Research Students

I have supervised many PhD students to successful completion. A brief description of completed projects and current projects are provided below.

Current Students MPhil/Phd

Hasan Alkhadafe

H. Alkhadafe, A Al-Habaibeh, S Daihzong and A Lotfi, “Optimising Sensor Location for an Enhanced Gearbox Condition Monitoring System”, Journal of Physics: Conf. Ser. Vol. 364, [doi: 10.1088/1742-6596/364/1/012077]

Hasan Alkhadafe, Ahmad Lotfi, Amin AL-Habaibeh, Daizhong SU, “An Investigation into Loose Bearing Prediction and Diagnosis on Gearbox System Performance Measurement” the 24th International Congress on Condition monitoring engineering management (COMDEM 2011), in 1st June 2011, PP. 977-807

Giovanna Martinez Arellano

Thesis title: Optimal integration of wind power in the real-time operationof power systems using enhanced wind forecasting models

Milad Elgargni

Thesis title: Intelligent Fusion of Sensor Signals and Images for Condition Monitoring of Manufacturing Processes

Jia (Michelle) Cui

Thesis title: A study of the energy related occupancy activities in a sample of monitored domestic buildings in the UK

PhD Completed

Saifullizam Puteh

PhD Thesis – User Profiling in the Intelligent Office [Download]

Synopsis – The research aim is to investigate different methods of profiling user activities in an office environment. This will allow optimal use of resources in future Intelligent Office Environments while still taking account of user preferences and comfort. To achieve the goal of this research, a data collection system is designed and built. This required a wireless Sensor Network to monitor a wide range of ambient conditions and user activities, and a software agent to monitor user’s Personal Computer activities. Collected data from different users are gathered into a central database and converted into a meaningful format for description of the worker’s Activity of Daily Working (ADW) and office environment conditions.

Different techniques including Approximate Entropy (ApEn), consistency measures, linear similarity measures and Dynamic Time Warping (DTW) are employed to quantify a user’s behaviour and extract a user profile. The individual user profile is representative of a user’s preferences, consisting of user routine activities, consistency of office usage and their thermal comfort. Using the statistical techniques, consistency and ApEn, it is possible to characterise different users with only a few parameters. Using similarity techniques one can assess the interrelationship of different aspects of a user’s behaviour. This helps to assess the importance of those aspects within the profile. The novel contribution is the use of these techniques within the context of ADW. 

This research investigates soft computing techniques to enhance user profiling. A novel fuzzy characteristic matrix is proposed to summarised the ADW. The activity recognition models using an event-driven and a fuzzy inference system are proposed to recognise a worker’s activities during times when the office is occupied and unoccupied during a workday. The experimental results demonstrate the models recognise a worker’s activities and can classify into six categories (home, lunch, short break, out of office duties, not use computer/lighting and use computer/lighting) with accuracy of more than 90%.

Anthony Ntaki

MPhil Thesis – Autonomous mobility scooter as an assistive tool for the elderly [Download]

Synopsis – he aim of this research is to investigate the development of an autonomous navigation system that could be used as an assistive tool for elderly and disabled people in their activities of daily living. The navigation environment is an urban environment and the platform is a Mobility Scooter (MoS).

To achieve this aim, a differentially steered MoS was modified to receive motion commands from a computer and outfitted with onboard sensors that included a Global Positioning System (GPS) receiver and two 2D planar laser range sensors. Perception methods were developed to detect the presence of an outdoor pedestrian walkway. These methods achieved this by processing the range data produced by the laser sensors to identify features that are typically found around walkways like curbs, low vegetation, walls and barriers. A method that utilises GPS localisation information to plan and navigate a route in an outdoor urban environment was also developed.

Extensive experimental work was conducted to test the accuracy, repeatability and usefulness of the sensory devices. The developed perception methodologies were evaluated in real world environments while the navigation algorithms were predominantly tested in virtual environments.

A navigation system that plans a route in an urban environment and follows it using behaviours arranged in a hierarchy is presented and shown to have the ability to safely navigate an MoS along an outdoor pedestrian path.

Ahmad Al Shami

PhD Thesis – Computational Intelligence for Measuring Macro-Knowledge Competitiveness [Download]

Synopsis – The aim of this research is to investigate the utilisation of Computational Intelligence methods for constructing Synthetic Composite Indicators (SCI). In particular for delivering a Unified Macro-Knowledge Competitiveness Indicator (UKCI) to enable consistent and transparent assessments and forecasting of the progress and competitiveness of Knowledge Based Economy (KBE). SCI are assessment tools usually constructed to evaluate and contrast entities performance by aggregating intangible measures in many areas such as economy, education, technology and innovation. SCI key value is inhibited in its capacity to aggregate complex and multi-dimensional variables into a single meaningful value. As a result, SCIs have been considered as one of the most important tools for macro-level and strategic decision making. Considering the shortcomings of the existing SCI, this study is proposing an alternative approach to develop Intelligent Synthetic Composite Indicators (iSCI). The suggested approach utilizes Fuzzy Proximity Knowledge Mining technique to build the qualitative taxonomy initially, and Fuzzy c-mean is employed to form the new composite indicators.

To illustrate the method of construction for the proposed iSCI, a fully worked application is presented. The presented application employs Information and Communication Technology (ICT) real variables to form a new unified ICT index. The weighting and aggregation results obtained were compared against classical approaches namely Vector Quantisation and Principal Component Analysis, Factor Analysis and the Geometric mean to weight and aggregate synthetic composite indicators. This study also compares and contrasts Optimal Completion Strategy and the Nearest Prototype Strategy to substitute missing values. The validity and robustness of the techniques are evaluated using Monte Carlo simulation.

The developed iSCI concept is generalised to build the suggested UKCI which ultimately is equipped with short-term forecasting capabilities. This achieved by a hybridised model consisting of Artificial Neural Networks and Panel Data: Time Series Cross Sectional to predict and forecast the competitiveness of KBE. The proposed model has the capability of forecasting and aggregating seven major KBE indicators into a unified meaningful map that places any KBE in its league even with limited data points. The Unified Knowledge Economy Forecast Map reflects the overall position of homogeneous knowledge economies, and it can be used to visualise, identify or evaluate stable, progressing or accelerating KBEs. In order to show the value added by the new development techniques, the UKCI is applied to fifty-seven countries initially, then expanded to include the Middle East and North Africa (MENA) region as a special case study. In total seventy-three countries were included, that are representative of developed, developing and underdeveloped economies. The final and overall results obtained, suggest novel, intelligent and unbiased results compared to traditional or statistical methods when building, not only the UKCI, but for any future composite indicator for many other fields.

Sawsan Mahmoud

PhD Thesis – Identification and Prediction of Abnormal Behaviour Activities of Daily Living in Intelligent Environments [Download]

Synopsis – The aim of this research is to investigate efficient mining of useful information from a sensor network forming an Ambient Intelligence (AmI) environment. In this thesis, we investigate methods for supporting independent living of the elderly (and specifically patients who are suffering from dementia) by means of equipping their home with a simple sensor network to monitor their behaviour and identify their Activities of Daily Living (ADL). Dementia is considered to be one of the most important causes of disability in the elderly. Most patients would prefer to use non-intrusive technology to help them to maintain their independence. Such monitoring and prediction would allow the caregiver to see any trend in the behaviour of the elderly person and to be informed of any abnormal behaviour.

Employing a sensor network system allows us to extract daily behavioural patterns of the occupant in an Intelligent Inhabited Environment (IIE). This information is then used to build a behavioural model of the occupant which ultimately is applied to predict the future values representing the expected occupancy in the monitored environment. Challenges of employing wired and wireless sensor network have been widely researched. However, pattern analysis and prediction of sensory data is becoming an increasing scientific challenge and this research investigates appropriate means of pattern mining and prediction within the IIE.

Door entry and occupancy sensors are used to extract the movement patterns of the occupant. These sensors produce long sequences of data as binary time series, indicating presence or absence of the occupant in different areas. It is essential to convert these binary series into a more flexible and efficient format before they are processed for any further analysis and prediction. Different ways of representing and visualizing the large sensor data sets in a format suitable for predicting and identifying the behaviour patterns are investigated.

A two-stage integration of Principal Component Analysis (PCA) and Fuzzy Rule-Based System (FRBS) is proposed to identify important information regarding outliers or abnormal behaviours in ADLs. In the first stage, binary dissimilarities or distance measures are used to measure the distances between the activities. PCA is then applied to find two indices of Hotelling’s T2 and Squared Prediction Error (SPE). In the second stage of the process, the calculated indices are provided as inputs to FRBSs to model them heuristically. They are used to identify outliers and classify them. The proposed system identifies user activities and helps in distinguishing between the normal and abnormal behavioural patterns of the ADLs.

Data provided for this investigation was from real environments and from a previously developed simulator. The simulator was modified to include trending behaviour in the activities of daily living. Therefore, in the occupancy signal generated by the simulator, both seasonality and trend are included in occupant’s movements. Prediction models are built through Recurrent Neural Networks (RNN) after converting the occupancy binary time series. RNN have shown a great ability in finding the temporal relationships of input patterns. In this thesis, RNN are compared to evaluate their abilities to accurately predict the behaviour patterns. The experimental results show that Echo State Network (ESN) and Non-linear Autoregressive netwoRk with eXogenous (NARX) inputs correctly extract the long term prediction patterns of the occupant and outperformed the classical Elman network.

Javad M. Akhlaghinia

PhD Thesis – Occupancy monitoring and prediction in ambient intelligent environment [Download]

Synopsis – Occupancy monitoring and prediction as an influential factor in the extraction of occupants’ behavioural patterns for the realisation of ambient intelligent environments is addressed in this research. The proposed occupancy monitoring technique uses occupancy detection sensors with unobtrusive features to monitor occupancy in the environment. Initially the occupancy detection is conducted for a purely single-occupant environment. Then, it is extended to the multipleoccupant environment and associated problems are investigated. Along with the occupancy monitoring, it is aimed to supply prediction techniques with a suitable occupancy signal as the input which can enhance efforts in developing ambient intelligent environments. By predicting the occupancy pattern of monitored occupants, safety, security, the convenience of occupants, and energy saving can be improved. Elderly care and supporting people with health problems like dementia and Alzheimer disease are amongst the applications of such an environment. In the research, environments are considered in different scenarios based on the complexity of the problem including single-occupant and multiple-occupant scenarios. Using simple sensory devices instead of visual equipment without any impact on privacy and her/his normal daily activity, an occupant is monitored in a living or working environment in the single-occupant scenario. ZigBee wireless communication technology is used to collect signals from sensory devices such as motion detection sensors and door contact sensors. All these technologies together including sensors, wireless communication, and tagging are integrated as a wireless sensory agent.

The occupancy data is then collected from different areas in the monitored environment by installing a wireless sensory agent in each area. In a multiple-occupant scenario, monitored occupants are tagged to support sensory signals in distinguishing them from nonmonitored occupants or visitors. Upon enabling the wireless sensory agents to measure the radio signal strength of received data from tags associated with occupants, wireless localising sensory agents are formed and used for occupancy data collection in the multiple-occupant scenario. After the data collection, suitable occupancy time-series are generated from the collected raw data by applying analysis and suitable occupancy signal representation methods, which make it possible to apply time-series predictors for the prediction of reshaped occupancy signal. In addition, an occupancy signal generator is proposed and implemented to generate sufficient occupancy signal data for choosing the best amongst the prediction techniques. After converting the occupancy of different areas in an environment to an occupancy timeseries, the occupancy prediction problem is solved by time-series analysis and prediction techniques for the single-occupant scenario. The proposed technique has made it possible to predict the occupancy signal for 530 seconds in a real environment and up to 900 seconds for a virtual environment. The occupancy signal generator created based on the proposed statistical model is proved to be able to generate different occupancy signals for different occupant profiles incorporating different environmental layouts.

This can give a good understanding of the occupancy pattern in indoor spaces and the effect of the uncertainty factors in the occupancy time-series. In the multiple-occupant scenario, the tagging technology integrated with data acquisition system has made it possible to distinguish monitored occupants and separate their occupancy signals. Separated signals can then be treated as individual time-series for prediction. All the proposed techniques and models are tested and validated by real occupancy data collected from different environments.

Hemin Shekh Omer – Personal Web Link

PhD Thesis – Mobile robot teleoperation through eye-gaze (TELEGAZE) [Download]

Synopsis – In most teleoperation applications the human operator is required to monitor the status of the robot, as well as, issue controlling commands for the whole duration of the operation. Using a vision based feedback system, monitoring the robot requires the operator to look at a continuous stream of images displayed on an interaction screen. The eyes of the operator therefore, are fully engaged in monitoring and the hands in controlling. Since the eyes of the operator are engaged in monitoring anyway, inputs from their gaze can be used to aid in controlling. This frees the hands of the operator, either partially or fully, from controlling which can then be used to perform any other necessary tasks. However, the challenge here lies in distinguishing between the inputs that can be used for controlling and the inputs that can be used for monitoring. In mobile robot teleoperation, controlling is mainly composed of issuing locomotion commands to drive the robot. Monitoring on the other hand, is looking where the robot goes and looking for any obstacles in the route. Interestingly, there exist a strong correlation between human’s gazing behaviours and their moving intentions. This correlation has been exploited in this thesis to investigate novel means for mobile robot teleoperation through eye-gaze, which has been named TeleGaze for short.

The contribution of this thesis is a well designed and extensively evaluated novel interface for TeleGaze, that enables hands-free mobile robot teleoperation. Since the interface is the only part of an interactive system that the remote user comes into direct contact, the thesis covers different phases of design, evaluation, and critical analysis of the TeleGaze interface. Three different prototypes (Native, Multimodal & Refined Multimodal) have been designed and evaluated using observational and task-oriented studies. The result is a novel interface, that interprets the gazing behaviour of the human operator into controlling commands in an intuitive manner. The interface demonstrates a comparable performance to that of a conventional joystick operated system, with the significant advantage of hands free control, for a number of mobile robot teleoperation applications; provided the limitations of calibration and drift are taken into account.

Leong Ping Tan

PhD Thesis – Dynamic Modelling and Intelligent Control of A Single Screw Extrusion Process

Synopsis – In the plastics industry, single screw extruders are widely used to melt the solid polymer. The extruder contains a helical screw with a varying channel depth along the barrel. It is designed to optimise the efficiency of energy conversion, and the consistency of the molten polymer during the operation. The relative motion between the rotating screw and the stationary barrel continuously shears, melts and pumps the molten polymer out of the extruder die. The extrusion process is generally steady, but it is very difficult to maintain constant operating conditions. This is mainly because the process is subjected to various sources of process disturbances including variations in the quality and quantity of the feed polymer, which can result in poor quality product. Therefore, an effective extrusion controller needs to be developed.

The present extrusion controllers have been mostly concentrated on Proportional-Integral (PI) controllers and Self Tuning Regulators (STR). Generally, the resulting control systems are in Single-Input-Single-Output (SISO) structure. The SISO control systems exhibit a major shortcoming that only one process output could be regulated at each control c)de. Past experience suggests that strong interactions exist between the process parameters. This implies that an encouraging control performance could only be attained if the parameter interactions are taken into consideration while calculating a control action.

In this thesis, an intelligent control system namely Fuzzy supervisory indirect Learning – Predictive Control (FsiLPC) system is proposed. The system is designed based on Model Based Predictive Control (MBPC), Controller Output Error Method (COEM) and Fuzzy Rule Based System (FRBS). The basic operating mechanism of the FsiLPC system is similar to the MBPC system, with one distinctive operating strategy. A control action in the FsiLPC system is calculated by a fuzzy supervisory unit, rather than using a control law as in a MBPC system. To improve the control action, the COEM is employed to tune the parameters of the fuzzy supervisory unit. This strategy allows the system to accept a predictive model of any structure.

The predictive model in the FsiLPC system needs to predict the behaviour of the extrusion process, and also be adaptive to the varying operating conditions. A semi-physical dynamic extrusion model is developed for the needs. The model is governed by a set of partial differential equations, algebraic equations and FRBS sub-models. A hybrid GA-Fuzzy algorithm is implemented to produce an optimal structure for each FBRS sub-model. The sub-models thus obtained show advantages including simpler rule-base and fewer membership functions. These help to improve their interpretability and adaptive ability.

The implementation of the FsiLPC system for the extrusion process has been evaluated by means of simulation studies. The simulation studies include a parametric study and a comparative study. In the parametric study, the characteristics of the FsiLPC system are examined. The results of the study also help in finding suitable settings of the system parameters. The FsiLPC system is then compared with the PI and S1R systems in the comparative study. These three control systems are evaluated based on the performance in tracking the changes of desired process output and minimising the impact of process disturbances. The performance of the FsiLPC system is relatively encouraging.