Multi-valued control problems and mixture density network

We have proposed a novel robust inversion-based neurocontroller that searches for the optimal control law by sampling from the estimated Gaussian distribution of the inverse plant model. However, for problems involving the prediction of continuous variables, a Gaussian model approximation provides only a very limited description of the properties of the inverse model. This is usually the case for problems in which the mapping to be learned is multi-valued or involves hysteritic transfer characteristics. This often arises in the solution of inverse plant models. In order to obtain a complete description of the inverse model, a more general multicomponent distributions must be modeled. In this paper we test whether our proposed sampling approach can be used when considering an arbitrary conditional probability distributions. These arbitrary distributions will be modeled by a mixture density network. Importance sampling provides a structured and principled approach to constrain the complexity of the search space for the ideal control law. The effectiveness of the importance sampling from an arbitrary conditional probability distribution will be demonstrated using a simple single input single output static nonlinear system with hysteretic characteristics in the inverse plant model.

Combining analytic hierarchy process and goal programming for logistics distribution network design

Logistics distribution network design is one of the major decision problems arising in contemporary supply chain management. The decision involves many quantitative and qualitative factors that may be conflicting in nature. This paper applies an integrated multiple criteria decision making approach to design an optimal distribution network. In the approach, the analytic hierarchy process (AHP) is used first to determine the relative importance weightings or priorities of alternative warehouses with respect to both deliverer oriented and customer oriented criteria. Then, the goal programming (GP) model incorporating the constraints of system, resource, and AHP priority is formulated to select the best set of warehouses without exceeding the limited available resources. In this paper, two commercial packages are used: Expert Choice for determining the AHP priorities of the warehouses, and LINDO for solving the GP model. © 2007 IEEE.

Neural network control of a gas turbine

In this paper we explore the practical use of neural networks for controlling complex non-linear systems. The system used to demonstrate this approach is a simulation of a gas turbine engine typical of those used to power commercial aircraft. The novelty of the work lies in the requirement for multiple controllers which are used to maintain system variables in safe operating regions as well as governing the engine thrust.

Modelling tribology of slider and disk in a computer hard drive system

Issues of wear and tribology are increasingly important in computer hard drives as slider flying heights are becoming lower and disk protective coatings thinner to minimise spacing loss and allow higher areal density. Friction, stiction and wear between the slider and disk in a hard drive were studied using Accelerated Friction Test (AFT) apparatus. Contact Start Stop (CSS) and constant speed drag tests were performed using commercial rigid disks and two different air bearing slider types. Friction and stiction were captured during testing by a set of strain gauges. System parameters were varied to investigate their effect on tribology at the head/disk interface. Chosen parameters were disk spinning velocity, slider fly height, temperature, humidity and intercycle pause. The effect of different disk texturing methods was also studied. Models were proposed to explain the influence of these parameters on tribology. Atomic Force Microscopy (AFM) and Scanning Electron Microscopy (SEM) were used to study head and disk topography at various test stages and to provide physical parameters to verify the models. X-ray Photoelectron Spectroscopy (XPS) was employed to identify surface composition and determine if any chemical changes had occurred as a result of testing. The parameters most likely to influence the interface were identified for both CSS and drag testing. Neural Network modelling was used to substantiate results. Topographical AFM scans of disk and slider were exported numerically to file and explored extensively. Techniques were developed which improved line and area analysis. A method for detecting surface contacts was also deduced, results supported and explained observed AFT behaviour. Finally surfaces were computer generated to simulate real disk scans, this allowed contact analysis of many types of surface to be performed. Conclusions were drawn about what disk characteristics most affected contacts and hence friction, stiction and wear.

Distributed processing, reconfigurable processes and active network

The fast spread of the Internet and the increasing demands of the service are leading to radical changes in the structure and management of underlying telecommunications systems. Active networks (ANs) offer the ability to program the network on a per-router, per-user, or even per-packet basis, thus promise greater flexibility than current networks. To make this new network paradigm of active network being widely accepted, a lot of issues need to be solved. Management of the active network is one of the challenges. This thesis investigates an adaptive management solution based on genetic algorithm (GA). The solution uses a distributed GA inspired by bacterium on the active nodes within an active network, to provide adaptive management for the network, especially the service provision problems associated with future network. The thesis also reviews the concepts, theories and technologies associated with the management solution. By exploring the implementation of these active nodes in hardware, this thesis demonstrates the possibility of implementing a GA based adaptive management in the real network that being used today. The concurrent programming language, Handel-C, is used for the description of the design system and a re-configurable computer platform based on a FPGA process element is used for the hardware implementation. The experiment results demonstrate both the availability of the hardware implementation and the efficiency of the proposed management solution.

Efficiency improvements to the Groupe Special Mobile (GSM) digital mobile radio system

Groupe Spécial Mobile (GSM) has been developed as the pan-European second generation of digital mobile systems. GSM operates in the 900 MHz frequency band and employs digital technology instead of the analogue technology of its predecessors. Digital technology enables the GSM system to operate in much smaller zones in comparison with the analogue systems. The GSM system will offer greater roaming facilities to its subscribers, extended throughout the countries that have installed the system. The GSM system could be seen as a further enhancement to European integration. GSM has adopted a contention-based protocol for multipoint-to-point transmission. In particular, the slotted-ALOHA medium access protocol is used to coordinate the transmission of the channel request messages between the scattered mobile stations. Collision still happens when more than one mobile station having the same random reference number attempts to transmit on the same time-slot. In this research, a modified version of this protocol has been developed in order to reduce the number of collisions and hence increase the random access channel throughput compared to the existing protocol. The performance evaluation of the protocol has been carried out using simulation methods. Due to the growing demand for mobile radio telephony as well as for data services, optimal usage of the scarce availability radio spectrum is becoming increasingly important. In this research, a protocol has been developed whereby the number of transmitted information packets over the GSM system is increased without any additional increase of the allocated radio spectrum. Simulation results are presented to show the improvements achieved by the proposed protocol. Cellular mobile radio networks commonly respond to an increase in the service demand by using smaller coverage areas. As a result, the volume of the signalling exchanges increases. In this research, a proposal for interconnecting the various entitles of the mobile radio network over the future broadband networks based on the IEEE 802.6 Metropolitan Area Network (MAN) is outlined. Simulation results are presented to show the benefits achieved by interconnecting these entities over the broadband Networks.

Capacity improvements for a direct-sequence code division multiple access network

The rapidly increasing demand for cellular telephony is placing greater demand on the limited bandwidth resources available. This research is concerned with techniques which enhance the capacity of a Direct-Sequence Code-Division-Multiple-Access (DS-CDMA) mobile telephone network. The capacity of both Private Mobile Radio (PMR) and cellular networks are derived and the many techniques which are currently available are reviewed. Areas which may be further investigated are identified. One technique which is developed is the sectorisation of a cell into toroidal rings. This is shown to provide an increased system capacity when the cell is split into these concentric rings and this is compared with cell clustering and other sectorisation schemes. Another technique for increasing the capacity is achieved by adding to the amount of inherent randomness within the transmitted signal so that the system is better able to extract the wanted signal. A system model has been produced for a cellular DS-CDMA network and the results are presented for two possible strategies. One of these strategies is the variation of the chip duration over a signal bit period. Several different variation functions are tried and a sinusoidal function is shown to provide the greatest increase in the maximum number of system users for any given signal-to-noise ratio. The other strategy considered is the use of additive amplitude modulation together with data/chip phase-shift-keying. The amplitude variations are determined by a sparse code so that the average system power is held near its nominal level. This strategy is shown to provide no further capacity since the system is sensitive to amplitude variations. When both strategies are employed, however, the sensitivity to amplitude variations is shown to reduce, thus indicating that the first strategy both increases the capacity and the ability to handle fluctuations in the received signal power.

Neural Network Modelling with Input Uncertainty: Theory and Application

It is generally assumed when using Bayesian inference methods for neural networks that the input data contains no noise. For real-world (errors in variable) problems this is clearly an unsafe assumption. This paper presents a Bayesian neural network framework which accounts for input noise provided that a model of the noise process exists. In the limit where the noise process is small and symmetric it is shown, using the Laplace approximation, that this method adds an extra term to the usual Bayesian error bar which depends on the variance of the input noise process. Further, by treating the true (noiseless) input as a hidden variable, and sampling this jointly with the network’s weights, using a Markov chain Monte Carlo method, it is demonstrated that it is possible to infer the regression over the noiseless input. This leads to the possibility of training an accurate model of a system using less accurate, or more uncertain, data. This is demonstrated on both the, synthetic, noisy sine wave problem and a real problem of inferring the forward model for a satellite radar backscatter system used to predict sea surface wind vectors.

An investigation into the network structures of eThylene propylene terpolymer elastomers

A series of ethylene propylene terpolymer vulcanizates, prepared by varying termonomer type, cure system, cure time and cure temperature, are characterized by determining the number and type of cross-links present. The termonomers used represent the types currently available in commercial quantities. Characterization is carried out by measuring the C1 constant of the Mooney Rivlin Saunders equation before and after treatment with the chemical probes propane-2-thiol/piperidine and n-hexane thiol/piperidine, thus making it possible to calculate the relative proportions of mono-sulphidic, di-sulphidic and poly- sulphidic cross-links. The cure systems used included both sulphur and peroxide formulations. Specific physical properties are determined for each network and an attempt is made to correlate observed changes in these with variations in network structure. A survey of the economics of each formulation based on a calculated efficiency parameter for each cure system is included. Values of C1 are calculated from compression modulus data after the reliability of the technique when used with ethylene propylene terpolymers had been established. This is carried out by comparing values from both compression and extension stress strain measurements for natural rubber vulcanizates and by assessing the effects of sample dimensions and the degree of swelling. The technique of compression modulus is much more widely applicable than previously thought. The basic structure of an ethylene propylene terpolymer network appears to be independent of the type of cure system used ( sulphur based systems only), the proportions of constituent cross-links being nearly constant.

Neural network enhanced self tuning adaptive control application for non-linear control of dynamic systems

The main theme of research of this project concerns the study of neutral networks to control uncertain and non-linear control systems. This involves the control of continuous time, discrete time, hybrid and stochastic systems with input, state or output constraints by ensuring good performances. A great part of this project is devoted to the opening of frontiers between several mathematical and engineering approaches in order to tackle complex but very common non-linear control problems. The objectives are: 1. Design and develop procedures for neutral network enhanced self-tuning adaptive non-linear control systems; 2. To design, as a general procedure, neural network generalised minimum variance self-tuning controller for non-linear dynamic plants (Integration of neural network mapping with generalised minimum variance self-tuning controller strategies); 3. To develop a software package to evaluate control system performances using Matlab, Simulink and Neural Network toolbox. An adaptive control algorithm utilising a recurrent network as a model of a partial unknown non-linear plant with unmeasurable state is proposed. Appropriately, it appears that structured recurrent neural networks can provide conveniently parameterised dynamic models for many non-linear systems for use in adaptive control. Properties of static neural networks, which enabled successful design of stable adaptive control in the state feedback case, are also identified. A survey of the existing results is presented which puts them in a systematic framework showing their relation to classical self-tuning adaptive control application of neural control to a SISO/MIMO control. Simulation results demonstrate that the self-tuning design methods may be practically applicable to a reasonably large class of unknown linear and non-linear dynamic control systems.

Parallel implementation of image segmentation algorithms on a network of transputers

Image segmentation is one of the most computationally intensive operations in image processing and computer vision. This is because a large volume of data is involved and many different features have to be extracted from the image data. This thesis is concerned with the investigation of practical issues related to the implementation of several classes of image segmentation algorithms on parallel architectures. The Transputer is used as the basic building block of hardware architectures and Occam is used as the programming language. The segmentation methods chosen for implementation are convolution, for edge-based segmentation; the Split and Merge algorithm for segmenting non-textured regions; and the Granlund method for segmentation of textured images. Three different convolution methods have been implemented. The direct method of convolution, carried out in the spatial domain, uses the array architecture. The other two methods, based on convolution in the frequency domain, require the use of the two-dimensional Fourier transform. Parallel implementations of two different Fast Fourier Transform algorithms have been developed, incorporating original solutions. For the Row-Column method the array architecture has been adopted, and for the Vector-Radix method, the pyramid architecture. The texture segmentation algorithm, for which a system-level design is given, demonstrates a further application of the Vector-Radix Fourier transform. A novel concurrent version of the quad-tree based Split and Merge algorithm has been implemented on the pyramid architecture. The performance of the developed parallel implementations is analysed. Many of the obtained speed-up and efficiency measures show values close to their respective theoretical maxima. Where appropriate comparisons are drawn between different implementations. The thesis concludes with comments on general issues related to the use of the Transputer system as a development tool for image processing applications; and on the issues related to the engineering of concurrent image processing applications.

Novel hardwired distributive tactile sensing system for medical applications

This thesis described the research carried out on the development of a novel hardwired tactile sensing system tailored for the application of a next generation of surgical robotic and clinical devices, namely a steerable endoscope with tactile feedback, and a surface plate for patient posture and balance. Two case studies are examined. The first is a one-dimensional sensor for the steerable endoscope retrieving shape and ‘touch’ information. The second is a two-dimensional surface which interprets the three-dimensional motion of a contacting moving load. This research can be used to retrieve information from a distributive tactile sensing surface of a different configuration, and can interpret dynamic and static disturbances. This novel approach to sensing has the potential to discriminate contact and palpation in minimal invasive surgery (MIS) tools, and posture and balance in patients. The hardwired technology uses an embedded system based on Field Programmable Gate Arrays (FPGA) as the platform to perform the sensory signal processing part in real time. High speed robust operation is an advantage from this system leading to versatile application involving dynamic real time interpretation as described in this research. In this research the sensory signal processing uses neural networks to derive information from input pattern from the contacting surface. Three neural network architectures namely single, multiple and cascaded were introduced in an attempt to find the optimum solution for discrimination of the contacting outputs. These architectures were modelled and implemented into the FPGA. With the recent introduction of modern digital design flows and synthesis tools that essentially take a high-level sensory processing behaviour specification for a design, fast prototyping of the neural network function can be achieved easily. This thesis outlines the challenge of the implementations and verifications of the performances.

Functional neuroimaging and behavioural studies on global form processing in the human visual system

Magnetoencephalography (MEG), functional magnetic resonance imaging (fMRI) and behavioural experiments were used to investigate the neural processes underlying global form perception in human vision. Behavioural studies using Glass patterns examined sensitivity for detecting radial, rotational and horizontal structure. Neuroimaging experiments using either Glass patterns or arrays of Gabor patches determined the spatio-temporal neural responseto global form. MEG data were analysed using synthetic aperture magnetometry (SAM) to spatially map event-related cortical oscillatory power changes: the temporal sequencing of activity within a discrete cortical area was determined using a Morlet wavelet transform. A case study was conducted to determine the effects of strbismic amblyopia on global form processing: all other observers were normally-sighted. The main findings from normally-sighted observers were: 1) sensitivity to horizontal structure was less than for radial or rotational structure; 2) the neural response to global structure was a reduction in cortical oscillatory power (10-30 Hz) within a network of extrastriate areas, including V4 and V3a; 3) the extend of reduced cortical power was least for horizontal patters; 4) V1 was not identified as a region of peak activity with either MEG or fMRI. The main findings with the strabismic amblyope were: 1) sensitivity for detection of radial, rotational, and horizontal structure was reduced when viewed with the amblyopic- relative to the fellow- eye; 2) cortical power changes within V4 to the presentation of rotational Glass patterns were less when viewed with the amblyopic- compared with the fellow- eye. The main conclusions are: 1) a network of extrastriate cortical areas are involved in the analysis of global form, with the most prominent change in neural activity being a reduction in oscillatory power within the 10-30 Hz band; 2) in strabismic amblyopia, the neuronal assembly associated with form perception in extrastriate cortex may be dysfunctional, the nature of this dysfunction may be a change in the normal temporal pattern of neuronal discharges; 3) MEG, fMRI and behavioural measures support the notion that different neural processes underlie the perception of horizontal as opposed to radial or rotational structure.

A process framework for selecting supply system architecture in manufacturing supply chains and networks

The process framework comprises three phases, as follows: scope the supply chain/network; identify the options for supply system architecture and select supply system architecture. It facilitates a structured approach that analyses the supply chain/network contextual characteristics, in order to ensure alignment with the appropriate supply system architecture. The process framework was derived from comprehensive literature review and archival case study analysis. The review led to the classification of supply system architectures according to their orientation, whether integrated; partially integrated; co-ordinated or independent. The classification was combined with the characteristics that influence the selection of supply system architecture to encapsulate the conceptual framework. It builds upon existing frameworks and methodologies by focusing on structured procedure; supporting project management; facilitating participation and clarifying point of entry. The process framework was initially tested in three case study applications from the food, automobile and hand tool industries. A variety of industrial settings was chosen to illustrate transferability. The case study applications indicate that the process framework is a valid approach to the problem; however, further testing is required. In particular, the use of group support system technologies to support the process and the steps involving the participation of software vendors need further testing. However, the process framework can be followed due to the clarity of its presentation. It considers the issue of timing by including alternative decision-making techniques, dependent on the constraints. It is useful for ensuring a sound business case is developed, with supporting documentation and analysis that identifies the strategic and functional requirements of supply system architecture.

Finding structures in information networks using the affinity network

This thesis proposes a novel graphical model for inference called the Affinity Network,which displays the closeness between pairs of variables and is an alternative to Bayesian Networks and Dependency Networks. The Affinity Network shares some similarities with Bayesian Networks and Dependency Networks but avoids their heuristic and stochastic graph construction algorithms by using a message passing scheme. A comparison with the above two instances of graphical models is given for sparse discrete and continuous medical data and data taken from the UCI machine learning repository. The experimental study reveals that the Affinity Network graphs tend to be more accurate on the basis of an exhaustive search with the small datasets. Moreover, the graph construction algorithm is faster than the other two methods with huge datasets. The Affinity Network is also applied to data produced by a synchronised system. A detailed analysis and numerical investigation into this dynamical system is provided and it is shown that the Affinity Network can be used to characterise its emergent behaviour even in the presence of noise.