Filters
Availability
Publication Year
Resource Type
Faculty
School/Institute
Publication Search Results
Sort
Results per page
1 - 10 of 238
-
(2021)ThesisNatural kidney filtration is a compact, multi-step filtration process which passes wastes and exceeded fluids via microscale vessels in glomerulus and tubules. The principal renal replacement therapy (RRT), commonly called dialysis, is a single-step filtration process based on diffusion to replace kidney failure. Conventional dialysis is limited in its effectiveness (not a continuous treatment), its impact on quality of life (typically requiring patients to spend several days per week in a clinic), and its cost (large systems, requiring frequent membrane replacement). This thesis is an investigation into the feasibility of using microfluidics and membrane technology to create portable alternatives to dialysis systems. It starts with a comprehensive review of the state-of-the-art in portable artificial kidneys, microfluidics, membrane science, and other related fields. An innovative, multi-step process was designed to mimic kidney filtration using two membranes; one to filter out large particles and one to remove urea and recycle water, thus mitigating the need for a dialysate system. The underlying physics (the mixing and shear stress) of the mechanisms which could enhance filtration performance at microscale was then studied. It was found that by adding microspacers into narrow-channel flows, it is possible to significantly enhance filtration. Optimized 3D-printed spacer designs (e.g., a ‘gyroid’ spacer) showed flux enhancement of up to 93% (compared to a plain channel) when using a plasma mimicking solution. The use of different blood and plasma mimicking solutions also suggested a prior step to separate large biological components (e.g., cells, proteins) is helpful to reduce cell contact and fouling in membrane filtration. The potential use of microfluidic diode valves and micropumps for pressure and flowrate regulation in the proposed small-format system was discussed. Membrane processes which mimic the filtration function of the tubules and have the potential for integration into portable systems (e.g., reverse osmosis and membrane distillation) are demonstrated to be useful potential alternatives to dialysis in toxin removal and in returning clean water to the blood stream.
-
(2021)ThesisA pseudolite (PL) is a ground-based positioning system that offers flexible deployment and accurate “orbits”. The PL system can carry on the role of the GNSS to provide precise positioning for indoor users. However, there are some unusual challenges that seriously affect the performance of a PL system in precise indoor positioning. To raise PL-based positioning accuracy up to the centimeter level or higher, the use of the PL carrier phase measurement with ambiguity resolution is a unique consideration. The PL phase ambiguities are also contaminated by clock bias, multipath errors, and cycle clips. Their existence destroys the integer nature of ambiguity and impedes the pursuit of further accuracy improvement. The major contributions in this research for addressing the above-mentioned challenging issues are specified as follows: 1. The ground-based AR methods are discussed. The impact of ground-based geometry on indoor AR is researched, and the influence of linearization error is also investigated. An efficient PL-based AR method is studied and verified in the balance of gaining convenience and avoiding linearization impact. 2. The clock bias between PL transmitters can be properly handled in a way that time synchronization can be achieved with a transmitter-only PL system at low cost and simplicity. Therefore, the PL-based the ambiguities are able to be fixed to correct integers, and centimeter-level indoor precise positioning can be reliably achieved. In addition, the proposed way for time synchronization is also applicable for other ground-based systems for precise positioning purposes. 3. The stochastic model for mitigation of indoor multipath and NLOS is investigated. The experimental results demonstrate that the proposed stochastic model is superior to other existing models in indoor multipath mitigation as it is competent to suppress the multipath errors mainly caused by multipath to the smallest in both static and kinematic results, respectively. Moreover, it is also verified to be efficient for NLOS mitigation. With the proposed new stochastic model, precise point positioning is confidently expected indoors. 4. The methods for PL-based cycle slips are extensively studied and discussed. Numerical results indicate that the integer-cycle slips can be efficiently and accurately detected and corrected. The concern about PL-based cycle slip is minimized, the reliability and sustainability of PL-based precise indoor positioning can be promised.
-
(2021)ThesisPolymer brushes are arrays of densely surface-tethered polymer chains, and are of interest for two reasons. Firstly, they possess interfacial characteristics, such as antifouling and lubrication, that are desirable in many applications. Secondly, they are model systems that can provide additional insight into polymer behaviour due to their unique geometry. Observing the interfacial structure of these brush layers is critically important for understanding both their properties and the mechanisms driving the polymer behaviour. To date, neutron reflectometry (NR) is the only technique that can demonstrably resolve the nanoscale structure of polymer brushes. However, these diffuse interfaces produce subtle features in the reflectometry data that challenge interpretation, with typical analyses failing to quantify the derived structure's uncertainty. Furthermore, the experimental potential of this technique for the study of brushes is only just being realised. This Thesis advances NR as a tool for studying polymer brush systems by establishing a robust analysis methodology that overcomes previous hurdles and demonstrating novel experimental techniques. In both cases, poly(N-isopropylacrylamide) (PNIPAM) brushes are used as model systems. First, the polymer system is characterised through the novel observation of surface-initiated ARGET ATRP using time-resolved NR, and a study of the dry brush as a function of humidity and temperature. Second, methodologies are developed that allow for robust determination of both solvated and confined brush structures. Lastly, NR is used to elucidate the behaviour of PNIPAM brushes in complex environments. A novel confinement apparatus is used to investigate the structure of a PNIPAM brush under mechanical confinement and contrast-variation provides unparalleled insight into PNIPAM–surfactant systems. In each case, complementary techniques are essential in guiding reflectometry experiments and fully understanding the polymer system. This work develops and demonstrates techniques that enhance the study of diffuse interfaces with the NR technique. Moreover, the holistic structural examination of PNIPAM undertaken sheds new light on the phase behaviour of this ostensibly well-understood polymer and highlights its rich interaction with surfactants.
-
(2021)ThesisThis thesis investigated the ecology and metabolism of microorganisms, especially yeasts, during the wet fermentation of Australian coffee beans, and their contribution to coffee quality. Pulped coffee beans were fermented underwater for 36 h where yeast growth was suppressed by the addition of Natamycin at 300 mg/L. Spontaneous fermentation without the addition of Natamycin was conducted as control. The growth and diversity of microorganisms during fermentation were monitored by both culture dependent and independent methods. Major non-volatile metabolites during fermentation were monitored by high performance liquid chromatography (HPLC) and volatiles in the green and roasted beans were measured by solid phase microextraction coupled with gas chromatography tandem mass spectrometry (SPME GC-MS). Both bacteria and yeasts grew significantly during spontaneous fermentation while yeast growth was restricted in the Natamycin treated fermentation without significant impact on bacterial growth. The bacterial community was dominated by Citrobacter sp., Gluconobacter cerinus, Leuconostoc mesenteroides and Lactococcus lactis with maximum populations between 4-7.2 log CFU/g, while Hanseniaspora uvarum and Pichia kudriavzevii were the predominant yeasts at 4.5-5 CFU/g. During fermentation, the microflora utilized sugars in the mucilage and produced mannitol, glycerol and essential volatiles, mainly alcohols, esters, aldehydes and organic acids, with their concentrations generally lower in beans fermented with yeast suppression. Coffee produced from yeast suppressed fermentation received lower sensory scores in flavour and aroma and overall quality by 3 Q-Grade coffee masters. When H. uvarum and P. kudriavzevii were inoculated individually and in combination, they dominated the fermentation by growing to 9-10 log CFU/ml, and produced greater amounts of glycerol and flavour volatiles in the green beans which remained in higher levels after roasting compared with the control. Coffee brewed from these beans received significantly high scores of flavour, aroma, acidity and overall quality. Mucilage degradation seems to be initiated by endogenous enzymes and microbial contributions to the process occurred subsequently either enzymatically or by acidification. These findings demonstrated the crucial contribution of yeasts to successful coffee fermentation and high-quality coffee, and the potential of developing the two yeasts into starter cultures for coffee fermentation.
-
(2021)ThesisRural electrification demands the use of inexpensive technologies such as single wire earth return (SWER) networks. There is a steadily growing energy demand from remote consumers, and the capacity of existing lines may become inadequate soon. Besides, the existing SWER networks are very inefficient and experience poor voltage regulation. Furthermore, high-impedance arcing faults (HIF) from SWER lines can ignite bushfires such as the catastrophic 2009 Black Saturday bushfires in Victoria (Australia). Replacing SWER lines by cables as recommended by the Victorian Royal Commission comes at an astronomical cost and service providers are not able to comply with. As a solution, reliable remote electricity networks can be established through splitting the existing system into microgrids, and existing SWER lines can be utilised to interconnect those microgrids. The development of such reliable networks with better energy demand management will rely on having an integrated network-wide condition monitoring system. As the first contribution of this thesis, a distributed online monitoring platform is developed that incorporates power quality monitoring, real-time HIF identification and transient classification in SWER network. Characteristic features are extracted from the current and voltage signals, and Artificial Intelligence (AI) based classification techniques are developed to classify the faults and transients. The proposed approach demonstrates higher HIF detection accuracy (98.67%) and reduced detection latency (115.2 ms). Secondly, to facilitate electricity demand management, a remote consumer load identification methodology is developed to detect the load type from its turn-on transients. An edge computing-based architecture is proposed to facilitate the high-frequency analysis for load identification with minimum data transmission. Computationally efficient load identification methodologies are developed to enable their real-time deployment on resource constrained devices. The proposed approach is evaluated in real-time, and it achieves an average accuracy of 98% in identifying different loads. Finally, a deep neural network-based energy disaggregation framework is developed to separate the load specific energy usage from an aggregated signal. A generative approach is applied to model energy usage patterns. The proposed framework is evaluated using a real-world data set. It improves the signal aggregate error by 44% and mean aggregate error by 19% in comparison with other state-of-the-art techniques.
-
(2021)ThesisIn recent years, electric power generation using renewable energy sources has experienced an exponential growth in the world energy market. Their unprecedented large scale penetration foresees a 100% renewable based power generation in near future. These sources require power electronic converters at various levels for power conversion and grid integration. The converters are fast acting devices and use advance control techniques in a hierarchical manner. With the retirement of the conventional synchronous generators, the modern power electronics based power system lacks the inertia property. Hence, they are more vulnerable to several grid transient events; for instance grid faults. Control functionalities classify these converters as grid-forming and grid-following. Both these types can act as grid supporting devices during grid fault. In contrast to grid-forming type, grid-following type converters are mostly used to support the grid. The act of supporting the grid with reactive power instead of tripping during a fault for a pre-defined duration is known as fault ride-through. These converters rely on a separate synchronization unit to inject grid current during both normal and fault condition. Recent grid fault events across the globe have revealed the inefficiency of such synchronization units. This is attributed to the delayed grid parameter estimation that eventually leads to the tripping of the converters rather ride-through. It indicates that the performance robustness of the synchronizing unit while considering the fault ride-through of converter needs to be further investigated thoroughly. In lieu of the above, accurate and fast grid voltage parameter estimation is essential for grid-connected converters. To achieve this objective, the contributions of this thesis are classified into two parts. The first part of the thesis deals with the fault detection for converters during a grid fault using digital signal processing (DSP) techniques. Faster fault detection is vital to safeguard the power converter as they have limited fault current carrying capacity. Hence a hybrid fault detection technique is proposed. The technique combines the features of two DSP techniques, Hilbert-Huang Transform (HHT) and Teager Energy Operator (TEO). This is called Teager-Huang in this thesis. With this proposed technique, several grid faults, balanced and unbalanced, in both the grid voltage magnitude and phase-angle jumps are detected. Further, comparisons of the fault energies are presented, which provides a benchmark for the severity of the grid faults. In the second part of the thesis, the synchronization aspect of the converter is investigated. For the purpose of analysis, the synchronization using the classical synchronous reference frame phase-locked loop (SRFPLL) is considered. Initially, the synchronization inefficiency of SRFPLL during the grid fault is explained in regards to the loss of synchronization (LOS) instability. It is shown that the cause for LOS during a fault may be initiated as results of very low grid voltage magnitude, high grid impedance or high current injection. The analysis emphasises on the occurrence of phase-angle jump (PAJ) during a fault. The thesis indicates that the conventional SRFPLL design parameters result in synchronization delay and insufficient damping to ride-through such PAJs. The decrease in the SRFPLL synchronization robustness highly affects the grid-connected converters. To enhance the grid synchronization performance during a grid fault with PAJ, a hybrid grid synchronization concept is proposed. It consists of both hybrid phase-angle estimators and hybrid frequency estimators. The hybrid frequency estimators contain several improved adaptive and PLL independent frequency estimation techniques. The proposed technique is designed to be compatible with both the three-phase and single-phase grid synchronization. To avoid voltage transients during the transition between the estimators, a transition scheme is presented. This is controlled based on the instantaneous phase-angle error measured by the estimators. The three-phase grid-connected converter is modelled using the proposed hybrid grid synchronization technique. The current controller of the converter is designed both in stationary and synchronous reference frame. Further, the fault ride-through (FRT) strategy is embedded in the converter controller. With the developed model, the FRT of the converter is tested during a fault. Both symmetrical and asymmetrical grid faults with PAJs are considered. The efficacy of the proposed technique is evaluated using both simulation and experimental validations. The last part of the thesis explores the FRT of single-phase power converters employing the proposed hybrid grid synchronization transition. The synchronization performance along with the current controller robustness during FRT is investigated.
-
(2021)ThesisAmong various physiological and behavioural traits, speech has gained popularity as an effective mode of biometric authentication. Even though they are gaining popularity, automatic speaker verification systems are vulnerable to malicious attacks, known as spoofing attacks. Among various types of spoofing attacks, replay attack poses the biggest threat due to its simplicity and effectiveness. This thesis investigates the importance of 1) improving front-end feature extraction via novel feature extraction techniques and 2) enhancing spectral components via adaptive front-end frameworks to improve replay attack detection. This thesis initially focuses on AM-FM modelling techniques and their use in replay attack detection. A novel method to extract the sub-band frequency modulation (FM) component using the spectral centroid of a signal is proposed, and its use as a potential acoustic feature is also discussed. Frequency Domain Linear Prediction (FDLP) is explored as a method to obtain the temporal envelope of a speech signal. The temporal envelope carries amplitude modulation (AM) information of speech resonances. Several features are extracted from the temporal envelope and the FDLP residual signal. These features are then evaluated for replay attack detection and shown to have significant capability in discriminating genuine and spoofed signals. Fusion of AM and FM-based features has shown that AM and FM carry complementary information that helps distinguish replayed signals from genuine ones. The importance of frequency band allocation when creating filter banks is studied as well to further advance the understanding of front-ends for replay attack detection. Mechanisms inspired by the human auditory system that makes the human ear an excellent spectrum analyser have been investigated and integrated into front-ends. Spatial differentiation, a mechanism that provides additional sharpening to auditory filters is one of them that is used in this work to improve the selectivity of the sub-band decomposition filters. Two features are extracted using the improved filter bank front-end: spectral envelope centroid magnitude (SECM) and spectral envelope centroid frequency (SECF). These are used to establish the positive effect of spatial differentiation on discriminating spoofed signals. Level-dependent filter tuning, which allows the ear to handle a large dynamic range, is integrated into the filter bank to further improve the front-end. This mechanism converts the filter bank into an adaptive one where the selectivity of the filters is varied based on the input signal energy. Experimental results show that this leads to improved spoofing detection performance. Finally, deep neural network (DNN) mechanisms are integrated into sub-band feature extraction to develop an adaptive front-end that adjusts its characteristics based on the sub-band signals. A DNN-based controller that takes sub-band FM components as input, is developed to adaptively control the selectivity and sensitivity of a parallel filter bank to enhance the artifacts that differentiate a replayed signal from a genuine signal. This work illustrates gradient-based optimization of a DNN-based controller using the feedback from a spoofing detection back-end classifier, thus training it to reduce spoofing detection error. The proposed framework has displayed a superior ability in identifying high-quality replayed signals compared to conventional non-adaptive frameworks. All techniques proposed in this thesis have been evaluated on well-established databases on replay attack detection and compared with state-of-the-art baseline systems.
-
(2021)ThesisUnmanned aerial vehicles (UAVs) have become very popular for many military and civilian applications including in agriculture, construction, mining, environmental monitoring, etc. A desirable feature for UAVs is the ability to navigate and perform tasks autonomously with least human interaction. This is a very challenging problem due to several factors such as the high complexity of UAV applications, operation in harsh environments, limited payload and onboard computing power and highly nonlinear dynamics. Therefore, more research is still needed towards developing advanced reliable control strategies for UAVs to enable safe navigation in unknown and dynamic environments. This problem is even more challenging for multi-UAV systems where it is more efficient to utilize information shared among the networked vehicles. Therefore, the work presented in this thesis contributes towards the state-of-the-art in UAV control for safe autonomous navigation and motion coordination of multi-UAV systems. The first part of this thesis deals with single-UAV systems. Initially, a hybrid navigation framework is developed for autonomous mobile robots using a general 2D nonholonomic unicycle model that can be applied to different types of UAVs, ground vehicles and underwater vehicles considering only lateral motion. Then, the more complex problem of three-dimensional (3D) collision-free navigation in unknown/dynamic environments is addressed. To that end, advanced 3D reactive control strategies are developed adopting the sense-and-avoid paradigm to produce quick reactions around obstacles. A special case of navigation in 3D unknown confined environments (i.e. tunnel-like) is also addressed. General 3D kinematic models are considered in the design which makes these methods applicable to different UAV types in addition to underwater vehicles. Moreover, different implementation methods for these strategies with quadrotor-type UAVs are also investigated considering UAV dynamics in the control design. Practical experiments and simulations were carried out to analyze the performance of the developed methods. The second part of this thesis addresses safe navigation for multi-UAV systems. Distributed motion coordination methods of multi-UAV systems for flocking and 3D area coverage are developed. These methods offer good computational cost for large-scale systems. Simulations were performed to verify the performance of these methods considering systems with different sizes.
-
(2021)ThesisCognitive services provide artificial intelligence (AI) technology for application developers, who are not required to be experts on machine learning. Cognitive services are presented as an integrated service platform where end users bring abilities such as seeing, hearing, speaking, searching, user profiling, etc. to their own applications under development via simple API calls. As one of the above abilities, recommender systems serve as an indispensable building brick, especially when it comes to the information retrieval functionality in the cognitive service platform. This thesis focuses on the novel recommendation algorithms that are able to improve on recommendation quality measured by accuracy metrics, e.g., precision and recall, with advanced deep learning techniques. Recent deep learning-based recommendation models have been proved to have state-ofthe-art recommendation quality in a host of recommendation scenarios, such as rating prediction tasks, top-N ranking tasks, sequential recommendation, etc. Many of them only leverage the existing information acquired from users’ past behaviours to model them and make one or a set of predictions on the users’ next choice. Such information is normally sparse so that an accurate user behaviour model is often difficult to obtain even with deep learning. To overcome this issue, we invent various adversarial techniques and apply them to deep learning recommendation models in different scenarios. Some of these techniques involve generative models to address data sparsity and some improve user behaviour modelling by introducing an adversarial opponent in model training. We empirically show the effectiveness of our novel techniques and the enhancement achieved over existing models via thorough experiments and ablation studies on widely adopted recommendation datasets. The contributions in this thesis are as follows: 1. Propose the adversarial collaborative auto-encoder model for top-N recommendation; 2. Propose a novel deep domain adaptation cross-domain recommendation model for rating prediction tasks via transfer learning; 3. Propose a novel adversarial noise layer for convolutional neural networks and a convolutional generative adversarial model for top-N recommendation.
-
(2021)ThesisThe development of science and technology brings novel directions, data sources and methods to deal with transport issues. In the past few decades, especially, a large amount of researches in the transport field have been undertaken based on applying emerging data sources or methods. In light of their findings, emerging data sources are considered to have relatively easier accessibility, larger user coverage and richer information. Meanwhile, novel modelling methodologies are better at mining the latent relationship between the various aspects of the data. Therefore, combining emerging data sources and novel large-scale data analysis methodologies is considered to be valuable for improving existing models and solving emerging problems in transport research. This thesis will mainly concentrate on three related problems. The first is to apply these models to a traditional problem – travel demand estimation. This study demonstrates that social media is an appropriate data source for improving the travel demand estimation model. It also presents the factors which may affect the usage of social media data in the existing models. The second problem addressed is an emerging problem – carsharing system design and demand forecasting. The aim is to estimate the latent demand by spatio-temporal autocorrelation model using historical big data. The last problem is potentially a future problem called a ‘Network Inferring Problem’ which is helpful in sketch network designing. This problem is defined as finding the number of links which can feasibly carry all the point-to-point demand. Meanwhile the travel time on those links is closest to what is actually observed. Following the ‘storyline’ of ‘traditional-emerging-future’, we aim to demonstrate that novel data sources and the analysis methods to enable them are appropriate for the different requirements of various transport research problems. Even though gaps could still be found in both quantity and quality between emerging data sources and traditional data sources, there is enough evidence to believe that with the improvement of information technology and analysis methods, they will have promising prospects for different studies in the future.