Filters
Availability
Publication Year
Resource Type
Faculty
School/Institute
Publication Search Results
Sort
Results per page
1 - 10 of 179
-
(2022)ThesisDecline in cognitive functions including memory, processing speed and executive processes, has been associated with ageing for sometime. It is understood that every human will go through this process, but some will go through it faster, and for some this process starts earlier. Differentiating between cognitive decline due to a pathological process and normal ageing is an ongoing research challenge. According to the definition of the World Health Organization (WHO), dementia is an umbrella term for a number of diseases affecting memory and other cognitive abilities and behaviour that interfere significantly with the ability to maintain daily living activities. Although a cure for dementia has not been found yet, it is often stressed that early identification of individuals at risk of dementia can be instrumental in treatment and management. Mild Cognitive Impairment (MCI) is considered to be a prodromal condition to dementia, and patients with MCI have a higher probability of progressing to certain types of dementia, the most common being Alzheimer's Disease (AD). Epidemiological studies suggest that the progression rate from MCI to dementia is around 10-12\% annually, while much lower in the general elderly population. Therefore, accurate and early diagnosis of MCI may be useful, as those patients can be closely monitored for progression to dementia. Traditionally, clinicians use a number of neuropsychological tests (also called NM features) to evaluate and diagnose cognitive decline in individuals. In contrast, computer aided diagnostic techniques often focus on medical imaging modalities such as magnetic resonance imaging (MRI) and positron emission tomography (PET). This thesis utilises machine learning and deep learning techniques to leverage both of these data modalities in a single end-to-end pipeline that is robust to missing information. A number of techniques have been designed, implemented and validated to diagnose different types of cognitive impairment including mild cognitive impairment and its subtypes as well as dementia, initially directly from NM features, and then in fusion with medical imaging features. The novel techniques proposed by this thesis build end-to-end deep learning pipelines that are capable of learning to extract features and engineering combinations of features to yield the best performance. The proposed deep fusion pipeline is capable of fusing data from multiple disparate modalities of vastly different dimensions seamlessly. Survival analysis techniques are often used to understand the progression and time till an event of interest. In this thesis, the proposed deep survival analysis techniques are used to better understand the progression to dementia. They also enable the use of imaging data seamlessly with NM features, which is the first such approach as far as is known. The techniques are designed, implemented and validated across two datasets; an in-house dataset and a publicly available dataset adding an extra layer of cross validation. The proposed techniques can be used to differentiate between cognitively impaired and cognitively normal individuals and gain better insights on their subsequent progression to dementia.
-
(2022)ThesisAs a dominant terrestrial ecosystem of the Earth, forest environments play profound roles in ecology, biodiversity, resource utilization, and management, which highlights the significance of forest characterization and monitoring. Some forest parameters can help track climate change and quantify the global carbon cycle and therefore attract growing attention from various research communities. Compared with traditional in-situ methods with expensive and time-consuming field works involved, airborne and spaceborne remote sensors collect cost-efficient and consistent observations at global or regional scales and have been proven to be an effective way for forest monitoring. With the looming paradigm shift toward data-intensive science and the development of remote sensors, remote sensing data with higher resolution and diversity have been the mainstream in data analysis and processing. However, significant heterogeneities in the multi-source remote sensing data largely restrain its forest applications urging the research community to come up with effective synergistic strategies. The work presented in this thesis contributes to the field by exploring the potential of the Synthetic Aperture Radar (SAR), SAR Polarimetry (PolSAR), SAR Interferometry (InSAR), Polarimetric SAR Interferometry (PolInSAR), Light Detection and Ranging (LiDAR), and multispectral remote sensing in forest characterization and monitoring from three main aspects including forest height estimation, active fire detection, and burned area mapping. First, the forest height inversion is demonstrated using airborne L-band dual-baseline repeat-pass PolInSAR data based on modified versions of the Random Motion over Ground (RMoG) model, where the scattering attenuation and wind-derived random motion are described in conditions of homogeneous and heterogeneous volume layer, respectively. A boreal and a tropical forest test site are involved in the experiment to explore the flexibility of different models over different forest types and based on that, a leveraging strategy is proposed to boost the accuracy of forest height estimation. The accuracy of the model-based forest height inversion is limited by the discrepancy between the theoretical models and actual scenarios and exhibits a strong dependency on the system and scenario parameters. Hence, high vertical accuracy LiDAR samples are employed to assist the PolInSAR-based forest height estimation. This multi-source forest height estimation is reformulated as a pan-sharpening task aiming to generate forest heights with high spatial resolution and vertical accuracy based on the synergy of the sparse LiDAR-derived heights and the information embedded in the PolInSAR data. This process is realized by a specifically designed generative adversarial network (GAN) allowing high accuracy forest height estimation less limited by theoretical models and system parameters. Related experiments are carried out over a boreal and a tropical forest to validate the flexibility of the method. An automated active fire detection framework is proposed for the medium resolution multispectral remote sensing data. The basic part of this framework is a deep-learning-based semantic segmentation model specifically designed for active fire detection. A dataset is constructed with open-access Sentinel-2 imagery for the training and testing of the deep-learning model. The developed framework allows an automated Sentinel-2 data download, processing, and generation of the active fire detection results through time and location information provided by the user. Related performance is evaluated in terms of detection accuracy and processing efficiency. The last part of this thesis explored whether the coarse burned area products can be further improved through the synergy of multispectral, SAR, and InSAR features with higher spatial resolutions. A Siamese Self-Attention (SSA) classification is proposed for the multi-sensor burned area mapping and a multi-source dataset is constructed at the object level for the training and testing. Results are analyzed by different test sites, feature sources, and classification methods to assess the improvements achieved by the proposed method. All developed methods are validated with extensive processing of multi-source data acquired by Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR), Land, Vegetation, and Ice Sensor (LVIS), PolSARproSim+, Sentinel-1, and Sentinel-2. I hope these studies constitute a substantial contribution to the forest applications of multi-source remote sensing.
-
(2022)ThesisAdhesively bonded joints have been widely used to manufacture aircraft components. However, its application to single load-path airframe structure is costly to certify as extensive validation testing is required. Certification of bonded joints or patch repairs for primary aircraft structures requires demonstration of damage tolerance. In recent years, a damage slow growth management strategy has been considered acceptable by Federal Aviation Administration to reduce the maintenance cost. This thesis evaluates the applicability of a damage slow growth management strategy to bonded joints/patch repairs of primary aircraft structures through both experimental and computational study. The investigation was carried out first by 2-D strip specimen assessment and finally using 3-D analysis of wide bonded metal joint. This research was a collaborative project between ARC Training Centre for Automated Manufacture of Advanced Composites (AMAC) at the University of New South Wales (UNSW) and Defence Science and Technology (DST) Group. The double overlap tapered end specimen (DOTES) specimen which represents both disbond tolerant zone and safe-life zone in bonded patch repair was investigated first through a detailed computational and experimental investigation. The residual static strength of the joint as a function of disbond length was established using finite element modelling based on the characteristic distance approach. The virtual crack close technique (VCCT) approach was utilised to assess the strain energy release rates (SERRs) as a function of disbond crack length. Fatigue tests of the DOTES coupon specimen were conducted to investigate the entire process of disbond growth from initiation up to ultimate failure of the joint. The measured disbond growth rates were correlated with the SERRs using a modified Paris law that enabled prediction of joint fatigue life. The fatigue test results indicated that for a joint having a sufficient static strength safety margin under a typical fatigue loading that would propagate disbond, the disbond growth would remain stable within a particular length range. Thus, the slow growth approach would be feasible for bonded joints/patch repairs if the patch is designed to be sufficiently large to allow extended damage propagation. Cohesive zone element (CZE) technique was utilised to assess the SERRs and estimate the disbond growth of 3-D wide bonded metal joint analysis. The impact of local or partial width disbond (load shedding effect) was investigated in detail. The results indicate that for a local or part width disbond, some load was redistributed to the adjacent regions (load shedding effect) that causes a slower disbond growth and accordingly longer fatigue life compared to the full width disbond. The key outcomes from this research are: (a) accurate prediction of the disbond growth behaviour in bonded patch repairs through the developed generic patch repair specimen i.e DOTES, (b) fatigue life prediction of the joints has been established through modified Paris law, by conducting numerical integration and (c) the effect of initial disbond size in 3-D wide bonded metal joint specimen was investigated through computational assessment using a cohesive fatigue model.
-
(2022)ThesisElements from cognitive psychology have been applied in a variety of ways to artificial intelligence. One of the lesser studied areas is in how episodic memory can assist learning in cognitive robots. In this dissertation, we investigate how episodic memories can assist a cognitive robot in learning which behaviours are suited to different contexts. We demonstrate the learning system in a domestic robot designed to assist human occupants of a house. People are generally good at anticipating the intentions of others. When around people that we are familiar with, we can predict what they are likely to do next, based on what we have observed them doing before. Our ability to record and recall different types of events that we know are relevant to those types of events is one reason our cognition is so powerful. For a robot to assist rather than hinder a person, artificial agents too require this functionality. This work makes three main contributions. Since episodic memory requires context, we first propose a novel approach to segmenting a metric map into a collection of rooms and corridors. Our approach is based on identifying critical points on a Generalised Voronoi Diagram and creating regions around these critical points. Our results show state of the art accuracy with 98% precision and 96% recall. Our second contribution is our approach to event recall in episodic memory. We take a novel approach in which events in memory are typed and a unique recall policy is learned for each type of event. These policies are learned incrementally, using only information presented to the agent and without any need to take that agent off line. Ripple Down Rules provide a suitable learning mechanism. Our results show that when trained appropriately we achieve a near perfect recall of episodes that match to an observation. Finally we propose a novel approach to how recall policies are trained. Commonly an RDR policy is trained using a human guide where the instructor has the option to discard information that is irrelevant to the situation. However, we show that by using Inductive Logic Programming it is possible to train a recall policy for a given type of event after only a few observations of that type of event.
-
(2022)ThesisEmerging modes of air transport such as autonomous airport shuttle and air taxi are potentially efficient alternatives to current transport practices such as bus and train. This thesis examines bus shuttle service within an airport and air metro as two examples of network design. Within an airport, the bus shuttle serves passengers between the terminals, train stations, parking lots, hotels, and shopping areas. Air metro is a type of pre-planned service in urban air mobility that accommodates passengers for intra- or inter-city trips. The problems are to optimise the service, and the outputs including the optimal fleet size, dispatch pattern and schedule. Based on the proposed time-space networks, the service network design problems are formulated as mixed integer linear programs. The heterogeneous multi-type bus fleet case and stochastic demand case are extended for the airport shuttle case, while a rolling horizon optimisation is adopted for the air metro case. In the autonomous airport inter-terminal bus shuttle case, a Monte Carlo simulation-based approach is proposed to solve the case with demand stochasticity, which is then further embedded into an "effective" passenger demand framework. The "effective" demand is the summation of mean demand value and a safety margin. By comparing the proposed airport shuttle service to the current one, it is found that the proposed service can save approximately 27% of the total system cost. The results for stochastic problem suggest estimating the safety margin to be 0.3675 times of the standard deviation brings the best performance. For the second case, the service network design is extended with a pilot scheduling layer and simulation is undertaken to compare the autonomous (pilot-less) and piloted service design. The results suggest that an autonomous air metro service would be preferable if the price of an autonomous aircraft is less than 1.6 times the price of a human-driven one. The results for rolling horizon optimisation suggest to confirm the actual demand at least 45 minutes prior to departure. Based on data from the Sydney (Australia) region, the thesis provides information directly relevant for the service network design of emerging modes of air transport in the city.
-
(2022)ThesisIn the as-built condition, laser powder bed fusion (LPBF) Ni superalloy 718 and electron beam melted (EBM) titanium aluminide (TiAl) have distorted, non-equilibrium structures that have negative consequences for the mechanical properties. These issues can be mitigated through post build heat treatments, and opportunities also exist to create additive components that exceed the properties of conventional ones. Fatigue crack growth (FCG) performance of LPBF 718 was characterized after applying two common post-build heat treatments. The solution and duplex aging (S+DA) treatment retained many of the features of the as-built condition including large residual stresses, an elongated, textured structure, and solute segregation and lattice distortion within grains. Applying hot isostatic pressing (HIP) prior to the S+DA treatment resulted in complete relief of residual stress as well as significant but incomplete recrystallization and homogenization. At 20 °C, low load ratio FCG was accelerated in the S+DA condition due to residual stresses and lack of significant crack path roughness induced closure. At high load ratios, the intrinsic FCG resistance of the S+DA material was slightly reduced compared to the HIP+S+DA condition which was comparable to high quality wrought material. At 650 °C, the 30 Hz fatigue crack growth performance followed similar trends to the 20 °C performance. Under constant and 0.1 Hz loading, all materials exhibited intergranular oxidation crack growth. However, compared to wrought and HIP LPBF materials, the highly elongated S+DA LPBF grain structure nearly halted crack growth perpendicular to the build direction and gave much higher fatigue resistance for that orientation with cracks deflecting off of the mode I loading direction. EBM fabricated TiAl (Ti-47Al-2Cr-2Nb) has a fine, distorted, and non-equilibrium single phase structure in the as-built condition. Three novel HIP cycles were developed targeting microstructures that are expensive or impossible to achieve through conventional manufacturing. The first treatment achieved a fine, homogeneous, equiaxed dual phase structure with high yield stress and low failure strain. The other HIP cycles achieved a duplex structure with regions of lamellar colonies and equiaxed structures, lower yield stress, and higher failure strains. Different cooling rates were employed to control the lamellar spacing and were shown to increase the yield stress.
-
(2022)ThesisRecently, blockchain becomes a disruptive technology of building distributed applications (DApps). Many researchers and institutions have devoted their resources to the development of more effective blockchain technologies and innovative applications. However, with the limitation of computing power and financial resources, it is hard for researchers to deploy and test their blockchain innovations in a large-scape physical network. Hence, in this dissertation, we proposed a peer-to-peer (P2P) networking simulation framework, which allows to deploy and test (simulate) a large-scale blockchain system with thousands of nodes in one single computer. We systematically reviewed existing research and techniques of blockchain simulator and evaluated their advantages and disadvantages. To achieve generality and flexibility, our simulation framework lays the foundation for simulating blockchain network with different scales and protocols. We verified our simulation framework by deploying the most famous three blockchain systems (Bitcoin, Ethereum and IOTA) in our simulation framework. We demonstrated the effectiveness of our simulation framework with the following three case studies: (a) Improve the performance of blockchain by changing key parameters or deploying new directed acyclic graph (DAG) structure protocol; (b) Test and analyze the attack response of Tangle-based blockchain (IOTA) (c) Establish and deploy a new smart grid bidding system for demand side in our simulation framework. This dissertation also points out a series of open issues for future research.
-
(2022)ThesisApple trees commonly require the removal of excessive flowers by thinning to produce marketable fruit. Estimating the flower counts and phenology is important for chemical thinning decisions. Farmers generally inspect flowers in randomly sampled trees within the orchard, which is time-consuming, labour-intensive, and unreliable. Existing algorithms for estimating flower density have proven to be of low efficacy. No published algorithms for estimating flower phenology distributions exist. This thesis first presents a novel pixel-level flower segmentation algorithm named FCNs-Edge to estimate flower density in apple orchards, which showed an improvement over State-Of-The-Art (SOTA) methods. A side-view apple flower density map is then generated for a variable rate chemical sprayer. The FCNs-Edge is then extended to multiple classes - green, pink and white flowers, demonstrating the feasibility to estimate three flower stage distributions grouped by colour. A novel method named DeepPhenology is proposed to estimate phenology distributions over eight apple flower stages by building direct relationships with real counts in field. The proposed method removes the need to label images, which overcomes difficulties in distinguishing overlapping or hidden flower clusters on 2D imagery. The proposed model was shown to outperform a SOTA object detection model for this task. This thesis finally delivers a novel data-centric analysis of on-tree fruit detection based on a SOTA object detection model using public datasets; of interest given the wide usage of deep learning in agriculture which lacks in-depth data-centric analysis. The results indicated that 2500 annotated objects are generally sufficient for single-class fruit training. A novel similarity score is also proposed to easily predict the AP score without any training. The proposed FCNs-Edge and DeepPhenology have demonstrated accurate, robust and efficient estimation of apple flower density and phenology distributions. Such methods can replace traditional human inspection, significantly reduce labour requirements and thus reduce costs. Such estimation of flower information can also provide tree-level decision support to manage the variable growth between trees. Ultimately, the data-centric analysis of on-tree fruit detection will help practitioners better understand the influence on the training accuracy from data-centric attributes, and thus prepare better quality datasets and achieve higher training accuracy.
-
(2022)ThesisElectron and nuclear spins are highly controllable and coherent quantum objects. They are therefore an excellent platform to study both fundamental physics and quantum information. Semiconductor quantum devices leverage the vast infrastructure that currently exists to produce our everyday electronics. With spins integrated into semiconductor devices, coherent control of individual electrons and nuclei has been demonstrated. Further development of these devices is essential to propel novel quantum technologies, such as quantum computers, beyond the lab. This thesis focuses on three themes: spin physics, quantum information processing and the foundations of quantum theory. We explore these topics with donors in silicon, phosphorus (31 P) and antimony (123 Sb). With a ‘Maxwell’s demon’ observing a single electron spin, its knowledge of the spin state heralds high-fidelity electron spin initialisation without requiring additional quantum resources. We benchmark the electron initialisation with high-fidelity nuclear spin readout by first mapping the electron state to the nucleus. We then motivate further improvements to the measurement apparatus to further enhance electron spin initialisation and readout. Recent advances have demonstrated embryonic two-qubit gates between donor-bound electrons. Alongside the high-fidelity readout afforded by nuclear spins, we are rapidly approaching the fault-tolerant threshold for some error-correcting codes, e.g. the surface code. We also discuss an electrical technique to coherently control quadrupolar nuclei, which we discovered for the first time in silicon with 123 Sb. With electrical control of donor electrons, this discovery could pave the way to an all-electrical donor spin quantum computer. Finally, with the highly-coherent ionised nuclear spin 123 Sb, we explore a foundational question in quantum mechanics. Originating from the Einstein-Podolsky-Rosen paradox, we discuss the reality of the quantum state. The quantum state is an incredibly accurate predictive tool, however the predictions are inherently probabilistic. The key question we seek to answer: is the probabilistic nature of the quantum state a representation of our (lack of) knowledge of the true state of reality? We propose an experimental test with 123 Sb that constrains the degree to which the quantum state can only represent knowledge.
-
(2022)ThesisBuilding fire accidents, as a continuing menace to the society, not only incur enormous property damage but also pose significant threats to human lives. More recently, driven by the rapid population growth, an increasing number of large-capacity buildings are being built to meet the growing residence demands in many major cities globally, such as Sydney, Hong Kong, London, etc. These modern buildings usually have complex architectural layouts, high-density occupancy settings, which are often filled with a variety of flammable materials and items (i.e., electrical devices, flammable cladding panels etc.). For such reasons, in case of fire accidents, occupants of these buildings are likely to suffer from an extended evacuation time. Moreover, in some extreme cases, occupants may have to escape through a smoke-filled environment. Thus, having well-planned evacuation strategies and fire safety systems in place is critical for upholding life safety. Over the last few decades, due to the rapid development in computing power and modelling techniques, various numerical simulation models have been developed and applied to investigate the building evacuation dynamics under fire emergencies. Most of these numerical models can provide a series of estimations regarding building evacuation performance, such as predicting building evacuation time, visualising evacuation dynamics, identifying high-density areas within the building etc. Nevertheless, the behavioural variations of evacuees are usually overlooked in a significant proportion of such simulations. Noticeably, evacuees frequently adjust their egress behaviours based on their internal psychological state (i.e., the variation of stress) and external stimulus from their surrounding environments (i.e., dynamical fire effluents, such as high-temperature smoke). Evidence suggests that evacuees are likely to shift from a low-stress state to a high-stress state and increase their moving speed when escaping from a high-temperature and smoke-filled environment. Besides, competitive behaviours can even be triggered under certain extremely stressful conditions, which can cause clogging at exits or even stampede accidents. Without considering such behavioural aspects of evacuees, the predicted evacuation performance might be misinterpreted based on unreliable results; thereby, misleading building fire safety designs and emergency precautions. Therefore, to achieve a more realistic simulation of building fire evacuation processes, this research aims to advance in modelling of human dynamical behaviour responses of each evacuee and integrating it into building fire evacuation analysis. A dynamical egress behaviour-based evacuation model that considering the evacuee’s competitive/cooperative egress movements and their psychological stress variation is developed. Furthermore, a fire hazard-integrated evacuation simulation framework is established by coupling with the fire dynamics simulator (i.e., FDS). By means of tracking dynamical interactions between evacuees and the evolutionary fire dynamics within the building space, evacuees’ local fire risks and stress levels under the impacts of locally encountered fire hazards (i.e., radiation, temperature, toxic gas, and visibility) can be effectively quantified. In this study, the developed simulation tool can provide a further in-depth building fire safety assessment. Thus, it contributes to performance-based fire safety engineering in designs and real applications, including reducing budgets and risks of participating in evacuation drills, supporting emergency evacuation strategy planning, mitigating fire risks by identifying risk-prone areas associated with building fire circumstances (e.g., putting preventative measures in place beforehand to intervene or mitigate safety risks, such as mass panic, stampede, stress evoked behaviours).