- "Adaptive Machine Learning with Imbalanced Data Sets for Precognitive Maintenance," NUS PI, with A*STAR Singapore Institute of Manufacturing Technology (SIMTech), Singapore, 2015 - 2018, S$300,000.00
Abstract — The imbalanced data learning problem typically occurs during classification due to skewness in the class distribution.
Standard classifiers that are trained to maximize the classification accuracy assume equal misclassification costs, and are overwhelmed by the imbalance in majority which tend to ignore the minority class.
This problem is intrinsic in many application domains such as fault detection, risk management, text classification, and medical diagnosis, etc., and obtaining data for training and prediction plays a vital role in the machine learning pipeline.
This is especially true for sensors in equipment or a manufacturing facilities which work in dynamic environments that evolve over time.
In this project, we aim to develop machine learning tools which improve the performance of adaptive learners for imbalanced data learning and precognitive maintenance of equipment which generates imbalanced datasets.
- "Keppel-NUS Joint Lab on Productivity Enhancement of Yard Operations," Subtheme PI, National Research Foundation (NRF) and Keppel Offshore & Marine Technology Centre (KOMtech), Singapore, 2014 - 2019, S$17,898,040.00
Abstract — Automated fabrication is either not feasible or too costly for large-scale products made to specifications, e.g., large tanks, bridges, power plant equipment, etc.
Offshore and marine is another example where ship and rig components are bulky and difficult to manoeuver.
Traditional stationary robots deployed in the automotive industries are not useful to automate yard operations.
It is necessary to explore mobile and semi-automated welding systems which can navigate through narrow pathways often laced with obstacles.
In many situations, skilled human welders are indispensable for accessing confined spaces to produce quality weld seams.
It is difficult, if not impossible, to replace skilled workers completely with robots as some of the yard operations are highly complex and are beyond the programming and control which can be used to command the automated production systems.
As such, it is necessary to explore human-machine system collaboration in order to achieve optimal results.
In this programme, we explore solutions for increasing productivity and efficiency of yard operations leading to higher productivity in the areas of welding, blasting, painting, and steelwork fabrication and keep Singapore yards ahead of regional and global competition.
Singapore yards are amongst the global leaders in the production of jack-up rigs.
- "Identification and Control of Critical Mechanical Resonant Modes Above Nyquist Frequency (Phase III)," NUS PI, with A*STAR Data Storage Institute (DSI), Singapore, 2014 - 2015, S$490,400.00
Abstract — Mechanical resonant modes are critical to the stability of mechatronic systems when they are out-of-phase (with the rigid body) and are near or above the Nyquist frequency, as they cannot be detected nor controlled due to the limits of sampling and aliasing imposed by the famous Shannon's Sampling Theorem.
With the advancement in high-speed Analog-to-Digital Converters (ADCs) and Digital-to-Analog Converters (DACs), most mechatronic systems have the capabilities of multirate sampling, i.e., different sampling rates exist for different parts of the sampled-data system.
The main objectives of this research project are to identify and control these critical mechanical resonant modes of the mechanical plant, and compensate for high frequency mechanical vibrations in next generation of ultra-high performance mechatronics.
In this project, we aim to study and optimize the identification and control of critical mechanical resonant modes above the Nyquist frequency.
Our results will be validated with rigorous theoretical derivations, coupled with extensive simulation and experimental verifications on Hard Disk Drives (HDDs).
- "Advanced Sensor Fusion and Vibration Control Technologies for Ultra-High Density Hard Disk Drives," PI, EDB-IPP with Western Digital Media Pte. Ltd., Singapore, 2014 - 2017, S$204,000.00
Abstract — With the advancements in magnetic recording technologies, the achievable recording density for perpendicular recording in Hard Disk Drives (HDDs) is reaching the limit of 1 Tbit/in2 due to magnetic physics constraints imposed by the renowned superparamagnetic effect.
To support a recording density of more than 10 Tbit/in2 for next generation of ultra-high density HDDs with specific applications to big data applications, e.g., cloud computing and data centres, etc., advanced magnetic recording technologies with read/write head-positioning accuracies in the Angstrom level are required.
As such, the accuracy of the head-positioning systems needs to be improved and strengthened in presence of external vibrations and measurement noises, as well as mechanical vibrations in the actuators of dual-stage HDDs.
In this project, advanced sensor fusion techniques and vibration control technologies are developed to meet the increasing demand for larger storage capacity via compensating for these vibrations and improving positioning accuracies.
- "Optimization of Non-Fixed Periodic Condition Monitoring/Inspection Interval," NUS PI, with A*STAR Singapore Institute of Manufacturing Technology (SIMTech), Singapore, 2013 - 2016, S$251,500.00
Abstract — Machinery health prognosis is vital to reduce unexpected machine downtime, spares inventory, maintenance costs, and safety hazards, etc.
Currently, there are two types of condition monitoring, namely, continuous and periodic.
The risks of periodic monitoring include the possibility of missing failure events which occur between successive inspections, which could lead to expensive consequences on equipment damage and safety hazards.
If the inspection period is shorter than necessary, additional unnecessary inspection costs and overheads are also incurred on the end user.
As such, the main challenge pertaining to periodic monitoring is the determination of the optimal condition monitoring interval or inspection interval.
In this project, we further our diagnosis and prognosis capabilities to develop intelligent prognosis algorithms for periodical inspection.
Our results will be benchmark against existing monitoring/inspection practices commonly used in many industries for various machine condition, vibration, and electrical inspection services, etc., on realistic engineering systems.
- "Energy-Efficient Flight Control Strategies and Multi-Agent Treatment for Nano-Satellites," PI, Ministry of Education (MOE), Singapore, 2013 - 2016, S$155,600.00
Abstract — With the advancement in consumer electronics, nano-satellites or nanosats are now commonly being used in swarm platforms to work together in formation and achieve objectives previously accomplished by the more costly and bulkier micro- or milli-satellites.
Once propelled into a Keplerian orbit, the nanosats essentially lose many degrees-of-freedom and are degenerated into nonholonomic agents.
In order for the tasks to be scheduled and executed correctly, advanced control laws are required to achieve a desired collective emergent behavior of the nanosat swarm.
As such, it is imperative for the control algorithms to provide the essential scalability and robustness, since nanosats are included and removed from the swarm in an add-on ad-hoc manner.
In this project, we conduct systems-level research on the control aspects of a swarm of nano-satellites.
We develop the essential high-endurance control and mathematical machineries in path-planning and collision avoidance, fully justified with mathematical rigour.
- "A Holistic Systems Design Framework for Ultra-High Performance Precision Mechatronics," PI, Ministry of Education (MOE), Singapore, 2013 - 2016, S$121,160.00
Abstract — Mechatronic systems-the synergetic integration of mechanical, electrical, computational, and control systems-have pervaded into consumer products ranging from large-scale avionics in aircrafts to small-scale integrated sensors in portable consumer devices.
To boost sales and increase revenue in competitive consumer electronics and precision engineering industries, improvements in research and development capabilities to provide a unified systems design framework for ultra-high performance mechatronic systems are essential.
This project is concerned with using a holistic approach and systems thinking to integrate the various components in these sampled-data systems, as it is imperative to consider the crosstalk between different interconnected components for design and manufacturing of next generation of precision mechatronics.
In this project, we will provide the answers to important research questions as the required positioning accuracies in current high-precision mechatronic systems are in the order of Angstrom levels.
Our results will be validated with rigorous theoretical derivations, coupled with extensive simulation and experimental verifications on commercial HDDs, since HDDs are classical examples of high-performance and high-precision mechatronic systems.
The proposed methodologies are also portable and scalable to other mechatronic systems manufactured in current industries.
- "SIMTech-NUS Joint Lab on Precision Motion Systems," Co-PI, A*STAR Singapore Institute of Manufacturing Technology (SIMTech), Singapore, 2012 - 2017, S$4,941,900.00
Abstract — Precision manufacturing has been steadily gathering momentum and attention over the last century in terms of research, development, and application to product innovation.
The driving force in this development stems from requirements for much higher performance of products, higher reliability, longer life, lower cost, and miniaturization.
Today, ultra-precision machine tools under computer control can position the tool relative to the workpiece to a resolution and positioning accuracy in the order better than micrometers.
The new trends in technological advancement can be collectively referred to as nanotechnology, covering nano-fabrication processes, design, behavior, and modeling of nano structures, and methods of measurement and characterization at the nanometer scale.
In the joint laboratory, we will focus on the constituent technologies of precision motion control and their effective integration to fulfil a whole spectrum of industrial needs in the systems and applications relevant to the manufacturing sector, balancing precision, speed, cost, robustness, and configurability.
The constituent technologies include (i) drives and actuation systems, (ii) measurement and control systems, and (iii) system optimisation and applications, etc.
- "A Mechatronic Design Approach for MagLev Motor Development," NUS PI, with A*STAR Singapore Institute of Manufacturing Technology (SIMTech), Singapore, 2013 - 2015, S$580,000.00
Abstract — In recent years, Extreme Ultra-Violet (EUV) lithography processes have emerged as the only solution to fabricate smallest possible features on wafer chips through photo lithographical exposure.
As such, high-resolution wafer fabrication lithography systems are currently developed with an EUV light source.
Within these systems, the silicon wafers are exposed in a high vacuum environment to prevent contamination of optical elements and absorption of the EUV light by air.
Magnetic Levitation (MagLev) has become the only solution to provide frictionless support to the moving stages in such vacuum environment without risking contamination through air, mechanical wear, and mechanical ball-bearing guides.
In this project, we address the fundamental issues of MagLev technologies and enhance their performances with realistic engineering applications to other industries. The developed MagLev motor serves as a platform to (i) conduct fundamental research on mechatronic design approach, (ii) design new concepts in MagLev technologies for coarse-fine positioning/scanning, (iii) propose new control algorithms for dual-actuation systems, and (iv) develop new power electronics devices, etc.
- "Robotics for Autism Assessment," Co-PI, Ministry of Education (MOE), Singapore, 2012 - 2015, S$179,978.00
Abstract — Autism is a complex developmental disability that causes problems with social interaction and communication.
Symptoms usually start before the age of three and can cause delays or problems in many different skills that develop from infancy to adulthood.
Autistic disorders refer to (i) inability to relate to other people, (ii) little use of eye contact with other people, (iii) difficulty in understanding gestures and facial expressions, (iv) difficulties with verbal and non-verbal communication, and (v) difficulty in understanding others' intentions, feelings, and mental states, etc.
Different people with autism can have very different symptoms, and health care providers think of autism as a "spectrum" disorder, i.e., a group of disorders with similar features.
In this project, we make use of a multimodal perception approach (e.g., eye gaze, speech, touch, and gestures, etc.) to analyse the experimental data from two groups of children as they interact with a robotic system, and with robots as the embodiment.
We will also implement data analytics techniques to extract features and classify patterns, by comparing the data from normally developing children and children who have been earlier diagnosed with autism.
- "Collision Avoidance Strategy For LEO (Low Earth Orbit) Satellite," Co-PI, Defence Science Organisation (DSO) National Laboratories, Singapore, 2012 - 2015, S$150,000.00
Abstract — Over the past decades, the number of space objects orbiting the Earth has increased tremendously.
Coupled with a rapid increase in small Cubesats being launched into Low Earth Orbits (LEOs), the risk of existing operational and/or new satellites colliding with other space objects, e.g., orbiting satellites and debris, etc., is becoming a growing concern.
This prompts for the need to develop advanced collision avoidance control laws for maneuvering satellites and avoiding collision with other space objects.
In this project, we develop the essential collision avoidance schemes for small spacecrafts at LEO.
The proposed technology is formulated as an optimal control problem, and the mathematically-justified solution is verified with extensive simulations.
- "Robust Methods for Large-Scale Systems Engineering," Co-PI, Ministry of Education (MOE), Singapore, 2011 - 2014, S$133,500.00
Abstract — Systems Engineering (SE) as a discipline emerges from the need of a scientific, robust and holistic framework to conceptualize, design, manage and implement increasingly complex products and services demanded in the modern world successfully.
Systems complexity arises from the challenging and often-conflicting user requirements, the scale, scope and interconnectivity with different subsystems and the environment, the inter-disciplinary nature of the problems of concern, and the presence of many poorly-perceived system structures.
As a consequence, many new emerging systems and problems are wrought with uncertainty, from both a lack of precedent data, limited knowledge and ambiguity of the system structures.
There is hence a constant need for the development and advancement of SE modelling and analysis methods to effectively mitigate the corrupting issues of uncertainties in the design, analysis and optimization of large-scale complex systems.
Indeed, the issues of uncertainty modelling and management, robust design and multi-criteria optimization are identified as some of the most critical research challenges and opportunities in system-of-systems engineering and design.
In this project, we propose the research to advance and enhance some commonly-used modelling and analysis tools of SE, in order to adapt them to meet the new challenges of large-scale systems complexity and uncertainty.
The system dynamics method, in particular, will be used extensively for system designers and policy-makers to support the elucidation and learning of complex and even 'chaotic' behaviour of large-scale socio-technical-economical systems.
- "Identification and Control of Critical Mechanical Resonant Modes Above Nyquist Frequency (Phase II)," NUS PI, with A*STAR Data Storage Institute (DSI), Singapore, 2013 - 2014, S$300,000.00
- "Identification and Control of Critical Mechanical Resonant Modes Above Nyquist Frequency (Phase I)," NUS PI, with A*STAR Data Storage Institute (DSI), Singapore, 2013, S$186,000.00
- "Intelligent Design and Control of High-Performance Mechatronic Systems," NUS PI, with A*STAR Data Storage Institute (DSI), Singapore, 2011 - 2012, S$636,500.00
Abstract — Currently, the achievable recording density for perpendicular recording in Hard Disk Drives (HDDs) is reaching the limit of 0.75 - 1 Terabit/in2 due to the renowned superparamagnetic effect.
To support 10 Terabit/in2 recording density for next generation of ultra-high density HDDs, advanced magnetic recording technologies like Bit-Patterned Media Recording (BPMR) technology with energy-assisted writing are proposed to reduce the bit-aspect ratio to be less than 4, corresponding to a track density of 1.6 - 3.2 million Tracks-Per-Inch (TPI) with track accuracy at Angstrom levels.
As with all product designs, HDD engineers in various Research and Development (R&D) teams (i.e., Servo, Mechanical, and Production Team) are continuously challenged to arrive at the "optimal" design individually that meets the Business Unit's (BU's) criteria for the various components driven by the HDD technology roadmap.
However, studying the integrity of individual components in an interwoven HDD production process environment solely is insufficient.
In this project, we consider the crosstalk between different teams to architect and facilitate actions amongst various groups of experts in order to progress and support such a high tracking density.
We propose an integrated servo-mechanical approach towards intelligent design and manufacturing of next generation of ultra-high data storage density HDDs.
- "Intelligent Diagnosis and Prognosis of Industrial Networked Systems for Fault Detection," PI, Ministry of Education (MOE), Singapore, 2010 - 2012, S$179,999.00
Abstract — In an era of intensive competition where asset usage and plant operating efficiencies must be maximized, unexpected downtime due to machinery failure has become more costly and unacceptable than before.
To cut operating costs and increase revenues, industries have an urgent need for prediction of fault progression and remaining lifespan of industrial machines. As such, predictive maintenance has been actively perused in the manufacturing industries in recent years where equipment outages are forecasted, and maintenance is carried out only when necessary.
Prediction leads to improved management and hence effective usage of equipment, and multifaceted guarantees are increasingly being given for industrial machines, processes, products, and services, etc.
To ensure successful condition-based maintenance, it is necessary to detect, identify, and classify different kinds of failure modes in the engineering process as early as possible.
In this project, we aim to develop the essential theory and mathematical machineries to provide formal decision software tools which can be readily used for intelligent diagnosis and prognosis of industrial networked systems.
The proposed tools will be applied to realistic engineering processes and systems, fully verified with theoretical simulations and experimental results.
- "Intelligent Monitoring for Energy Efficient Manufacturing," NUS PI, with A*STAR Singapore Institute of Manufacturing Technology (SIMTech), Singapore, 2010 - 2011, S$127,000.00
Abstract — Manufacturing industry has the biggest share in energy consumption and emission of green house gases.
Energy efficient manufacturing will become the priority action because of increasing energy price, emergence of new environmental regulation and growing trend in consumers towards buying "green" products and services.
In the future of green manufacturing, manufacturers will not only compete in product quality and cost, but also in energy efficiency and carbon footprint in manufacturing process.
In order to improve energy efficiency in shop floor manufacturing, companies need to be equipped with intelligent energy measurement and control technology to track energy performance and provide decision support based on accurate and up-to-date information.
This research project aims to investigate the issues associated with energy-aware control, operation and management of the future energy efficient manufacturing.
The research project is targeted at scenarios in which energy consumption will be reduced, where intelligent sensing devices allow users to make informed choices about the control of machines and processes in the factory.
In this project, we develop computational algorithms for online analysis of energy data in the context of manufacturing activities.
The technology will help the factory to assess the energy usage and identify the key energy consumption machines and components, and this information will be used for optimization in the production and enhance the factory's competitiveness.
- "Development of Simulation Toolkit for HTT Head Tester," CI, Hitachi High-Technologies Corporation, Guzo, Kanagawa, Japan, 2007
- "Investigation of Probabilistic Small Signal Stability," CI, Electric Power Research Institute (EPRI), Palo Alto, CA, USA, 2003