University of Macau / Faculty of Science and Technology / Department of Mathematics

M.Ing. / It has facilitated human endeavor in every facet of his existence. A staple of civilized society and central to any industrialized or developing economy, the iron and steel industry was, and remains a truly fundamental facet ofhuman culture. It's technological development has paralleled that of humankind, spanning the millennia of his civilized existence. Today's global steel market is plagued with heavy competition from numerous consortiums, where national and political influences serve to further rock an overcrowded boat. The emergence of global steel giants have emphasized a need in smaller players to obtain and maintain a competitive advantage through differentiation, increased throughput, reduced costs and superior quality. These competitive concerns have been augmented by slowing world steel production which amounted to a 6% slowdown over the past 5 years.

Concrete possesses inherently durable characteristics due to having chemical and dimensional stability in most environments. This leads to the perception that reinforced concrete structures are expected to be relatively maintenance free. Unfortunately this is not the case as recent years have shown increased emphasis on the repair and refurbishment of all types of concrete structures, in preference to demolition and rebuilding. New concrete repair methods and repair materials have been developed in order to keep up with the growing demand of the concrete repair industry. Diagnostic techniques are constantly upgraded in the hope of quantifying the extent and nature of the repair work to be undertaken. However, contract documents for concrete rehabilitation contracts are currently drawn up with a flexible approach, which is in favour of the contractor and not the client, as the volume and cost of the contract could escalate to unacceptable levels. This dissertation investigates the development of a new technique to accurately predict concrete repair volumes. Artificial neural networks, digital image processing and software creation is combined to achieve what can be seen as the first step towards a quicker and more accurate concrete repair volume estimation. Once implemented, this could result in a revolution of current quantity surveying techniques used for the estimation of quantities in concrete repair projects. / Prof. P.C. Pretorius

M.Ing. (Electrical and Electronics) / The increased use of power electronic equipment in power networks prompted the development of various topologies to compensate for the distortion in the power networks. The various compensator topologies employ a vast range of converters for the compensation of the different non-active power components. The compensators are either designed to eliminate a specific non-active power component, or a combination of converters is used to simultaneously compensate for several non-active components. The choice of compensator depends largely on the type of load, the distortion levels in the power network, the effectiveness of the compensator and very importantly, the cost implications for the user. Under constant load conditions a particular compensator would suffice. It is however not the case when the load and the accompanied distortion varies with time, which is the case with present non-linear, dynamic high power loads on the network. In these cases,. a need for another compensator or compensation strategy, that is more effective in compensating the changing load condition, exists. It would therefore be advantageous to construct a single compensator from various converters -the hybrid compensator -, so as to enable the user to compensate effectively at all times the distortion caused by his load. In order to be able to operate such a hybrid compensator cost-effectively an intelligent control system capable of constantly monitoring the load and updating the compensation strategy, is needed. Keeping in mind that, with the technology available today, compensators can effectively operate for periods in excess of twenty years, it makes sound economical sense to operate the compensator as cost-optimally as possible. This dissertation investigates the development of an artificial neural network based controller for the cost-optimal control of a hybrid compensator. The hybrid compensator considered consists of the following: A 21 kVAR three phase FF-TCR compensator with LC-fiIters tuned at the 5th, 7th,11th and 13th harmonic frequencies and a 6 kVA three-phase dynamic power filter. The hybrid compensator is to be applied for the compensation of a 25 KVA non-linear load (Inductively loaded controlled rectifier). The above mentioned compensators have been modelled to agree with experimental pilot plants. The complete system with low-level controllers was simulated with EMTP (The Electromagnetic Transients Program). This simulation was used to verify the intelligent controller operation. The neural network based controller that is investigated, consists of a Backpropagation-trained neural network, that continuously analyses the load conditions, considers the operational characteristics and losses of the hybrid compensator and proposes a cost-optimal compensation strategy for the hybrid compensator. The modelling of the hybrid compensator's operational losses and characteristics to enable the cost-effective operation thereof is discussed. Special attention is given to the modelling of the cost-effective control strategy, in the training data used for the training of the neural network controller. The training of the neural network controller, and an evaluation of its behaviour when applied to two different hybrid compensator structures, is also given.

Reliable communication is an important part of everyday living. When transmitting a message over a physical channel, noise is introduced that causes errors in the message. These errors can be inversion errors or synchronization errors. The aim of this thesis is to investigate coding techniques to minimise the effect of synchronization errors that occurred in a transmitted message. Watermarks are inserted into the encoded sequence. A neural network system is implemented before the decoder to detect the watermark and regain synchronization. / Prof. H. C. Ferreira

M.Ing. / This project investigates the use of Self-Organizing Maps (SOM) for feature recognition and analysis in 3D objects. Object data was generated to simulate data obtained from 3D scanning and trained using SOM. The trained data was analysed using speci cally developed software. The feature recognition and analysis process can be summarized as follows: a 3D object le is converted to a pure 3D data le, this data le is trained using the SOM algorithm after which the output is analyzed using a 3D object viewer and SOM data display.

In this thesis an eye-directed controller is developed that slaves the narrow field display within a split-foveal-peripheral-display system to the operator's gaze position. A neural network controller is proposed that directly maps the gaze position to the narrow field projection co-ordinates without the need for any axis or co-ordinate transformations. A novel image feature-extraction algorithm, for extraction of the pupil-purkinje difference measure, has been developed that exhibits robust and reproducible real-time performance. By providing foveal and peripheral vision in a far-field teleoperator through the eye-directed split-foveal-peripheral-display, visual information is sufficiently and naturally provided for the establishment of telepresence. / Dissertation (M Eng (Electronic Engineering))--University of Pretoria, 2007. / Electrical, Electronic and Computer Engineering / unrestricted

This work presents the design, evaluation and comparison of evolution and reinforcement learning models, in isolation and combined in Darwinian and Lamarckian frameworks, with a particular emphasis being placed on their adaptive nature in response to environments that become increasingly unstable. Our ultimate objective is to determine whether hybrid models of evolution and learning can demonstrate adaptive qualities beyond those of such models when applied in isolation. This work demonstrates the limitations of evolution, reinforcement learning and Lamarckian models in dealing with increasingly unstable environments, while noting the effective adaptive nature of a Darwinian model to assimilate increasing levels of instability. This is shown to be a result of the Darwinian evolution model's ability to separate learning at two levels, the population's experience of the environment over the course of many generations and the individual's experience of the environment over the course of its lifetime. Thus, knowledge relating to the general characteristics of the environment over many generations can be maintained in the population's genotypes with phenotype (reinforcement) learning being utilized to adapt a particular agent to the particular characteristics of its environment. Lamarckian evolution, though, is shown to demonstrate adaptive characteristics that are highly effective in response to the stable environments. Selection and reproduction combined with reinforcement learning creates a model that has the ability to utilize useful knowledge produced by reinforcements, as opposed to random mutations, to accelerate the search process. As a result the influence of individual learning on the populations evolution is shown to be more successful when applied in the more direct Lamarckian form. Based on our results demonstrating the success of Lamarckian strategies in stable environments and Darwinian strategies in unstable environments, hybrid Darwinian/Lamarckian models are created with a view towards combining the advantages of both forms of evolution to produce a superior adaptive capability. Our investigation demonstrates that such hybrid models can effectively combine the adaptive advantageous of both Darwinian and Lamarckian evolution to provide a more effective capability of adapting to a range of conditions, from stable to unstable, appropriately adjusting the required degree of inheritance in response to the requirements of the environment.

Please read the abstract in the section 00front of this document / Dissertation (MSc)--University of Pretoria, 2005. / Computer Science / unrestricted

Application of EMG-controlled functional neuromuscular stimulation to a denervated muscle depends largely on the successful discrimination of the EMG signal by which the subject desires to execute control over the impeded movement. This can be achieved by an adaptive and flexible interface regardless of electrodes location, strength of remaining muscle activity or even personal conditions. Adaptability is a natural and important characteristic of artificial neural networks. This research work is restricted to the development of a real-time application of artificial neural network to the EMG signature recognition. Through this new approach, EMG features extracted by Fourier analysis are presented to a multilayer perceptron type neural network. The neural network learns the most relevant features of the control signal. For real-time operation, a digital signal processor operates over the resulting set of weights from the learning process, and maps the incoming signal to the stimulus control domain. Results showed a highly accurate discrimination of the EMG signal over interference patterns.