DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
The mountain was there: imagined landscapes, memory, and the expansion of photography
No full textThe Mountain Was There explores the construction of imagined landscapes through the lens of
memory and desire, questioning how one can belong to a place that does not physically exist. Drawing from
real landscapes encountered in August 2024, the work employs experimental darkroom techniques, collage,
and alternative processes such as coffee developing to create tactile, painterly photographic images. Guided
by philosophical ideas of interpretation and material transformation, the project treats photography as a
medium not of documentation but of introspection and emotional resonance. The resulting images— “photo
paintings”—reveal varied depths of looking into place, reflecting longing and the quiet pull of the unknown.
By positioning photography as a transformative process, this work offers a poetic and process-driven
framework for reimagining landscape as both memory and question, contributing new insights into the role
of image-making in constructing internal geographies.Bachelor's degre
IGZO-based transparent optoelectronic synapse for stealth mode trajectory tracking and registration
No full textWe report a transparent indium gallium zinc oxide (IGZO)-based optoelectronic synapse that exhibits strong persistent photoconductivity and tunable synaptic plasticity. The device, built on a quartz substrate with indium tin oxide electrodes, maintains over 70% transparency in the visible range, enabling stealthy operation. By modulating light duration, intensity, and frequency, we achieve key neuromorphic behaviors. A 3 × 3 device array further demonstrates trajectory tracking by mapping real-time ultraviolet illumination sequences. Notably, upon re-illumination, the device shows enhanced current, mirroring relearning in biological synapses. This work highlights the feasibility of IGZO-based synapses for next-generation transparent neuromorphic devices, providing new avenues for covert sensing for military purpose, interactive displays, and adaptive wearable electronics.Agency for Science, Technology and Research (A*STAR)Submitted/Accepted versionA*STAR, Singapore, Advanced Manufacturing and Engineering (AME) Individual Research Grant (IRG) (M23M6c0099)
Polarimetric tree radar measurements and signal processing
No full textSingapore, a modern city-state, is one of the greenest countries in the world, where Singapore is working towards creating a network of green spaces, transforming Singapore into a “City in a Garden”. The government and National Parks Board of Singapore (NParks) are working to plant trees to help provide shades to cool down harsh temperatures in the urban environment.
However, there were reports of tree falling over the years. Precious lives were lost, and some were injured due to the issue of tree falling. Therefore, there is a need to assess the health of the tree trunk by conducting measurements on tree trunk samples and process it with signal processing algorithms detect defects in the tree trunk.
In this project, several simulations, radar measurements have been conducted to analyse the defects in the tree trunk. Signal processing algorithms are applied to the data and results are generated to provide insights about the tree trunk condition based on the trunk’s interior defects.
More measurements and quantitative analysis need to be done to provide a comprehensive insight to make a more informed decision on preventive measures on fall tree issues.Bachelor's degre
Design optimization of permanent magnet synchronous motor
No full textModern electric vehicles and renewable energy systems require high-efficiency electric motors, which
drive continuous development of Permanent Magnet Synchronous Motors (PMSMs). However, the
reliance on rare earth materials, thermal challenges, electromagnetic losses, and mechanical stress present
significant obstacles to their widespread adoption. The paper investigates how design optimization
strategies can boost PMSMs' performance, lower material expenses, and establish better ecological
sustainability.
The research centers on decreasing dependence on rare earth magnets by developing new magnet
composition techniques and material replacements. It investigates progressive cooling systems and
methods to decrease electromagnetic losses, increase efficiency, and extend motor operational lifetime. The
investigation includes systematic enhancements of materials that fight against mechanical stresses to
achieve operational stability when operating at high speeds.
Different optimization techniques, such as multi-objective design and topology modification applied with
advanced control methods, allow engineers to reach optimal efficiency, durability, and cost-effectiveness
levels. Multiple tests through simulation and experimental procedures assess how healthy optimization
methods function.
These design improvements enable PMSMs to continue providing sustainable energy solutions and electric
mobility through their cost-effective and environmentally friendly operation. The research outcomes
advance sustainable high-performance electric motor design solutions that meet the technical requirements
of upcoming environmental and technological needs.Master's degre
Cementitious paint for energy-saving cooling
No full textWith the increasing effects of climate change, the demand for cooling technologies for buildings has significantly increased. However, traditional active cooling methods such as air conditioning rely on coal and gas-fired power, driving up greenhouse gas emissions and threatening our living conditions. Therefore, there is an urgent to develop eco-friendly and energy-saving cooling technologies.
To achieve the COP 26 goal of carbon neutrality and sustainability, passive cooling technologies are favored over active cooling methods. Daytime radioactive and evaporative
cooling are the primary methods utilised due to their energy and emission-free characteristics. However, the radioactive cooling power in tropical countries such as Singapore is more than
half due to its high relative humidity (84% on average) and strong atmospheric radiation. Hence, an effective cooling solution is necessary to maximise the available cooling power.
In the following pages, the report will introduce the theory behind passive daytime radiative cooling (PDRC) and evaporative cooling. Findings from previous research will also be
analysed to understand the limitations of previous passive cooling strategies. Studies on Integrated Passive Cooling Strategies on cementitious paint with optofluidic properties will be
explored, through various experiments to demonstrate the effectiveness of sub-ambient cooling even under Singapore's strong solar radiation. The process of product optimisation will also be
discussed and explored to prove that such technologies would be applicable commercially.Bachelor's degre
Enhancing top-down RGB image-based worker detection in challenging sea-port scenario
No full textIn modern seaport operations, ensuring the safety and efficiency of human workers remains a critical challenge. This dissertation focuses on enhancing worker detection accuracy from top-down RGB images captured in complex maritime environments, particularly under visually degraded conditions such as low-light, rain, and blur. By integrating image classification, enhancement, and retraining strategies, the proposed pipeline improves the reliability of small-object detection without requiring additional hardware.
A key characteristic of this work is its focus on top-down small-object detection. The RGB data used in this study were collected in advance by mounting cameras on the trolleys and spreaders of quay cranes in a real seaport. These cameras provide overhead surveillance footage of workers operating in close proximity to heavy container-handling equipment. The goal is to reduce the risk of accidents involving moving mechanical structures such as spreaders.
While depth sensors offer valuable spatial information, their limitations in adverse weather and low-light conditions necessitate complementary RGB-based approaches. In this work, a novel image enhancement and classification pipeline is proposed to identify and optimize challenging images such as those captured in low-light, rainy, or blurred scenarios. Using advanced enhancement models including DarkLighter and MPRNet, I pre-process the overhead seaport images before feeding them into a YOLOv11-based worker detection network.
To validate the effectiveness of these enhancements, I introduce a two-stage annotation strategy. In many low-light cases, original images suffer from annotation errors due to poor visibility. Therefore, I re-annotate the enhanced images to generate reliable ground truth and conduct comparative training and evaluation on both original and enhanced datasets. This allows me to assess whether the optimized images and their updated annotations lead to better detection performance.
The proposed system is evaluated on a custom-collected RGB dataset, and significant improvements are observed in mAP, precision, and recall metrics. These results confirm the benefit of image enhancement and re-annotation strategies in improving detection robustness for top-down, small-scale human targets under challenging environmental conditions. Future extensions of this work may integrate multi-modal fusion with depth sensors for enhanced 3D localization.Master's degre
Design of PCB-based wireless power transfer for underwater application
No full textThis dissertation presents a design scheme for a PCB-based coil structure tailored for lightweight and compact wireless power transfer (WPT) systems, targeting future underwater applications. By adding a relay unit on the transmitter side, the system forms a three-coil configuration that offers clear advantages over traditional two-coil designs. This structural enhancement improves both power transfer efficiency and output power, effectively addressing key limitations of conventional WPT architectures and enabling more stable energy delivery over extended distances. Experimental results verify the performance benefits of the proposed design in various environments. In air, the relay unit increases maximum efficiency by 3% and boosts output power by 2.94 times. In seawater, despite higher electromagnetic losses, the system achieves a 9% efficiency improvement and a 5.75-fold increase in power output. These findings confirm the effectiveness and robustness of the three-coil design in both terrestrial and underwater settings, highlighting its potential for applications such as powering autonomous underwater vehicles (AUVs) and other submerged electronic systems.Master's degre
A hydrogen embrittlement study of stainless steel
No full textThe prolonged dependence on fossil fuels has resulted in significant carbon emissions, speeding up the degradation of the environment and climate change. Due to the critical need to decrease carbon emissions, the switch to clean hydrogen being a sustainable energy source has become more urgent. Nevertheless, despite hydrogen’s potential, hydrogen-based application faces multiple challenge, with one of the most important
factors being hydrogen embrittlement. Hydrogen embrittlement is a vital challenge to materials used in transportation system and hydrogen storage as it degrades metallic structures through initiating cracks and premature failure. Stainless Steel 316 (SS316), a well-known material that is widely adopted for these applications are hydrogen embrittlement prone after extended exposure. It is curial to comprehend this vulnerability in order to guarantee safety and dependability of materials in hydrogen infrastructure. To assess the impacts of hydrogen embrittlement on stainless steel 316, this study involves conducting of Vickers hardness testing on pre-charged samples. This approach assesses hardness variations through hydrogen diffusion, providing insights on SS316 vulnerability to hydrogen embrittlement. By analysing hardness variations, the goal of this research is to deepen the understanding of hydrogen embrittlement behaviour. This study will also look into potential variables affecting the degree of embrittlement in order to contribute to refining previous established literature. The results of this research will offer insights on SS316’s susceptibility to hydrogen embrittlement and could help improve current models and mitigation methods for applications involving hydrogen. Addressing hydrogen embrittlement is important in advancing hydrogen as a substitute for fossil fuel, assuring its reliability and safety in energy systems.Bachelor's degre
Masculinity, online radicalisation, and the far right: a global perspective
No full textThe intersection of masculinity, online radicalisation, and far-right extremism has become an increasingly pressing concern, as digital platforms provide a gateway for disaffected young men to be drawn to extremist ideologies. Hyper-masculinity plays a central role in this process, with both global and local case studies demonstrating the transnational nature of far-right narratives. A comprehensive approach to counter-extremism is required, encompassing the redefinition of masculinity, the promotion of community engagement, and the regulation of online spaces to disrupt extremist recruitment strategies.Published versio
Machine learning for statistical arbitrage
No full textStatistical arbitrage strategies aim to exploit temporary price inefficiencies in financial markets using statistical techniques. While the existing literature has primarily focused on equities, this project extends statistical arbitrage to the foreign exchange (FX) market, specifically targeting G20 currencies. Our contributions are twofold:
1. Non-linear Return Decomposition: Using deep learning models, we propose a novel approach to decompose FX currency returns into statistical factors. These are compared against a traditional Principal Component Analysis (PCA) benchmark.
We then model each currency’s returns using Ordinary Least Squares (OLS) regression on these extracted factors. The regression residuals, which represent the unexplained component of returns, are modelled using an Ornstein-Uhlenbeck (OU) process to generate mean-reverting trading signals.
2. Deep Learning-Based Weight Allocation: We introduce a Long Short-Term Memory (LSTM) model to dynamically allocate portfolio weights based on the residuals and factor betas, aiming to enhance the strategy’s responsiveness to changing market conditions.
Our backtests over 10 years show that:
1. The PCA and Autoencoder approach achieves a consistent returns and significant Sharpe ratio when traded using the OU-based residual strategy with s-score thresholds for opening a position set at 1.25.
2. The LSTM-based weight allocation achieves significant Sharpe ratios, particularly when combined with PCA residuals as input.Bachelor's degre