Permanent URI for this community
Browse
16 results
Search results
Publication INTENSIFIED PROPYLENE PRODUCTION FROM PROPANE VIA CHEMICAL LOOPING TECHNOLOGY(2023-08-07) WANG WEI; CHEMISTRY; Wei ChenPublication Learning and optimization of human motions based on kinematics(2008-04-24) WANG WEI; ELECTRICAL & COMPUTER ENGINEERING; XU JIANXINHUMAN MOTION LEARNING HAS BEEN STUDIED FOR MANY YEARS, AND WIDELY USED IN VARIOUS AREAS, SUCH AS ANIMATION, ROBOTICS, KINEMATICS & INVERSE KINEMATICS, REHABILITATION AND COMPUTATION VISION. THE GOAL OF THIS THESIS IS TO EXPLOIT NEW APPROACHES IN HUMAN MOTION LEARNING AND OPTIMIZATION AND APPLY THEM TO SOLVE SEVERAL HUMAN MOTION LEARNING PROBLEMS. THIS THESIS COMPRISES SEVERAL PARTS WITH EMPHASES ON ITERATIVE LEARNING CONTROL (ILC) IN HUMAN MOTION LEARNING, A GENERAL MODELING APPROACH FOR MOTION LEARNING AND ITS EXTENSIONS, AND OPTIMIZATION IN THE ESTIMATION OF INSTANTANEOUS HELICAL AXIS IN HUMAN ARM ROTATION.
Publication Moment identities and a right tail index in insurance(2003-10-29) WANG WEI; STATISTICS & APPLIED PROBABILITY; YATRACOS, YANNISPublication A TRAINING FRAMEWORK AND ARCHITECTURAL DESIGN OF DISTRIBUTED DEEP LEARNING(2016-08-10) WANG WEI; COMPUTER SCIENCE; OOI BENG CHINDeep learning has recently gained a lot of attention on account of its incredible success in many complex data-driven applications, such as image classification. However, deep learning is quite user-hostile and is thus difficult to apply. For example, it is tricky and slow to train a large model which may consume a lot of memory. This thesis introduces our investigations and approaches towards these challenges. First, we have conducted a comprehensive analysis of optimization techniques for deep learning systems, including stand-alone and distributed training. Second, we have designed and developed a distributed deep learning system, named SINGA, which tackles the usability problem and realizes optimization techniques for distributed training. SINGA provides a flexible system architecture for running different distributed training frameworks. Last, we have proposed deep learning based methods for effective multi-modal retrieval on top of SINGA, which outperform state-of-the-art approaches.Publication THE OPERATING OPTIMISATION OF DC RAILWAY SYSTEM BY FUZZY LOGIC CONTROL AND GENETIC ALGORITHM(1996) WANG WEI; ELECTRICAL ENGINEERING; CHANG CHE SAU; LIEW AH CHOYCost effectiveness is always an important criterion in the operation of a modem rapid transit system (MRTS), and there is a great incentive to reduce the energy cost of railway operation for any MRTS Corporation. In DC railway systems, trains are often equipped with both regenerative and rheostatic braking systems. During regenerative braking, the released power can be either used by nearby trains or returned to an AC power supply system through traction substations (TSSs) equipped with inverters. On the other hand, should over-voltage occur, rheostatic braking will be activated. The recovery of power will thus be limited. In order to achieve maximum power recovery, rheostatic braking must be avoided whenever possible. The selection of braking mode depends on the receptivity condition of the railway system. Railway receptivity is time-varying, and is a function of the positions and power demands of trains, control settings (transformer tap positions and inverters' firing angles) of TSSs, and the operating status of the AC supply distribution network. It is feasible to improve the receptivity of the system by controlling the operating status of TSSs. Other control measures, such as adjustments of the train departure time, can also be considered. In this thesis, the operating optimisation of DC railway systems by fuzzy logic control and genetic algorithms (GAs) is presented. Chapter 2 establishes the mathematical model describing the DC railway operation, and presents the essential characteristics of a typical DC railway system. Chapter 3 develops a simulation algorithm for DC railway operation. The key part is an AC/DC load flow, by which the operating status of the railway system at any instant can be obtained. This is the basis for gain control and TSS operational optimisation, which will be discussed in Chapters 4 through 7. These two Chapters lay a solid foundation for the succeeding Chapters. Chapter 4 proposes a dwell time control strategy for trains at passenger stations. In particular, predictive fuzzy set theory has been applied to achieve multiobjective decision making. The energy-related performance index and its interaction with the regularity performance index are addressed. A robust fuzzy dwell time control over a wide range of railway configurations and operating conditions is verified by test systems. The developed control strategy can be extended to deal with more complicated circumstances. Some other operating performance indices, such as comfort requirement, waiting time limitation and headway, can be taken into account. Simulation results have shown that the method is fast and may be implemented for online control. Chapters 5 through 7 develop a bicriterion optimisation model and a GAs based method to the optimal co-ordination control of firing angles at TSSs. At first, the two objectives, i.e., power saving of the DC railway system and load sharing among TSSs, are clarified, and a bicriterion optimisation model is developed for compromising between these two objectives. Then, a brief survey of optimisation techniques is presented, and GAs are identified to be an efficient tool for solving this bicriterion operating optimisation problem. A typical MRTS is used to test the developed method. The simulation results have shown that the developed mathematical model is correct, and the proposed GAs-based method is efficient and flexible. As much as 6% of power saving can be achieved, and load sharing can be greatly improved at the same time through the proposed co-ordinated control of firing angles.Publication MULTI-SCALE OPTIMAL DESIGN OF SUSTAINABLE FUTURE URBAN ENERGY SYSTEM(2020-01-09) WANG WEI; CHEMICAL & BIOMOLECULAR ENGINEERING; Wang XiaonanWith the background of continuous urbanization and global transformation of energy systems towards decentralized, smart and low carbon manners, the traditional supply-side oriented single-direction urban energy supply paradigm is out of date. The integrated urban energy system emerges, which integrates multiple energy sectors to satisfy the urban electricity, heating, cooling and gas demands simultaneously. However, to determine the best design of the complicated integrated urban energy system is a challenge in the energy system research field. Hence, this thesis establishes the optimization model for the integrated urban energy system with a particular focus on the multi-scale modelling and optimization perspective. Efficient problem-solving strategies are proposed and comprehensive analysis is conducted for such a system to explore the best system design and dispatch strategies. Through a case study, the proposed approach is verified that can efficiently model and solve the optimization problem for the large-scale integrated urban energy systems.Publication SOY SAUCE POWDER PRODUCTION BY SPRAY DRYING AND POWDER PROPERTY CHARACTERIZATION FOR ANTI-CAKING(2012-08-24) WANG WEI; CHEMISTRY; ZHOU WEIBIAOThis research started with choosing maltodextrins as a carrier agent and compared the anti-caking effect of maltodextrins with different dosage and DE value. It further studies the moisture adsorption isotherms and glass transition temperatures of the soy sauce powders by fitting classic mathematical models from which the critical storage conditions of the powders were predicted. After that, some alternative drying carriers were studied for their anti-caking effects on soy sauce powders. On one hand, adding crystalline or semi-crystlline carbohydrates as complementary drying carrier with maltodextrin was investigated. On the other hand, dairy whey protein as a complementary drying carrier was incorporated into soy sauce powders to enhance the storage stability of them. In both studies, improved caking characteristics in the soy sauce powders were demonstrated by adding the complementary drying carriers respectively. Finally, the spray drying process and other heat-treatments of liquid soy sauce on the acrylamide formation in the liquid soy sauce and spray dried powders were explored.Publication Performance analysis of hybrid mobile sensor networks(2008-08-01) WANG WEI; ELECTRICAL & COMPUTER ENGINEERING; VIKRAM SRINIVASAN; CHUA KEE CHAINGIn this thesis, we focus on performance analysis of hybrid mobile sensor networks, where a small number of mobile sensors work with a large number of static sensors to achieve network performance requirements. Three performance metrics, network lifetime, network coverage and localization beacon density, are studied in this thesis. Firstly, we show that the network lifetime can be improved by 4 times with a single mobile relay in an ideal dense network. In randomly deployed networks, an LP-based mobility schedule provides two times network lifetime than static networks. Secondly, we show that the sensor network can be fully covered with the help of mobile sensors which only move once for a short distance. A distributed movement scheduling algorithm is proposed to coordinate the movements of mobiles. Finally, we introduce a new coverage model, called sector coverage, for estimating the beacon density for mobile localization over the network field.Publication Mitigating the Impact of Physical Layer Capture and ACK Interference in Wireless 802.11 Networks(2014-08-01) WANG WEI; COMPUTER SCIENCE; LEONG WING LUP, BENThe interference among 802.11 devices adversely affects network performance. In this thesis, we address two major sources of interference that have received little attention in the literature: i) physical layer capture and ii) MAC Acknowledgment (ACK) frames. Physical layer capture can cause MAC unfairness. We developed a CWmin adjustment protocol, called FairMesh, to slow down the nodes causing unfairness. FairMesh can achieve approximate max-min fairness for arbitrary set of links in 802.11 mesh networks. We also investigated the Message In Message (MIM) mechanism of 802.11 hardware, which could be detrimental to the reception of aggregate frames. We proposed a method to dynamically toggle MIM to achieve near-optimal throughput. The second source of interference is the interference due to MAC ACK frames. We propose MinPACK to minimize ACK interference without affecting the original throughput. MinPACK is complementary to existing data frames power control algorithms and adapts rapidly to dynamic environments.Publication ECONOMIC AND HEALTH RELATED QUALITY OF LIFE OUTCOMES AMONG COMMUNITY-DWELLING DEMENTIA PATIENTS IN SINGAPORE(2016-03-31) WANG WEI; PHARMACY; WEE HWEE LINWe evaluated the annual economic burden of dementia in Singapore overall and stratified it by young-onset dementia (YOD) and late-onset dementia (LOD), and also explored patient factors associated with various cost components. Both EQ-5D-3L and HUI3 were shown to be valid and reliable health utility measures for caregiver ratings of patients with dementia, but less so for self-ratings. HUI3 performed better overall than EQ-5D-3L. Patients with dementia were found to have substantially impaired HRQoL compared to the Singapore general population. The mental health of caregivers could bias their ratings of patient preference and EQ-5D-3L was more affected by the bias than HUI3. We also demonstrated that the caregivers of patients with dementia had significant impairment in the mental aspects of HRQoL, compared to the Singapore general population. YOD caregivers suffered significantly more in the mental aspects of HRQoL and it was probably due to more behavior problems in YOD patients.