Sofia Chaudry, Murdoch University
Xueli Bai, Curtin University
As a generalization of the linear complementarity problem and a special case of the nonlinear complementarity problem, recently, the tensor complementarity problem has been investigated in the literature. From theoretical perspective, we study the non-emptiness and compactness of the solution set, the uniqueness of the solution as well as the stability and continuity analysis. And from algorithmically perspective, we focus on designing algorithms for the tensor complementarity problem which beyond algorithmic frameworks designed for general nonlinear complementarity problems. So far, by using properties of structured tensors, many good results for the tensor complementarity problem have been obtained.
Simon Bowly, The University of Melbourne
The results of empirical analysis of algorithm performance are highly dependent on the diversity of test instances used. Real-world test sets alone may not fully explore the range of easy and hard cases for a given algorithm. Furthermore, without sufficient diversity in instance properties, it can be difficult to determine the root cause of good or bad algorithm performance. This is particularly true for Mixed Integer Programming (MIP) solvers, where a wide range of strategies are employed even within a single run.
To augment existing test data sets, we use Genetic Algorithms (GAs) to generate test instances with challenging properties. GAs have been used in previous work to produce instances which challenge heuristics for combinatorial problems. This talk highlights two extensions developed to generate new test cases for MIP solvers. The first develops an encoding which restricts the search space to a subset of useful instances. The second chooses in-depth performance metrics relevant to particular components of the solver in order to generate instances from which useful insights can be gained. To demonstrate the method, new test instances are generated which challenge branch variable selection strategies used in the MIP solver SCIP.
Dr Philipp Braun, The University of Newcastle
The energy transition, from a centralized to a decentralized and sustainable power supply using small scale power plants, presents new challenges to the distribution grid provider who is responsible for maintaining the stability of the electricity network. New procedures to ensure the overall network performance need to be developed, which are flexible with respect to the underlying network and scalable, to be able to handle the amount of data of a fast growing network of renewable energy sources. In this talk we examine model predictive control (MPC) and hierarchical distributed optimization schemes to tackle these challenges. In particular, we use a network of residential energy systems (RESs), connected to a grid provider through a point of common coupling, where every resident is equipped with local generators and local storage devices to examine hierarchical distributed optimization algorithms with a focus on flexibility (plug and play capability), scalability and convergence. The performance of the distributed optimization schemes embedded in the MPC closed loop are analyzed and illustrated through numerical simulations.
Xiaoli Cao, Curtin University
The current highly changeable environment and market demands require much more change management in all kinds of organizations, even in universities and colleges. University changes and transformation also require their staff to support and involvement. Due to differences in individual characteristics, teachers’ attitudes and behaviors toward to change in the universities lead to different job satisfaction. The purpose of this study is to integrate and expand people-oriented organizational change management by examining the relationship among commitment to change, coping with change, and job satisfaction. On the basis of relevant literature review, the research hypothesis is put forward, and the research model is constructed. Data were collected from 597 teachers of 10 universities undergoing significant organizational change. Results from structural equation modelling indicate that (a) the relationship between affective commitment to change and job performance was fully mediated by coping with change, (b) the relationship between continuance commitment to change and job satisfaction was only partially mediated by coping with change, and (c) normative commitment to change had a direct impact on job satisfaction. Results are discussed in terms of implications for managing universities change.
Laura Cartwright, University of Wollongong
Co-author: Dr Andrew Zammit-Mangion, Dr Nicholas M. Deutscher, University of Wollongong
Dr Andrew Feitz, Geoscience Australia
The detection and quantification of greenhouse-gas fugitive emissions is of both national and global importance. Despite several decades of active research, it remains predominantly an open problem, largely due to model errors and misspecifications that appear at each stage of the flux-inversion processing chain. In 2015, a controlled-methane-release experiment headed by Geoscience Australia was carried out at the CO2CRC Ginninderra site, and a variety of instruments and methods were employed for quantifying the emission rate. In this talk I will present a fully Bayesian approach to atmospheric tomography for inferring the emission rate. The Bayesian framework is designed to account for uncertainty in the measurements, the meteorological data, the temporally varying background concentration, and the atmospheric model itself, when doing inversion using Markov chain Monte Carlo. We show the utility of our approach in detecting and quantifying methane emissions from both point and path instruments, and highlight its ability to reasonably quantify methane emissions in a timely manner, using data collected during the Ginninderra experiment. This work is joint with Andrew Zammit-Mangion, Andrew Feitz, and Nicholas M. Deutscher.
Peng Cheng, Curtin University
We propose a novel method for constructing probabilistic robust disturbance rejection control for uncertain systems in which a scenario optimization method is used to deal with the nonlinear and unbounded uncertainties. For anti-disturbance, a reduced order disturbance observer is considered and a state-feedback controller is designed. Sufficient conditions are presented to ensure that the resulting closed-loop system is stable and a prescribed H1 performance index is satisfied. A numerical example is presented to illustrate the effectiveness of the techniques proposed and analyzed.
Dr Minh N. Dao, The University of Newcastle
We propose a flexible approach for computing the resolvent of the sum of weakly monotone operators in real Hilbert spaces from individual resolvents. This relies on splitting methods where strong convergence is guaranteed. We also prove linear convergence under Lipschitz continuity assumption. The approach is then applied to computing the proximity operator of the sum of weakly convex functions, and particularly to finding the best approximation to the intersection of convex sets.
Ya-zheng Dang, Curtin University
The alternating direction method of multipliers (ADMM) is widely used to solve large-scale linearly constrained optimization problems, convex or nonconvex, in many engineering fields. However there is a general lack of theoretical understanding of the algorithm when the objective function is nonconvex. its theoretical convergence guarantee is still an open problem. In this paper we analyze the convergence of the ADMM for solving certain nonconvex consensus problems. We show that the classical ADMM converges to the set of stationary solutions, provided that the penalty parameter in the augmented Lagrangian is chosen to be sufficiently large and the objective function satisfies some conditions. In the problem, the objective is nonconvex and possibly nonsmooth. Our analysis does not impose any assumptions on the iterates generated by the algorithm as previous corresponding research, and flexible selection in the flexible block is employs in the process.
Dr Giovanni Firmani, Roy Hill Iron Ore
Water management is an integral part of operations at the Roy Hill mine. The water balance can be described in terms of tasks, including dewatering, water treatment, ore processing, tailing storage facility (TSF), dust suppression and surplus water disposal. The opportunity exists to optimise the cost and water footprint of the operation through integration of the water balance tasks.
The prediction of the water use is generally analysed by modelling the abovementioned tasks individually. However, the achievement of an optimal water balance plan is complex when, for example, the dewatering predictions, that are highly variable in terms of both volume and quality, may be at times in conflict with water demands for ore processing.
The purpose of this work was to optimisation the cost and water footprint of water management by analysing the results of each modelling task and evaluate the presence of critical scenarios for the future with an automated and mathematically rigorous tool. In this work, we describe the development of a software application that integrates the results of each modelling task, providing the optimal utilisation of water.
Chuanye Gu, Curtin University
This paper develops a distributionally robust joint chance constrained optimization model for motorway management. The optimization aims to minimize motorway delay via ramp metering with consideration of uncertainties in traffic demand and road capacity. The major contribution of this paper is to propose an approach to approximate a joint chance-constrained Cell Transmission Model based motorway optimization with only partial distributional information of uncertain demand and capacity. The resultant formulation is a semidefinite program. Some numerical experiments are conducted to demonstrate that the proposed approximation approach is efficient. The proposed approximation approach may provide useful insights and have broader applicability in traffic management and traffic planning problems under uncertainty.
Lingshuang Kong, Curtin University
The blending operation is a key process in alumina production. The real-time optimization (RTO) of ﬁnding an optimal raw material proportioning is crucially important for achieving the desired quality of the product. However, the presence of uncertainty is unavoidable in a real process, leading to much diﬃculty for making decision in real-time. This paper presents a novel robust real-time optimization (RRTO) method for alumina blending operation, where no prior knowledge of uncertainties is needed to be utilized. The robust solution obtained is applied to the real plant and the two-stage operation is repeated. When compared with the previous intelligent optimization (IRTO) method, the proposed two-stage optimization method can better address the uncertainty nature of the real plant and the computational cost is much lower. From practical industrial experiments, the results obtained show that the proposed optimization method can guarantee that the desired quality of the product quality is achieved in the presence of uncertainty on the plant behavior and the qualities of the raw materials. This outcome suggests that the proposed two-stage optimization method is a practically signiﬁcant approach for the control of alumina blending operation.
Chunjuan Li, Curtin University
Over the last two decades, disaster losses remain substantial because of increasing frequently of disasters. Man-made and natural emergencies cannot be prevented, but they can be better managed. The use of knowledge management (KM) and knowledge management system (KMS) functions for emergency management (EM) is supported and recommended by existing literature. KMS can play an important role in improving the speed and quality of response actions. The KMS Success Model is useful in the broad organizational context of KMS implementation. However, the model requires modification to match the unique nature of emergency situations. Based on KMS Success Model, this paper seeks to emphasize that a KMS for emergency management must incorporate features that enable role changes and allow people to access changes based on the situational requirement.
Chongyi Liu, Curtin University
2D computer-aided design (CAD) drawing compliance checking requires related specialists to do significant amount of workload. To save time and improve productivity, program-based CAD drawing compliance checking systems were introduced. However, they need people to input specifications manually which is time consuming and errors prone. This article introduces an NLP (Natural Language Processing) method to recognize and understand the official specification documents of 2D CAD. It uses up-to-date Convolutional Neural Network (CNN) algorithm to extract all related information from the documents. Also, a serial of test methods will be used to verify if the design complies with the specifications and standards in the documents. Thus, a complete automated road CAD drawing compliance checking system will be developed.
Jie Liu, Curtin University
The multi-time scale calculation and analysis of the power loss of distribution network are carried out by using the massive data collected by the new smart meter. In addition, big data mining technology and machine learning are applied to the analysis and prediction of distribution network loss to reveal the changing rules between the distribution network loss and the electrical/ non-electrical influence factors. Firstly, data cleaning is carried out on the collected data. Outliers are detected by the local factor algorithm based on k-means clustering pruning, and abnormal data are identified and set as vacant values. The missing values are then filled in together with other uncollected data. In view of the complex situation of data missing and the poor effect of the commonly used filling methods, the Random Forest algorithm was improved for missing data imputation, and other filling methods were compared for the data with different loss rates. This imputation algorithm based on Random Forest has high accuracy in filling effect, better robustness and generalization ability. Using the Pearson correlation, the Random Forest’s features importance rank, MIC (Maximal Information Coefficient), with methods of grey correlation analysis to find out the influential factors that have strong correlation with the distribution network loss. The similarity clustering of the distribution network loss is done before the strongly correlated influencing factors and historical time series data are input into LSTM (Long Short-Term Memory) neural network. The network losses are clustered into different categories, and the network losses are predicted separately under each category. Here, the Stacking is adapted by inputting the output of the previously trained models (such as 5 LSTM models) to the second layer model, so as to obtain the final prediction results. This model can gather the advantages of previous training models, and can improve the prediction accuracy. The research result will be of great guidance to the implementation of practical comprehensive loss reduction and efficiency improvement from the perspective of the overall operation of the distribution network. For example, provide the optimization decision for arranging the load access and distribution reasonably and planning of equipment modification.
Yuanyuan Liu, Curtin University
Maximum efficiency transfer (MET) is an important research in the practical application of a wireless power transfer (WPT) system. The impedance matching network on the receiving side plays a very important role in converting the dynamic load impedance to the optimal impedance to achieve MET control in WPT system. The nonlinearity of the rectifier bridge circuit has a significant impact on the transmission characteristics of the system, and the impedance matching network designed based on the pure resistive equivalent rectification load will also lead to inevitable errors. According to the optimal load theory of WPT system in MET control, this paper analyzes the nonlinear characteristics of the rectifying load, and optimizes the design method of T-type impedance matching network. Finally, the effectiveness of the optimization method is verified by simulation and experiments.
Xiumei Lyu, Curtin University, Chongqing Technology and Business University
In view of the fact that China’s Regtech is too lagging behind the Fintech innovation and therefore brings about some prominent problems such as regulatory loopholes, time-lag mismatches and increased financial risks, I study the synergistic innovation mechanism of Regtech and Fintech, and the use of Fintech to enhance Regtech. Through the synergistic innovation mechanism, it can ensure that Fintech can achieve sustainable and healthy development in the direction of improving financial service efficiency under the premise of risk control. Furthermore, it can solve the problem that lack of motivation and ability when Fintech authorities independently develop and use Regtech to enhance supervision.
Dr Elham Mardaneh, Curtin University
Co authors: Professor Ryan Loxton, Dr Qun Lin, Curtin University
Efficient transportation of cargo and personnel is crucial in the offshore oil and gas industry. This paper proposes a transportation scheduling model for offshore oil and gas
operations involving both vessels (for transporting cargo) and helicopters (for transporting personnel). The vessels and helicopters interact through constraints that prohibit simultaneous visits at the offshore production facilities due to safety regulations and limits on personnel availability. The scheduling challenge can be formulated as a mixed-integer linear programming (MILP) model and we define a set of pre-processing operations that exploit model structure to significantly streamline the model. We also present a heuristic algorithm for generating an initial feasible schedule, which can be used as a starting point
for commercial MILP solvers. Simulation results for the North West Shelf project in Australia show that the proposed computational approach can generate high-quality solutions to large, industrial-scale problem instances.
Dr Fred Roosta, University of Queensland
Establishing convergence of the classical Newton’s method has long been limited to making restrictive assumptions on (strong) convexity. Furthermore, smoothness assumptions, such as Lipschitz continuity of the gradient/Hessian, have always been an integral part of the analysis. In fact, it is widely believed that in the absence of well-behaved and continuous Hessian, the application of curvature can hurt more so that it can help.
Here, we show that two seemingly simple modifications of the classical Newton’s method result in an algorithm, called Newton-MR, which can be applied, beyond the traditional convex settings, to invex problems. Newton-MR appears almost indistinguishable from its classical counterpart, yet it offers a diverse range of algorithmic and theoretical advantages.
Furthermore, by introducing a weaker notion of joint regularity of Hessian and gradient, we show that Newton-MR converges globally even in the absence of the traditional smoothness assumptions. Finally, we obtain local convergence results in terms of the distance to the set of optimal solutions. This greatly relaxes the notion of “isolated minimum”, which is required for the local convergence analysis of the classical Newton’s method.
Numerical examples using several machine learning problems demonstrate the great potentials of Newton-MR compared with several other second-order methods.
Dr Bjorn Ruffer, The University of Newcastle
This talk will apply nonlinear stability analysis techniques to the Douglas-Rachford Algorithm, with the aim of shedding light on the interesting non-convex case, where convergence is often observed but seldom proven. The Douglas-Rachford Algorithm can solve optimisation and feasibility problems, provably converges weakly to solutions in the convex case, and constitutes a practical heuristic in non-convex cases. Lyapunov functions are stability certificates for difference inclusions in nonlinear stability analysis.
Ekta Sharma, University of Southern Queensland
Co-author: Dr Ravinesh Deo, Professor Alfio Parisi, University of Southern Queensland
Dr Ramendra Prasad, University of Fiji
Development of practical and efficient air quality regulating mechanisms is indeed a challenge. A pertinent air pollutant responsible for recurrent health-care costs with increased respiratory induced mortality in Australia is Particulate Matter (PM). In such a pressing public health problem, artificial intelligence can provide promising solution via developing models to obtain predictions through novel learning algorithms. This research project, unique in its own kind, focuses on generating real-time air quality forecasts of PM-2.5, PM-10 and the overall lower atmospheric visibility. To establish a robust optimisation model, an online sequential-extreme learning machine (OS-ELM), a powerful artificial intelligence algorithm was integrated with improved empirical mode decomposition with adaptive noise (ICEEMDAN) as a data pre-processing tool. The resulting novel data-driven, hybrid predictive model: ICEEMDAN-OS-ELM registers good predictive ability, particularly appropriate for near real-time forecasts of PM-2.5, PM-10 and visibility at key test sites in Australia. The excellent performance of the OS-ELM hybrid model indicates its suitability as a decision-support systems tool in air quality monitoring, forecasting and subsequent health risk mitigation. Future study aspires to develop robust predictive frameworks in facilitating air quality strategies in mitigating Australia’s public health risks in a coordinated way through advanced practical development of artificial intelligence approaches.
Shaymaa Shraida, Murdoch University
When fluid is pumped from an elevated source it flows downward and then outward once it hits the base. In this talk we consider a simple two dimensional model of flow from a single line source elevated above a horizontal base and consider its downward flow into a spreading layer on the bottom. A hodograph solution and linear solutions are obtained for high flow rates and full nonlinear solutions are obtained over a range of parameter space. It is found that there is a minimum flow rate beneath which no steady solutions exist. Overhanging surfaces are found for a range of parameter values. This flow serves as a model for a two-dimensional water fountain, or approximates a similar flow in a density stratified environment.
Juan Wang, Curtin University
This paper models glycerol metabolism in continuous fermentation as a nonlinear mixed-integer dynamic system (NMIDS) by defining the time-varying metabolic network structure as an integer-valued function. To identify the dynamic network structure and kinetic parameters, we establish a mixed-integer min-max dynamic optimization (MIMMDO) problem with concentration robustness as objective functional. By direct multiple shooting strategy and a decomposition approach consisting of convexification, relaxation and rounding strategy, the MIMMDO problem is transferred into an approximated multistage parameter optimization problem with a large number of variables, which is further solved through a competitive particle swarm optimization (CPSO) algorithm. We also prove that the relaxation problem yields the best lower bound for the MIMMDO problem, and its solution can be arbitrarily approximated by the solution obtained from rounding strategy. Numerical results indicate that the presented NMIDS can better describe cellular self-regulation and response to inhibitions of intermediate metabolites in continuous fermentation of glycerol. Our proposed numerical methods are proved to be effective in solving the large-scale MIDO problem.
Jiapei Wei, Curtin University
When I hold the meeting notice, I was thinking about the following questions: What topic am I good at? I work at a business school and do my research on resource management. What topics are you interested in? All of you are from around the world, focus on mathematics. I need to find something in common.so that I can exchange ideas and insights with you. After a period of observation. I found that everyone is doing a same thing every day, that is, writing articles. Writing an article is start with a review. So citespace is come to me.
Citespace is a bibliometric software. It was developed by Dr. Chaomei Chen, whom is a tenured professor at drexel university. It helps researchers identify scientific literature, discover new trends and developments in scientific development.
The first time I contact with citespace is at the time I’m reading Dr. Chen’s paper by chance. It was a paper on the field of Regeneration Medicine. It was published on Expert Opinion in May 2012.In this paper, Dr. Chaomei Chen systematically reviews the totally unfamiliar research field of “regenerative medicine” with the help of CiteSpace. In order to test the extent to which CiteSpace tools and methods allow a person without relevant expertise to give a valuable overview. Finally, with structural and temporal indicators in the CiteSpace software, professor chaomei Chen identified two papers having excellent performance and important influence in the field of “regenerative medicine”. The papers were written by shinya yamanaka’s team from Kyoto university in Japan. By coincidence, five months after their publication, yamanaka won the 2012 Nobel Prize in physiology for the two papers.
Although our research fields are diverse and our language and culture are quite different, there are also some common things worth discussing together. That’s why I chose to share citespace.
Citespace application example: A review on comprehensive utilization of agricultural resources in China from 1999 to 2018
Jianxiong Ye, Curtin University
Modelling of intracellular dynamics is helpful for understand cellular metabolism. In this talk, we will introduce some basic problems faced in modelling cellular metabolism. Some existing approach will be reviewed and an example will be given to illustrate how to simulate the intracellular fluxes dynamically.
Yanyan Yin, Curtin University
In this paper, the problem of event-triggered mixed H-infinity and passive control for a class of time delay stochastic Markov jump systems subject to input constraint is addressed. In order to reduce network burden, a useful event-triggered scheme is proposed. Then, due to network induced delays, a time-delay model analysis approach is used to reconstruct the system. Analysis and design methods of the state feedback event-triggered controller are derived to ensure that the resulting system is stochastically stable and satisfies mixed H-infinity and passive performance index. Sufficient conditions are obtained in terms of liner matrix inequalities. Finally, a numerical example is given to illustrate the effectiveness of the proposed approach.
Chuanlai Yuan, Curtin University
In this paper, optimization of new car engine lubrication system was discussed in our study. It concluded key analysis of noise pump energy, optimal control strategy of the pump, optimal allocation of real-time detection and sensor pump working parameters. Fuel consumption and operating noise relationship model was built for energy saving and noise reduction as well as energy-saving pump optimization model, which was proposed for solving an effective energy-saving optimization model. Discussion formed a more perfect pump optimization and energy saving noise control method for automotive engine energy saving. In our study, low-power car motor lubrication system consisted of energy noise vane pump was studied. Advanced optimization control theory was used in automotive motor lubrication system, the study of new energy-saving car pump noise optimal control strategy and oil pump working parameters achieved real-time detection and sensor optimize configuration. We built oil fuel with operating noise and working parameters of the relational model for energy-saving and noise reduction, it proposed an energy smart kinds of effective optimization model for forming a more perfect energy-optimized pump control system.