DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Assessment of effective stiffness in reinforced concrete walls subjected to reversed cyclic loading
No full textThe lateral load resistance of reinforced concrete (RC) walls is fundamental to structural integrity
in seismic and wind-prone regions. Precise determination of their effective stiffness under
reversed cyclic loading is essential for design, but current methods often lack accuracy due to
simplifications of cracking behaviour. This research addresses this by utilizing machine learning,
specifically Support Vector Regression (SVR), to model the effective stiffness. A comprehensive
database of 361 RC wall test results was compiled for training. Through comparative analysis of
feature selection and kernel choices, the SVR model employing the Radial Basis Function (RBF)
kernel and Recursive Feature Elimination (RFE) demonstrated superior predictive capability,
outperforming established models from Paulay & Priestley (1992) , Li & Xiang (2011) , and
Fenwick and Bull (2000).
The study demonstrated a substantial improvement in effective stiffness prediction through
machine learning. To understand the SVR-RBF-RFE model's predictions, SHAP analysis was
used, quantifying the influence of each input feature.
The findings of this study advocate for the adoption of machine learning as a superior method for
predicting the effective stiffness of RC walls under reversed cyclic loading. This highlights the
transformative potential of integrating data-driven techniques into structural engineering, paving
the way for more precise and reliable assessments.Bachelor's degre
Automating investigation checks for rehandybot, an upper limb rehabilitation device
No full textRobotic-assisted rehabilitation has become an imperative tool in physiotherapy, helping patients
recover from neurological and musculoskeletal impairments. Devices like the ReHandyBot are
used to assist guided upper limb exercises but require extensive manual investigation before they
can be deployed to customers. This investigation takes place in the manufacturing site and it
requires operators 60-70 manual intervention steps to complete, depending on the operator’s
technical expertise and familiarity with the device. This high frequency of manual interventions
introduces the potential for human error which can affect the consistency and accuracy of the
testing procedure.
The main aim of this project was to automate the investigative testing process of ReHandyBot by
leveraging the device’s inbuilt sensors and actuators. A software-driven solution was developed
to streamline the setup and reduce manual intervention while maintaining accuracy and safety.
The automation system interfaces with the device’s microprocessor to read sensor values, adjust
variables and verify data. This reduces human error by decreasing the reliance on manual input.
Unlike adaptive calibration methods used in other rehabilitation systems, the proposed solution
offers a scalable approach to automate its testing due to its compatibility with ReHandyBot’s
existing hardware.
Through the implementation of the automation system, the system successfully reduced manual
intervention to under 25 steps which essentially enhanced the consistency and repeatability of the
testing process while ensuring accuracy and safety. Despite these improvements, there are some
limitations that need to be addressed in future works. The most significant is the lack of cloud
connectivity for updates and scalability. Moreover, the system lacks built-in security measures,
especially encryption or containerization, which are necessary for broader deployment.
This project contributes to the efficiency of robotic rehabilitation systems by demonstrating that
automation can significantly reduce the effort and resources involved in investigative testing.
Future work will focus on further enhancing automation along with the integration of cloud
based updates to improve the system’s scalability and deployment efficiency.Bachelor's degre
Vision-based learning for drones: a survey
No full textDrones, as advanced cyber-physical systems (CPSs), are undergoing a transformative shift with the advent of vision-based learning, a field that is rapidly gaining prominence due to its profound impact on drone autonomy and functionality. Unlike existing task-specific surveys, this work offers a comprehensive overview of vision-based learning for drones, emphasizing its pivotal role in enhancing their operational capabilities across various scenarios. First, the fundamental principles of vision-based learning are elucidated, demonstrating how it significantly improves drones’ visual perception and decision-making processes. Vision-based control methods are then categorized into indirect, semidirect, and end-to-end approaches from the perception-control perspective. Various applications of vision-based drones with learning capabilities are further explored, ranging from single-agent systems to more complex multiagent and heterogeneous system scenarios, while highlighting the challenges and innovations characterizing each domain. Finally, open questions and potential solutions are discussed to guide future research and development in this dynamic and rapidly evolving field. With the growth of large language models (LLMs) and embodied intelligence, vision-based learning for drones provides a promising yet challenging road toward achieving artificial general intelligence (AGI) in the 3-D physical world.Economic Development Board (EDB)Submitted/Accepted versionThis work was supported by the Economic Development Board (EDB) through Space Technology Development Programme (STDP) Thematic Grant Call on Space Technologies under Award S23-020019-STDP.
Acts of resistance: exploring undergraduate medical students’ experiences with social spaces
No full textScholars in higher education have long theorised the relational and political dimensions of social spaces, including their potential impact on students’ wellbeing, sense of belongingness and identity, but this is a largely unexplored topic in health professions literature. To address this gap, we explored how student common rooms in a medical school were allocated and managed, and how these processes shaped student experiences. Our specific research question was: how does the design and management of student social space influence students’ experience of this space? We adopted a constructivist perspective and drew upon Lefebvre’s idea of space as a sensitising lens to guide our analysis.Published versio
Differential optical imaging of antigen presentation machinery using molecular optical reporters
No full textDetection of antigen presentation is central to understanding immunological processes and developing therapeutics for cancer, infectious diseases, and allergies. However, methods with the ability to dynamically and noninvasively distinguish between major histocompatibility complex class I (MHC-I) and MHC-II antigen presentations remain lacking. Herein, we develop activatable molecular optical reporters (MORs) for real-time differential imaging of antigen presentations in lymph nodes (LNs). These MORs are engineered to passively target LNs and activated through proteolytic cleavage by key enzymes in the MHC-I and MHC-II pathways, the immunoproteasome (iP) and cathepsin S (CTSS), respectively, triggering their chemiluminescent or fluorescent signals. Coupled with minimized signal crosstalk and high sensitivity, MORs delineate the subtle differences in the antigen presentation machinery across various disease models, including cancer and bacterial or viral infection, a feat unattainable for existing imaging methods. After systemic administration, MORs also allow real-time visualization of antigen presentation in the tumor microenvironment. Besides, MORs are validated to have potential for preclinical application in immunotherapeutics screening and clinical application in tissue biopsy. Thus, our study not only presents the first example of real-time, in vivo differential imaging of antigen presentation pathways but also opens new avenues for optical probes in immune contexture analysis.National Research Foundation (NRF)Ministry of Education (MOE)Submitted/Accepted versionY.Z. thanks the National Natural Science Foundation of China (22322406)for financial support. K.P. thanks Singapore National Research Foundation(NRF) (NRF-NRFI07-2021-0005), Singapore Ministry of Education, andAcademic Research Fund Tier 2 (MOET2EP30220-0010; MOE-T2EP30221-0004) for the financial support
Automated fabrication of 3D printed device for cellular therapy for Type 1 Diabetes
No full textThe use of macro-encapsulation platforms for cellular therapeutic delivery in Type 1 diabetes management offers advantages such as enhanced cell loading capacity and ease of device retrieval. In our laboratory, a macro-device has been developed to facilitate the even spatial arrangement of therapeutic islet tissues through the implementation of micro-patterned designs, addressing the common issue of limited oxygen and nutrient diffusion caused by micro-tissue clumping in traditional macro-encapsulation approaches. However, the current fabrication process is manual, making it prone to variability, human error, material waste, and inefficiencies in time and reproducibility. To address these limitations, we propose the automation of the device fabrication process using a 3D bioprinter. This study explores the feasibility of automating key steps such as GelMA coating and cell seeding, aiming to reduce variability while maintaining or improving performance relative to the manual standard. Our findings serve as proof of concept for an automated fabrication workflow, advancing the device toward a robust, reproducible, and clinically translatable solution for Type 1 diabetes treatment.Bachelor's degre
Adaptive edge intelligence for rapid structural condition assessment using a wireless smart sensor network
No full textCombining artificial intelligence and edge computing, edge intelligence is a promising computing paradigm for the Internet-of-Things-based Structural Health Monitoring (SHM), showing great potential to improve system responsiveness by reducing communication latency. Previously, very limited studies proposed, optimized, or verified edge intelligence approaches for SHM applications, where the overhead and efficiency of algorithms to manage limited onboard resources are the main gaps. In this study, an adaptive edge intelligence strategy is proposed to facilitate autonomous structural condition assessment, involving reference-free displacement estimation algorithm, Gaussian Process Regression, and stochastic process control. To facilitate algorithm deployment, both effective single-node independent computing and multi-node coordination are explored to deal with the limited onboard resources, utilizing the computing capacity of each node to speed up computation. Using the Xnode, a MEMS-based wireless sensor platform, lab tests and full-scale applications in railroad bridge monitoring were conducted to verify the proposed strategy, demonstrating the potential and suitability of the developed approach for rapid adaptive structural condition assessment in SHM practice.Ministry of Education (MOE)Nanyang Technological UniversityThe authors want to gratefully acknowledge the Federal Railroad Administration (FRA) for the financial support of this research under contract DTFR53-17-C-00007, and ZJU-UIUC Institute Research under Grant #ZJU083650, NTU Start-up Grant 021323-00001, MOE AcRF Tier 1 Grants, No. RG121/21
Azadihomocorannulene as a heptagon-embedded diradicaloid
No full textPolycyclic aromatic diradical(oid) molecules are attracting significant attention because of their unique electronic and magnetic properties as well as their applications as functional materials. While diradical(oid) molecules bearing five-membered rings have been extensively investigated, those bearing seven-membered rings are relatively fewer. Herein, we report the synthesis of azapentabenzodihomocorannulene dication and diradical molecules. The synthesis was achieved through a mechanochemical C(sp2)-H/C(sp3)-H coupling in the presence of sodium as a key reaction. Electron spin resonance studies revealed that the neutral azapentabenzodihomocorannulene adopts a singlet diradical (diradicaloid) ground state with a small singlet-triplet energy gap of 2.1 kcal/mol. The electronic and optical properties were investigated both experimentally and theoretically to elucidate their aromatic character.Ministry of Education (MOE)Nanyang Technological UniversityThis work was supported by Nanyang Technological University and the Ministry of Education, Singapore, under its Academic Research Fund Tier 1 (RG2/23 for S.I.)
Facilitated ammonia decomposition and enhanced hydrogen diffusion in 10Ni-Ce0.8Zr0.2O2 as anode catalytic functional layer for low-temperature direct ammonia fuel cells
No full textDue to higher volumetric energy density and more efficient transportation of ammonia in contrast to hydrogen, ammonia has propelled DAFCs into the spotlight of research and development in recent years. To enhance their performance, a novel design strategy is employed by introducing the catalytic functional layer onto the surface of anode in single cells to substitute traditional anode material modification approaches. In this study, xNi-Ce0.8Zr0.2O2 (xNi-CZ, x = 5 wt%, 10 wt%, 15 wt%) catalysts are synthesized and their phase structures are analyzed. The physiochemical properties including, microstructure, reduction capability, ammonia adsorption capacity, elemental valence states and ammonia decomposition conversions for as-prepared catalysts are characterized and evaluated by SEM&TEM, H2-TPR, NH3-TPD, XPS, fixed-bed reactor, respectively. Based on the TEM images, a theoretical calculation model is developed by utilizing DFT, and the reaction pathways and rate-determining steps for ammonia decomposition are analyzed and determined. Moreover, the 10Ni-CZ|Ni-YSZ|YSZ|GDC|LSCF-GDC structure is constructed by employing 10Ni-CZ as a catalytic functional layer, and the impedances and power outputs are tested at 550–650 °C. Experimental results indicate that the incorporation of 10Ni-CZ in LT-DAFCs can notably reduce the impedance by 24.46 % and significantly increase the maximum power density by 11.04 % at 650 °C as expected, demonstrating that adding 10Ni-CZ could be a highly effective and practical strategy for advancing LT-DAFCs technology.Singapore Maritime Institute (SMI)This work was financially supported by the National Natural Science Foundation of China (22078262, 52336009), the Singapore Maritime Institute through the Maritime Transformation Programme (SMI-2023- MTP-02) and the cooperative R&D Project from Shaanxi Yanchang Petroleum Northwest Rubber Co., Ltd. (2022610002005308)
Controllable neural text generation in persona-based dialogue systems
No full textPersona-based dialogue system, a specialized area of open-domain dialogue system, focuses on creating more engaging and personalized conversational agents. Distinguished from the open-domain dialogue system, it incorporates a consistent and distinct persona that directs its interaction style, which is either explicitly defined or learned from large datasets of human interactions. The recent advance of pre-trained language models boosts the dialogue system, significantly enhancing the user experience by delivering fluent, diversified, and human-like conversations. However, the challenge of maintaining persona consistency and contextual coherence still persists in persona-driven dialogue generation, due to the overlook of user intent and persona identification. Therefore, it is proposed to create a pipeline dialogue system consisting of natural language understanding module for precise user intent detection and persona extraction, and dialogue management and generation module for multiple source inputs integration. The research goals of this thesis include controllable sentence generation for persona-consistent and contextual coherent conversation generation, controllable structural extraction by exploiting persona attributes in conversation histories to facilitate natural language understanding in a dialogue system, and controllable data synthesis for low-resource extraction scenario.
Natural language understanding and dialogue management and generation, which elevated the performance of a task-oriented dialogue system, are introduced to the persona-based dialogue system. In the first work, a lightweight hierarchical intent-inferring pointer network is proposed for multi-source persona-driven dialogue generation. For natural language understanding, the proposed method involves detecting interlocutor intents, retrieving the most relevant persona attributes in chitchat, and utilizing pseudo-labeling and natural language inference techniques to generate intent labels. For dialogue management and generation, a multi-source pointer-generator is designed to leverage the useful information from multiple input sources and filter out irrelevant textual noises.
However, matching the relevant persona attribute is a retrieval-based method and requires predefined persona information, nontransferable to realistic conversations. This motivates the second framework, which formulates and investigates persona attribute extraction from dialogues utilizing pre-trained language models in generalized zero-shot setting. Moreover, a Meta-Variational-AutoEncoder sampler with contrastive structured constraint is presented to tackle the hard negative samples in this task. By leveraging more reliable text-label matching criteria, a task-specific dataset with more detailed relation types and consistent entity relation annotation is created for persona attribute extraction. The aforementioned method is applicable to cases when seen data and unseen relation types are available. A more stringent condition has not been explored where no information about the unseen data is accessible, which is genuine zero-shot persona attribute extraction. In the third work, thus, a Chain-of-Proposal prompting framework is developed for unseen relation generation, controllable support sentence synthesis conditioned on hypothesized relations, synthetic data denoising, and zero-shot persona triplet extraction. Extensive experiments on benchmark and built datasets demonstrate the superior performance of the proposed frameworks compared to strong baseline models. A further discussion and empirical study is conducted on the potential constructive effects of persona attribute understanding on personalized conversational sentiment analysis.Doctor of Philosoph