Filters
Availability
Publication Year
Resource Type
Faculty
School/Institute
Publication Search Results
Sort
Results per page
1 - 10 of 117
-
(2021)ThesisPer- and polyfluoroalkyl substances (PFAS), also known as fluorinated surfactants, are a class of emerging contaminants. Historical use of PFAS for commercial applications has caused widespread contamination in surface water, groundwater and soil/sediments worldwide, with broad environmental and human health implications. As such, understanding the mechanisms that control the fate and transport of PFAS compounds in a range of environmental conditions is of particular interest. Air-water interfacial adsorption is an important environmental process that contributes to PFAS fate and transport. It is well known that air-water interfacial behaviour of PFAS, and generally of surfactants, is strongly impacted by the molecular chemical structures (i.e., hydrophobicity of the carbon chains and hydrophilicity of functional groups) and environmental conditions (e.g., salinity). Two significant challenges related to the air-water interfacial adsorption of PFAS are (1) the large number of single PFAS compounds with diverse molecular structures in the environment, and (2) the influence of dynamic environmental conditions on interfacial behaviour. Limited research is currently available to adequately predict the air-water interfacial activity of PFAS compounds in natural conditions with varied salinity. Therefore, the aim of this thesis is to develop quantitatively predictive models to predict the interfacial behaviour for a wide range of environmentally relevant PFAS with differing composition and concentration of inorganic salts. To achieve this aim, the first part of the thesis presents a group contribution model to quantitatively predict the interfacial affinity for PFAS based on PFAS chemical structure. Literature values for air-water surface tensions were collected for a range of PFAS and conventional hydrocarbon surfactants, and then fitted to the Langmuir-Szyszkowski equation to quantify the interfacial affinity for single surfactants. These data were subsequently used as input to the group contribution model in order to determine the specific molecular component parameters. Using these parameters, the interfacial affinity is then calculated for any PFAS with known molecular components. In the next part of this thesis, a new model (named UNSW-OU) was developed based on the mass action law. This model predicts the air-water interfacial affinity for different salt concentrations from 0 to 0.5 M. This model was then expanded to predict the impact of salt composition, with measured surface tension data for PFAS solutions containing diverse salt compositions and concentrations collected via laboratory experiments. These data were then used as model inputs to calculate parameters that are specific to surfactants and different salt types. With these parameters, interfacial affinity is calculated for different anionic PFAS solutions containing monovalent and divalent salt components with an ionic strength up to 0.5 M. This thesis provides a quantitative approach to predict interfacial behaviour for a wide range of environmentally relevant PFAS under different inorganic salt concentrations. As salts are ubiquitous, and vary from site to site, a small change in salt concentration or composition is shown to have a substantial impact on PFAS interfacial behaviour. Therefore, the ability to calculate interfacial affinity in different salt conditions is important to achieve accurate predictions for PFAS transport in the vadose zone. Further, the knowledge obtained from this thesis is beneficial where the air-water interfacial area is significant, including in the long-range transport of PFAS due to interfacial adsorption via the sea-spray surface, and in PFAS treatment using gas bubbling and foam forming techniques.
-
(2021)ThesisGraphs are widely used to represent interactions (i.e., edges) between entities (i.e., nodes/vertices) in a large spectrum of applications, including social networks, biological protein-protein networks, and e-commerce networks. One fundamental task in graph analysis is to explore node-to-node relationships such as "how similar two proteins are in biological networks'' and "whether or not a user can influence another user in social networks''. In this thesis, we study the following three problems, which are of great importance in exploring these relationships. Firstly, we study the problem of role similarity computation. As one of the structural node similarity metrics, role similarity has the merit of indicating automorphism. However, existing algorithms cannot handle large-scale graphs. In this thesis, we propose an efficient algorithm StructSim, which admits a pre-computed index to query a node pair in O(k log D) time, where k is a small user-defined parameter, and D is the maximum node degree. To build the index efficiently, we further devise an FM-sketch-based technique that can handle billion-scale graphs. Secondly, we study how to quantify approximate simulation. Simulation and its variants are useful binary relations among nodes. However, all simulation variants are coarse "yes-or-no'' indicators that simply confirm or refute whether one node simulates another, which limits the scope of their utility. Therefore, it is meaningful to develop a fractional simulation measure to quantify the extent that a node simulates another. To this end, we propose a general fractional simulation computation framework that can be configured to quantify the extent of different simulation variants. Thirdly, we study the reachability problem on temporal bipartite graphs. Reachability, which studies if a node can reach the other node, has been extensively studied on (temporal) unipartite graphs, while it remains largely unexplored on temporal bipartite graphs. In this thesis, we study the temporal bipartite reachability problem. Specifically, a vertex u reaches a vertex w in a temporal bipartite graph if they are connected through a series of consecutive wedges with time constraints. We propose an index-based method to support fast reachability queries, and we also devise effective techniques to accelerate the index construction process.
-
(2021)ThesisA gravity dam is designed to retain water by using its self-weight to resist hydro-pressure of the reservoir. When an earthquake occurs, the energy emanated from the earthquake source may reach the dam site and cause the dam to vibrate. The earthquake action at the site is often the most critical loading case in the design of gravity dams. The estimation of the dynamic responses of gravity dams to earthquakes is necessary for achieving optimal upgrades and maintenance, and for improving our confidence in knowing that a dam will survive the impact of an earthquake of a specified magnitude. This thesis develops an efficient approach to the seismic analysis of gravity dam-reservoir-foundation systems with an emphasis on the seismic input modelling and adaptive damage simulation of dams. The whole system is divided into a bounded domain including the dam body and adjacent parts of reservoir and foundation, and an unbounded domain of reservoir and an unbounded domain of foundation. The dynamic properties of the unbounded domains are simulated by artificial boundaries formulated in the framework of Scaled Boundary Finite Element Method (SBFEM). The seismic waves are considered as plane waves in both two-dimensional and three-dimensional media. The seismic waves are inputted to the bounded domains by means of the Domain Reduction Method (DRM) through a single layer of elements adjacent to the interface between the bounded domain and the unbounded domain of foundation. The fully automatic quadtree/octree mesh technique is employed to discretize the complex geometry of the bounded domain including the dam and geological features in the foundation. The scaled boundary finite element method is applied in the bounded domain and overcomes the issue of hanging node faced by standard finite elements. The continuum damage mechanics is applied to model concrete and rocks as quasi-brittle materials. An h-adaptive strategy is developed for damage analysis to improve the computational efficiency. A progressive damage process is simulated through a series of optimal meshes. The proposed strategy simplifies the implementation of the adaptive analysis in automatic mesh refinement and data transfer. As the final outcome of this thesis, an automatic and efficient SBFEM formulation for seismic analysis of gravity dam-reservoir-foundation interaction systems has been developed. Case studies of gravity dams are performed.
-
(2021)ThesisAs the demands of autonomous mobile robots are increasing in recent years, the requirement of the path planning/navigation algorithm should not be content with the ability to reach the target without any collisions, but also should try to achieve possible optimal or suboptimal path from the initial position to the target according to the robot's constrains in practice. This thesis investigates path planning and control strategies for mobile robots with machine learning techniques, including ground mobile robots and flying UAVs. In this thesis, the hybrid reactive collision-free navigation problem under an unknown static environment is investigated firstly. By combining both the reactive navigation and Q-learning method, we intend to keep the good characteristics of reactive navigation algorithm and Q-learning and overcome the shortcomings of only relying on one of them. The proposed method is then extended into 3D environments. The performance of the mentioned strategies are verified by extensive computer simulations, and good results are obtained. Furthermore, the more challenging dynamic environment situation is taken into our consideration. We tackled this problem by developing a new path planning method that utilizes the integrated environment representation and reinforcement learning. Our novel approach enables to find the optimal path to the target efficiently and avoid collisions in a cluttered environment with steady and moving obstacles. The performance of these methods is compared with other different aspects. In addition, another important navigation problem, reconnaissance and surveillance problem for UAVs, is studied and two algorithms are presented. It requires drones to fully cover the area of interest along their trajectories. In the first method, a two-phase strategy is presented and enables to operate with a given altitude. Furthermore, an occlusion-aware UAV reconnaissance and surveillance approach is developed, which takes both UAV kinematics constraints and camera sensing limitations into consideration. We have implemented all the proposed algorithms by illustrative computer simulations in different scenarios, and the results have confirmed the effectiveness of these approaches. An extra study on the steering angle prediction algorithm for autonomous vehicles is presented. The proposed algorithm employs the convolutional neural network to extract features from the human driver and predicts the steering angle for autonomous driving. The performance of the algorithm is validated through simulations in different scenarios, we find the learned features can be transferred to the environment that has never been seen before.
-
(2021)ThesisThis century, power industry has been undergoing the revolution transitioning from coal-fired synchronous generators to renewable based generation which typically converts energy by voltage-source converters (VSCs). To address the emerging challenges due to significant integration of VSCs, the primary aim of this research is to model and design advanced control strategy for VSCs and understand and investigate the underlying principles of the interactions between the grid and power electronic inverters. The first part of this research elaborates on the design of control strategies for inverter-based microgrid and multi-microgrid cluster (MMC). As a basis of the control structure in this research, Chapter 2 proposes an autonomous and robust inverter control strategy to facilitate microgrid stable and secure operation, with the emphasis on alleviating fluctuations during microgrid state transitions. Furthermore, Chapter 3 presents a phase reformulation-based sliding framework to decouple inherent control interactions for inverter-based microgrid, and an improved frequency and phase synchronisation scheme to attenuate the interaction between the inverter and the weak grid. In Chapter 4, distributed consensus controllers with event-triggered mechanism are considered for MMC, which achieves reasonable power sharing and remarkably relieves communication and computation burdens. The second part of this research mainly analyses static voltage stability and small-signal stability of the inverter penetrated future power system. From the perspective of static voltage stability in Chapter 5, a systematic index is deduced to assess system loading status, with the impacts of inverter-based distributed generators (DGs) under various control modes rigorously discussed. From the perspective of small-signal stability in Chapter 6, the explicit state-space model for a multi-inverter system including different types of inverter-based DGs is developed by two-level component connection method (CCM), which modularised inverter control blocks at the primary level and inverter-based DGs at the secondary level. Through the analysis on eigenvalues of the partitioned subsystems, adverse impacts of generic inverter control prototypes on the power system stability are revealed. Case studies regarding the proposed methodologies have been undertaken on the benchmarking systems and compared to the existing literature where applicable to demonstrate the effectiveness and superiority.
-
(2021)ThesisArtificial micro/nanomotors (MNMs), inspired by mobile biomolecular entities, have demonstrated great potential as miniaturized robots performing diverse tasks from environmental remediation to biological treatment owing to their great mobility and versatility. The reported MNMs can be propelled using various power sources, including magnetic field, electric field, ultrasound, light, and chemical reaction. MNMs that operate on chemical reactions are usually equipped with higher velocity due to the superior energy conversion efficiency, which dominantly present as bubble propelled systems. However, the majority of the bubble propelled MNMs utilize bubble ejection and detachment force, which result in swarming and linear motion for Janus and tubular motors, respectively. It is still challenging for chemical propelled MNMs to have absolute control in direction without external field. A crafty design to circumvent this limitation is to develop biocatalytic MNMs with bubble buoyancy propulsion. This thesis focuses on the design, fabrication, and applications of submarine-like buoyancy-propelled MNMs that move in the vertical direction. I fabricated buoyancy propelled nanomotors with one-pot synthesis and provided the first work characterizing detailed motion behavior with electrochemistry. With coupled biocatalytic cascade reaction converting glucose as the fuel to oxygen bubbles, the nanomotor was propelled by buoyancy, which dominated the initiative collision at the electrode surface. Four representative electrical impact signals were observed and corresponded to four types of motion patterns. The corresponding relationship was confirmed with a numerical simulation. The integration of MNMs and electrochemistry provided a new dimension to characterize and understand the complex dynamics of the self-propelled nanoparticles. I further investigated the buoyancy propelled MNMs in biomedical applications. The pH-sensitive polymer incorporated micromotor exhibited regulated vertical motion via hydrophilic/hydrophobic phase shifting in different pH environments, and the system was proved to be applicable for anti-cancer drug delivery in a proof-of-concept three-dimensional cell culture. The proposed micromotor opens up new avenues in autonomous robotic fabrication for in vivo drug delivery in complex media. I also investigated the buoyancy propelled MNMs in water remediation. The buoyancy propelled nanomotor exhibited reversible vertical motion in low concentrations of H2O2, which induced the convection of the micro-environment and increased the pollutants to get in contact with the absorbents. The proposed nanomotor showed efficient removal of both inorganic heavy metal ions and organic per- and poly-fluoroalkyl substances (PFAS) in complex environments. At last, the buoyancy propelled MNMs were studied for the vertically spatial separation of targeted cancer cells in mixed samples. With the aid of antibody surface modification, the buoyancy-propelled nanomotors can autonomously attach to the targeted cells and endow the cancer cells with vertical motion. With a customized glass tube, the floated cells can be easily separated. The proposed nanomotor exhibited great isolating efficiency with facile operations, which broadened the development of cell separation methods towards biocompatible nanostructures. The findings presented in this thesis open up new avenues for the development of buoyancy propelled MNMs in diverse applications.
-
(2021)ThesisLearning node embedding for graphs has been proved essential for a wide range of applications, from recommendation to community search. However, most existing approaches mainly focus on either the graph structure information or the attribute (feature) information, and thus cannot make full use of the information of the graph data. In this thesis, we study four typical problems on node embedding in graphs, and utilize both structural and attribute information of the original graph to learn representative node embeddings to solve various tasks. Firstly, we investigate the problem of inductive link prediction on attributed graphs. Many real-world applications require inductive prediction for new nodes having only attribute information. It is more challenging since the new nodes do not have structure information and cannot be seen during the model training. To solve this problem, we propose a model called DEAL, which is versatile in the sense that it works for both inductive and transductive link prediction. Extensive experiments on several benchmark datasets show that our proposed model significantly outperforms existing inductive link prediction methods, and also outperforms the state-of-the-art methods on transductive link prediction. Secondly, we focus on the data imbalance problem of link prediction. Positive Unlabeled (PU) learning is a good choice to address this issue since only positive links are available. Unfortunately, the unknown class prior and data imbalance of graphs impede the use of PU learning. To deal with these issues, we propose a novel model-agnostic PU learning algorithm for Graph Neural Network (GNN)-based link prediction via PU-AUC optimization. The proposed method is free of class prior estimation and able to handle the data imbalance. Moreover, we propose an accelerated method to reduce the operational complexity of PU-AUC optimization from quadratic to linear. Extensive experiments back up our theoretical analysis and validate that the proposed method is capable of boosting the performance of the state-of-the-art GNN-based link prediction models. Thirdly, we focus on generating representative node embedding on dynamic graphs. Most of the existing graph embedding methods assume that the graph is static, where nodes and edges are fixed. However, in practice, real-world graphs (e.g., social networks, citation networks, and transportation networks) are often dynamic, meaning that both the topology and node attributes of the graph evolve over time. As a result, these methods ignore the evolution history and cannot perform well in dynamic graphs. Last but not least, we study the graph keyword search problem which aims to find subtrees or subgraphs containing all query keywords ranked according to some criteria. Existing studies all assume that the graphs have complete information. However, real-world graphs may usually contain some missing information (such as edges or keywords), thus making the problem much more challenging. To solve this, we propose a novel model named KS-GNN based on the graph neural network and the auto-encoder. By considering the latent relationships and the frequency of different keywords, KS-GNN alleviates the effect of missing information and is able to learn low-dimensional representative node embeddings that preserve both graph structure and keyword features. The experiments on four real-world datasets show that our model consistently achieves better performance than state-of-the-art baseline methods in graphs having missing information.
-
(2021)ThesisHeterogeneous information networks (HINs) are networks involving multiple typed objects and multiple typed links denoting different relations, and they are prevalent in various domains, such as bibliographic networks, social media networks, and knowledge networks. Recently, the topic of cohesive subgraphs discovery has gained plenty of attention. Typical cohesive subgraph models include k-core, k-truss, k-ECC, maximal-clique, quasi-clique, etc. Among the cohesive subgraph models, the k-core and the k-truss have been demonstrated to be outstanding, as they can achieve both high cohesiveness and high computational efficiency. Conceptually, the k-core of a graph is a subgraph in which each vertex participates in at least k-1 edges, while the k-truss of a graph is a subgraph in which each edge participates in at least k-2 triangles. Nevertheless, existing solutions mainly focus on homogeneous networks, where vertices are of the same type, and thus cannot be applied to HINs. In this thesis, we study the problem of cohesive subgraphs mining and analysis over large HINs. Firstly, we study the problem of k-core based community search over large HINs; that is, given a query vertex q, find a community from an HIN containing q, in which all the vertices are with the same type of q and have close relationships. To model the relationship between two vertices of the same type, we adopt the well-known concept of meta-path, which is a sequence of relations defined between different types of vertices. We then measure the cohesiveness of the community by extending the classic minimum degree metric with a meta-path. We further propose efficient query algorithms for finding communities using these cohesiveness metrics. Secondly, we study the problem of truss based cohesive subgraph computation over large HINs. As the classic truss model is based on the structure of triangle which does not exist in some HINs, we introduce two kinds of HIN tirangles for three vertices, regarding a specific meta-path P. The first one requires that each pair of vertices is connected by an instance of P, while the second one also has such a connectivity constraint but further needs that the three instances of P form a circle. Based on these two kinds of triangles, we propose two HIN truss models respectively. We further develop efficient truss computation algorithms. Finally, we study the problem of bi-triangle counting over a specific type of networks--a bipartite network, where a bi-triangle is a cycle with three vertices from one vertex set and three vertices from another vertex set. Counting bi-triangles has found many real applications such as computing the transitivity coefficient and clustering coefficient for bipartite networks. To enable efficient bi-triangle counting, we first develop a baseline algorithm relying on the observation that each bi-triangle can be considered as the join of three wedges. Then, we propose a more sophisticated algorithm which regards a bi-triangle as the join of two super-wedges, where a wedge is a path with two edges while a super-wedge is a path with three edges. We further optimize the algorithm by ranking vertices according to their degrees.
-
(2021)ThesisDeep learning has achieved great success in many real-world applications, e.g., computer vision and human healthcare. Current deep learning usually relies on large-scale datasets with large amounts of carefully annotated data, however, objects in the real world follow a long-tailed distribution, i.e., a tremendous number of classes have little data and hard to be collected. Besides, new classes keep emerging that collecting, and particularly annotating examples is impossible. The data insufficiency poses a bottleneck in the robustness of deep learning methods. Targeting this challenge, I propose to ease data insufficiency in two general data situations: single dataset and multiple datasets. For a single dataset, I propose to discover extra knowledge to enrich the learnable information for models (i.e., knowledge discovery). For multiple datasets, I propose to transfer generalizable knowledge to enhance the analysis of limited datasets (i.e., knowledge transfer). To discover extra new knowledge in a single dataset, I propose to implement the discovery process in three levels. 1) Dataset level. Using contrastive relationships as new information to improve data efficiency on small-scale datasets. 2) Subject/class level. Enabling models to be aware about class/subject and thus be robust handling class/subject variance. 3) Instance level. Automating expert analysis as a new supervisor to boost model training. To enable models to transfer knowledge between datasets, I propose to tackle three common transfer problems in the real world. 1) Lacking data in one of the datasets. I transfer cross-domain knowledge from sufficient to insufficient information domains to augment new data and ease data inefficiency. 2) Multi-view aggregation problem. I propose a general multi-view algorithm to transfer and unite different information from hierarchical views. 3) Biased knowledge transfer among datasets. I propose a task-aligned meta-learning model to learn generalizable knowledge for transferring. In this thesis, I have conducted extensive experiments and ablation studies to validate the proposed methods on diverse real-world applications, e.g., limited-scale datasets of human activity recognition, image classification, neural signal analysis, and commercial analysis.
-
(2021)ThesisKesterite Cu2ZnSn(S,Se)4 (CZTSSe) materials have attracted considerable interest as a promising candidate for future thin film photovoltaic technology. This material family enjoys its compelling features of inexpensive and non/low toxic constituents, thermodynamically stable structure, and theoretical high power conversion efficiency defined by its suitable optoelectronic properties. However, on the other hand, a flexible quaternary kesterite structure induces much more complex defect chemistry in the crystal lattice, which leads to the distinct band- or potential-fluctuation, short minority carrier lifetime, and associated severe bulk as well as interface recombination. This thesis aims to tackle these problems by the addition of extrinsic elements (Ag, Cd, Ge, and Li) to the kesterite matrix (so-called doping/alloying strategy) for improving the electronic properties of CZTSSe solar cells. First, the substitution of Cu by Ag for CZTSSe absorbers has been achieved by annealing Ag contained precursor. The formation mechanism of Ag alloyed CZTSSe has been studied in detail. Second, a small amount of Cd on CZTSSe has been successfully incorporated by depositing a CdS layer on precursor film prior to selenization. The incorporation of Cd dramatically passivated GBs due to Cd segregation. As a result, the electronic properties of CZTSSe have been greatly improved. Third, The incorporation of Ge has effectively improved the VOC and FF of the CZTSSe solar cell. This enhancement is found to be mainly associated with modified defect characteristics instead of improved Na content in the CZTSSe bulk region. Then, we studied the effect of double cations (Cd and Ge) incorporation in electronic properties CZTSSe solar cell. This strategy simultaneously reduced nonradiative recombination at different regions of CZTSSe solar cells. Finally, a feasible solution-based post-deposition treatment process is developed for incorporating lithium into the CZTSSe absorber. The dominant acceptor defects CuZn antisites are replaced by shallower LiZn antisites, thus leading to enhanced effective p-type doping and reduced recombination, especially in the space charge region. Based on the results and analysis, this thesis elucidates new insights into the underlying mechanism of different doping/alloying strategies in kesterite-based materials