficiency and quality.Bachelor's degre
Comandeering foreign policy: how the Thai military overshadows the foreign ministry
No full textHistorically, Thailand’s foreign policy has been described as “bamboo diplomacy” – rooted in interest and values, while swaying to the geopolitical context of the world. Yet, beneath the perceived view of Thailand’s external foreign policy lies a group of actors that seek to influence and undermine the maintenance of a coherent foreign policy vision. This paper explores the following question: why has Thailand struggled to articulate and maintain a clear foreign policy vision, which often appeared inconsistent and reactive? It argues that Thailand’s inconsistent foreign policy is due to the military’s pursuit of its own interests, which often conflict with the elected government’s. Utilising the concept of a "parallel state,” this paper examines the case studies of the Preah Vihear Temple dispute as well as the opaque procurement practices of the Thai military to illustrate how self-interest and the quest of political survival can overshadow diplomatic efforts and national strategic goals, resulting in the emergence of a reactive and inconsistent foreign policy.Bachelor's degre
Federated fine-tuning of foundation ai models: lightweight computing approaches
No full textThis research investigates the intersection of parameter-efficient fine-tuning (PEFT) methods and federated
learning algorithms to address the computational, privacy, and performance challenges in deploying
foundation AI models. Through systematic evaluation of four PEFT approaches (LoRA, P-Tuning, Prefix
Tuning, and Discrete Prompt Tuning) across seven federated learning algorithms, we demonstrate taskspecific optimal combinations with significant efficiency gains. LoRA achieved superior performance on NLP
tasks (84.87% accuracy with SCAFFOLD), while P-Tuning excelled in computer vision tasks (65.39%
accuracy with FedAvg). Our experiments reveal that selective combinations reduce communication overhead
by up to three orders of magnitude while maintaining competitive accuracy. We identify novel algorithmic
synergies, such as LoRA-SCAFFOLD for language tasks and P-Tuning-FedAvg for vision tasks, and establish
theoretical principles governing PEFT-federated algorithm interactions. These findings provide a
comprehensive framework for efficiently deploying foundation models in resource-constrained, privacysensitive distributed environments, enabling wider accessibility of state-of-the-art AI capabilities across
diverse computational landscapes. Code and models are publicly available at https://github.com/Tan-Yu/FYPFederated-Learning.Bachelor's degre
Experimental study on the rail damper for vibration and noise control and rail corrugation suppression
No full textWith the rapid construction of rail transportation infrastructure, vibration issues caused by wheel–rail interactions during train operation have become a significant concern. Excessive vibration can lead to rail defects, such as corrugation, and also cause environmental vibration and noise problems in surrounding buildings. Rail dampers have been proposed to mitigate these issues by reducing rail vibration across multiple frequency bands. This paper presents a comprehensive experimental study on evaluating the effectiveness of rail dampers in reducing vibration and noise, as well as in suppressing the development of rail corrugation. Specifically, numerical simulations were conducted to design the optimal mass distribution of the rail dampers. The designed dampers were then installed on two operating railway sections, and the dynamic characteristics of the rails were measured and analyzed before and after rail damper installation to demonstrate their effectiveness in mitigating train-induced vibration and noise. Additionally, rail corrugation was measured over a period of 463 days, comparing development with and without rail damper installation. The experiments demonstrate the practical effectiveness of rail damper in mitigating rail vibration issues, and these findings can serve as a reference for related research and engineering applications.Published versionThis work is supported by Beijing Natural Science Foundation (No. L221023), Science and Technology Project of Gansu Province (No.22CX8JA142), and Research Program of Beijing Infrastructure Investment Co., Ltd. (No.2022-10-06-01)
An app to encourage users to engage in interdisciplinary learning
No full textThis paper presents the development of an application to encourage users to engage in interdisciplinary learning. By assessing interdisciplinary-natured essays, the interdisciplinary analytical tool will leverage LLMs to generate feedback for the users. As a modern leading-edge application that harnesses the potential of Artificial Intelligence, it aims to address the gaps in conventional assessment frameworks and embrace interdisciplinary learning, a concept growing in importance amidst a dynamically evolving world. This work will also explore different assessment methodologies, such as the use of contextualisation in analysis and the effects of LLM rationalisation on feedback quality through the Self-Taught Reasoner (STaR) framework. The user studies, which involved students who had taken the CC0002 course in NTU, offer much positivity and optimism for the potential of such a tool in the classroom. The findings outline how Generative AI can be used effectively as a learning assistant to support students in their interdisciplinary coursework, as well as provide insights into how this tool can be improved upon and scaled up to cover a larger area of their university education.Bachelor's degre
A β-galactosidase-activatable photosensitiser for photodynamic therapy in glioblastoma
No full textGlioblastoma (GBM) is one of the most aggressive brain cancers, characterised by high tumour recurrence rates and poor patient outcomes despite administration of common treatment methods. Photodynamic therapy (PDT) has emerged as a promising treatment alternative which utilises a combination of light, oxygen, and photosensitisers (PS) to produce reactive oxygen species (ROS) for targeted cell death. Unfortunately, the clinical efficacy of 5-aminolevulinic (5-ALA), an existing PDT agent, remains variable, inspiring the development of new PSs such as SeNBD-oleic developed by the Vendrell group for PDT of GBM. Despite its good phototoxicity and rapid uptake, its lack of specificity for GBM limits its therapeutic potential. To overcome this, an activatable PS with increased uptake in GBM cells was designed, utilising an ‘OFF’ to ‘ON’ mechanism to limit PS activation to target cells. This project therefore reports the synthesis and characterization of β-galactosidase-activatable Gal-SeNBD-oleic, a caged derivative of the always ‘ON’ PS, SeNBD-oleic. Gal-SeNBD-oleic incorporates a β-galactosidase-cleavable D-galactose moiety and a benzyl carbamate self-immolative linker to quench reactive oxygen species (ROS) production in its inactive state. Enzymatic assays indicated selective uncaging of Gal-SeNBD-oleic in the presence of β-galactosidase, restoring its ROS generation and optical properties. Moreover, in vitro studies with E17 glioma stem cells demonstrated uptake and activation of the PS, leading to significant cytotoxicity under light irradiation. Taken together, these findings illustrate the potential of Gal-SeNBD-oleic as a phototoxic and selective PS for GBM PDT. Future work will focus on optimizing synthesis yields and evaluating the efficacy and selectivity of Gal-SeNBD-oleic in ex vivo and in vivo GBM models.Bachelor's degre
Health technology for all: examining health apps and wearables use through the lens of (in)equality
No full textThe rapid proliferation of health apps and wearables presents significant opportunities to enhance physical and mental health. However, the benefits of these technologies are not equally distributed, particularly among individuals with lower educational levels and older adults. This thesis investigates these disparities by integrating the extended Technology Acceptance Model (TAM) with Communication Inequality, exploring how technological, individual, and social factors influence the adoption and use of health apps and wearables among populations facing digital challenges.
Study One addresses a critical theoretical gap through a scoping review of frameworks and adoption antecedents in health apps and wearables research. Analyzing 61 empirical studies published between 2007 and 2022, the review reveals that while human-computer interaction theories dominate the literature, social factors and broader environmental influences receive limited attention. This insight underscores the need for a more integrated theoretical framework that incorporates these dimensions, setting the stage for the empirical investigations that follow.
Study Two builds on these theoretical findings by integrating TAM with Communication Inequality to examine how technological, individual, and social factors influence adoption across educational levels. Using data from a national survey in Singapore (N = 906), the study identifies promising pathways for narrowing adoption gaps, particularly through technological attributes (e.g., perceived usefulness and design aesthetics) and social norms (e.g., descriptive and injunctive norms). The results also suggest that health apps and wearables use may help reduce disparities in social well-being, highlighting the potential of targeted interventions for vulnerable populations.
Study Three responds to the need for translating the theoretical understanding to practical design considerations by employing a multi-stakeholder consultation approach involving older adults, caregivers, and community center managers. This process identifies key requirements for developing accessible health technologies, such as usability, data security, social support, and design preferences. These findings provide actionable insights for designing interventions that address the specific needs of older adults, a population often underrepresented in mainstream technology development.
Study Four operationalizes these insights through a four-week community-based intervention (N = 48) that explores conditions for enhancing health benefits among older adults in Singapore. The intervention compares four scenarios: (a) conventional exercise, (b) individual exergaming, (c) exergaming with a health coach, and (d) exergaming with peers. The findings demonstrate that gamification elements in health apps and wearables enhance engagement and improve physical and mental well-being. While health coach support appears to boost technology adoption and physical activity, peer collaboration proves particularly effective in enhancing emotional well-being, offering actionable strategies for narrowing digital health gaps among older adults.
By integrating the Technology Acceptance Model and Communication Inequality framework, this thesis advances the theoretical understanding of health apps and wearables adoption and provides practical insights for designing inclusive digital health interventions. The findings underscore the importance of user-centered approaches and social influences in creating accessible health technologies for diverse populations. These contributions also inform healthcare providers and technology developers in addressing inequalities and promoting equitable health outcomes in an increasingly digital healthcare landscape.Doctor of Philosoph
Misorientation and dislocation evolution in rapid residual stress relaxation by electropulsing
No full textThis study investigates the effect of high current density electropulsing on the material in a rapid stress relaxation process. An AISI 1020 steel was shot-peened to induce surface compressive residual stresses in a controlled manner and subsequently electropulsed to investigate the changes in microstructure and defect configuration. AISI 1020 steel was chosen as it has a simple microstructure (plain ferritic) and composition with low alloying conditions. It is an appropriate material to study the effect of transmitting electric pulses on the microstructural defect evolution. A combination of electron-backscattered diffraction and transmission electron microscopy proved to be an effective tool in characterizing the post-electropulsing effects critically. By application of electropulsing, a reduction in the surface residual stress layer was noticed. Also, reductions in misorientation and dislocation density together with the disentanglement of dislocations within the cold-worked layer were observed after electropulsing. Additionally, the annihilation of shot-peening-induced deformation bands beyond the residual layer depth was observed. These effects have been rationalised by taking into account the various possibilities of athermal effects of electropulsing.Nanyang Technological UniversityNational Research Foundation (NRF)This work was financially supported by the National Research Foundation of Singapore, Rolls-Royce Singapore Pte. Ltd., and Nanyang Technological University through grants #002123-00009 and #002124-00009
An efficient strategy to enhance air filtration through the synergistic effects of ultrasonics and seed particles
No full textThe individual and combined effects of ultrasonic acoustics agglomeration (AA) with the addition of seeding water droplets for particle agglomeration (DA) are demonstrated for their efficiency in manipulating the number concentration of airborne particles as a pre-treatment mechanism in filtration systems. Through a series of experiments in a purpose-built aerosol wind tunnel with particle separation and filtration stages, it is demonstrated that the combined effect of AA and DA can reduce the particle number concentration by up to 30 % as a result of extensive particle agglomeration. This reduction is significantly higher than that achieved either by the individual action of AA or DA and higher than the direct superposition of the reductions achieved individually by these two actions. In addition, the synergistic effects of combining AA and DA can extend filter operating lifespan significantly by up to 57 % due to lower mass deposition in the filter. A theoretical model is developed for the prediction of filters’ pressure drop patterns under various pre-conditioning methods. The current study sheds light on the development of filtration pre-conditioning systems that are capable of enhancing particle removal efficiencies alongside reduced pressure drop increment across filters, leading to both environmental and cost benefits.Energy Market Authority (EMA)National Research Foundation (NRF)Submitted/Accepted versionThis study is supported by the National Research Foundation, Singapore, and the Energy Market Authority, under its Energy Program (EP Award EMA-EP009-SEGC-007)