This thesis is concerned with the applicability of model reference adaptive control to the control of robot manipulators under a wide range of configuration and payload changes, and a comparison of the performance of this technique with that of the non-adaptive schemes. The dynamic equations of robot manipulators are highly nonlinear and are difficult to determine precisely. For these reasons there is an interest in applying adaptive control techniques to robot manipulators. In this work, the detailed performance of three adaptive controllers are studied and compared with that of a non-adaptive controller, namely, the computed torque control scheme. Computer simulation results show that the use of adaptive control improves the performance of the manipulator despite changes in the payload or in the manipulator configuration. Making use of these results, a rule-based controller is developed by dividing a given manipulation task into portions where a particular adaptive control scheme, based on a specific linearized subsystem model, performs best. This strategy of selecting the proper controller during each portion of the overall task yields a performance having the least deviation from the desired trajectory during the entire length of the task.

The rise of environmental awareness as well as the unstable global fossil fuel market has brought about government initiatives to increase electricity generation from renewable energy sources. These resources tend to be geographically and electrically remote from load centres. Consequently many Distributed Generators (DGs) are expected to be connected to the existing Distribution Networks (DNs), which have high impedance and low X/R ratios. Intermittence and unpredictability of the various types of renewable energy sources can be of time scales of days (hydro) down to seconds (wind, wave). As the time scale becomes smaller, the output of the DG becomes more difficult to accommodate in the DN. With the DGs operating in constant power factor mode, intermittence of the output of the generator combined with the high impedance and low X/R ratios of the DN will cause voltage variations above the statutory limits for quality of supply. This is traditionally mitigated by accepting increased operation of automated network control or network reinforcement. However, due to the distributed nature of RES, automating or reinforcing the DN can be expensive and difficult solutions to implement. The Thesis proposed was that new methods of controlling DG voltage could enable the connection of increased capacities of plant to existing DNs without the need for network management or reinforcement. The work reported here discusses the implications of the increasing capacity of DG in rural distribution networks on steady-state voltage profiles. Two methods of voltage compensation are proposed. The first is a deterministic system that uses a set of rules to intelligently switch between voltage and power factor control modes. This new control algorithm is shown to be able to respond well to slow voltage variations due to load or generation changes. The second method is a fuzzy inference system that adjusts the setpoint of the power factor controller in response to the local voltage. This system can be set to respond to any steady-state voltage variations that will be experienced. Further, control of real power is developed as a supplementary means for voltage regulation in weak rural networks. The algorithms developed in the study are shown to operate with any synchronous or asynchronous generation wherein real and reactive power can be separately controlled. Extensive simulations of typical and real rural systems using synchronous generators (small hydro) and doubly-fed induction generators (wind turbines) have verified that the proposed approaches improve the voltage profile of the distribution network. This demonstrated that the original Thesis was true and that the techniques proposed allow wider operation of greater capacities of DG within the statutory voltage limits.

Rotary planing, and moulding, machining operations have been employed within the woodworking industry for a number of years. Due to the rotational nature of the machining process, cuttermarks, in the form of waves, are created on the machined timber surface. It is the nature of these cuttermarks that determine the surface quality of the machined timber. It has been established that cutting tool inaccuracies and vibrations are a prime factor in the form of the cuttermarks on the timber surface. A principal aim of this thesis is to create a control architecture that is suitable for the adaptive operation of a wood planing machine in order to improve the surface quality of the machined timber. In order to improve the surface quality, a thorough understanding of the principals of wood planing is required. These principals are stated within this thesis and the ability to manipulate the rotary wood planing process, in order to achieve a higher surface quality, is shown. An existing test rig facility is utilised within this thesis, however upgrades to facilitate higher cutting and feed speeds, as well as possible future implementations such as extended cutting regimes, the test rig has been modified and enlarged. This test rig allows for the dynamic positioning of the centre of rotation of the cutterhead during a cutting operation through the use of piezo electric actuators, with a displacement range of ±15μm. A new controller for the system has been generated. Within this controller are a number of tuneable parameters. It was found that these parameters were dependant on a high number external factors, such as operating speeds and run‐out of the cutting knives. A novel approach to the generation of these parameters has been developed and implemented within the overall system. Both cutterhead inaccuracies and vibrations can be overcome, to some degree, by the vertical displacement of the cutterhead. However a crucial information element is not known, the particular displacement profile. Therefore a novel approach, consisting of a subtle change to the displacement profile and then a pattern matching approach, has been implemented onto the test rig. Within the pattern matching approach the surface profiles are simplified to a basic form. This basic form allows for a much simplified approach to the pattern matching whilst producing a result suitable for the subtle change approach. In order to compress the data levels a Principal Component Analysis was performed on the measured surface data. Patterns were found to be present in the resultant data matrix and so investigations into defect classification techniques have been carried out using both K‐Nearest Neighbour techniques and Neural Networks. The application of these novel approaches has yielded a higher system performance, for no additional cost to the mechanical components of the wood planing machine, both in terms of wood throughput and machined timber surface quality.

published_or_final_version / Mechanical Engineering / Doctoral / Doctor of Philosophy

published_or_final_version / abstract / Industrial and Manufacturing Systems Engineering / Doctoral / Doctor of Philosophy

The thesis investigates two domains, initially the kite domain and then part of a more demanding Rolls-Royce domain (jet engine design). Four main types of refinement rules that use the associated application conditions and domain ontology to support the maintenance of constraints are proposed. The refinement rules have been implemented in ConEditor and the extended system is known as ConEditor+. With the help of ConEditor+, the thesis demonstrates that an explicit representation of application conditions together with the corresponding constraints and the domain ontology can be used to detect inconsistencies, redundancy, subsumption and fusion, reduce the number of spurious inconsistencies and prevent the identification of inappropriate refinements of redundancy, subsumption and fusion between pairs of constraints.

D.Phil. / A modern day N-P complete problem is the assigning of frequencies to transmitters in a cellular network in such a manner that, ideally, no two transmitters in the same cell or neighbouring cells use the same frequency. Considering that an average cellular network provider has over 29 000 transmitters and only 55 frequencies, choosing these frequencies in an optimal way is a very difficult computational problem. Swarm intelligence allows the acceptable minimization and optimization of the frequency assignment problem (FAP). Swarm intelligence is a concept modelling the processes in natural systems such as ant colonies, beehives, human immune systems and the human brain. These systems are selforganizational and display high efficiency in the execution of their tasks. A number of simple automated agents interacting with each other and the environment form a collective. Specifically, there is no "central agent" directing the others. A collective can display surprising intelligence which emerges out of the interaction of the individual agents. This collective intelligence, referred to as swarm intelligence, is displayed in ant colonies when ants build elaborate nests, regulate nest temperature and efficiently search for food in very complex environments. In this thesis a proposal is made to utilize swarm intelligence to build a swarm automatic frequency planner (swarm AFP). The swarm AFP produces frequency plans that are better, or on par with existing frequency planning tools, and in a fraction of the time. A swarm AFP is presented through an in-depth investigation into complex adaptive systems, agent architectures and emergence. Based on an understanding of these concepts, a swarm intelligence model called ACEUS is constructed. ACEUS forms the platform of the swarm AFP. It is a contribution to multi-agent technology as it is a new multi-agent framework that exhibits swarm intelligence and complex distributed computation. What differentiates ACEUS from other multi-agent technologies is that ACEUS works on the basis that the tasks or constructions that have been created by the agents actually guide the agents in their endeavours. There is no centralised agent controlling or guiding the process. The agents in ACEUS receive information and stimulation from their tasks or constructions in the environment. As these constructions or tasks alter the environment, the agents receive stimulus from the changing environment and then react to the changing environment. The changing environment acts as an emergent guiding force to the agents. This is the important contribution that stigmergy contributes to ACEUS. Utilizing this concept, ACEUS is used to create a swarm AFP. The swarm AFP is benchmarked against the COST 259 Siemens benchmarks. In all the COST 259 Siemens scenarios the swarm AFP produced the best results in the shortest time. The swarm AFP was also tested in a real cellular network and the resulting statistics before and after the swarm AFP implementation are presented.

A dissertation submitted to the Faculty of Engineering, University of the Witwatersrand, .Johannesburg, in fulfilment of the requirements for the degree of Master of Science in Engineering. / As industrial plants become more complex. there is a growing need for new approaches to control and supervision. This research investigates the issues involved in applying Artificial Intelligence (AI) techniques in the real-world engineering problem of process control supervision. Current AI theory is examined and some techniques modified to design a general-purpose, reactive planner. The planner forms the basis of a supervisory control system. The system is implemented and interfaced with an existing Laboratory plant, so that its performance can he tested and evaluated by comparing with a conventional feedback controller This real life testing necessitates explicit treatment of issues such as data: sampling. situation assessment and CPU scheduling. The case study shows that by combining AI techniques with conventional control, a system can be built which displays superior performance under normal operating conditions and which can deal with abnormal conditions such as equipment failures. / Andrew Chakane 2018

Orientador: Carlos Roberto Minussi / Banca: Anna Diva Plasencia Lotufo / Banca: Maria do Carmo Gomes da Silveira / Banca: Edivaldo Romanini / Banca: Fernando Nogueira de Lima / Resumo: A evasão escolar é um dos problemas mais complexos e cruciais no âmbito da educação. Está presente e é motivo de preocupação nos vários níveis e modalidades de ensino, além de ferir o princípio da dignidade humana. No que tange ao ensino superior, internacionalmente, o fenômeno é objeto de atenção e de cuidado, no intuito de aumentar os índices de permanência e conclusão dos estudantes de graduação e minimizar os prejuízos sociais, econômicos, políticos, acadêmicos e financeiros causados a todos os envolvidos no processo educacional. Nesse contexto, é imprescindível o desenvolvimento de métodos e instrumentos eficientes e eficazes para predição, avaliação e acompanhamento de estudantes em risco de evasão, possibilitando o planejamento e a adoção de medidas proativas no intuito de minimizar a situação. Assim sendo, esta pesquisa tem por objetivo apresentar as potencialidades de um sistema inteligente capaz de identificar, de maneira proativa, continuada e acurada, o grupo de risco de evasão discente, da educação clássica-presencial, no ensino de nível superior. No desenvolvimento deste sistema foi utilizada uma das técnicas da inteligência artificial, as Redes Neurais Artificiais, mais especificamente, a Rede Neural ARTMAP-Fuzzy, uma rede neural da família ART (Adaptive Resonance Theory) que possibilita o aprendizado continuado do sistema. Para o treinamento e teste da Rede Neural e, posteriormente, a validação do sistema proposto foram utilizados os dados socioeconômicos e acadêmicos dos estudantes matriculados nos cursos superiores de tecnologia do Instituto Federal de Mato Grosso - IFMT. Os dados que compuseram os vetores de entrada do sistema foram coletados a partir de dois bancos de dados do sistema de informação do IFMT, respectivamente, o Q-seleção e o Q-Acadêmico. Este sistema faz a classificação dos padrões de entrada em propensos ... / Abstract: School dropout is one of the most complex and crucial problems in the field of education. It permeates and afflicts the several levels and teaching modalities, besides hurting the principle of human dignity. In relation to higher education, internationally, the phenomenon is an object of attention and care, aiming to increase the indexes of permanence and completion rate of the undergraduate students and minimize social, economic, political and financial damage caused to all involved in the educational process. In this context, it is fundamental to develop efficient and effective methods and instruments for prediction, assessment and monitoring of the students at risk of dropping out, making the planning and the adoption of proactive actions possible for the improvement of the situation. Thus, this study aims to present the potentialities of an intelligent system able to identify, in a proactive, continued and accurate way, the student dropout risk group in higher education classroom courses. In the development of this system one of the artificial intelligence techniques was used, the Artificial Neural Networks, more specifically, the Fuzzy-ARTMAP Neural network, a neural network of the ART (Adaptive Resonance Theory) family which makes the continued learning of the system possible. For the training and test of the Neural Network and, later, the validation of the system proposed the socio-economic and academic records of the students enrolled in the technology courses of the Federal Institute of Mato Grosso - IFMT were used. The data that constituted the input vectors of the system were extracted from two database of the IFMT information system, respectively, the Q-selection and the Q-Academic. This system classifies the input patterns in school dropout propensity. The consistence of the results, showing a success rate of the dropout group around 95% and 100% and the overall mean accuracy around ... / Doutor

novel approach to extend the decision-making capabilities of unmanned surface vehicles (USVs) is presented in this work. A multi-objective framework is described where separate controllers command different behaviors according to a desired trajectory. Three behaviors are examined – transiting, station-keeping and reversing. Given the desired trajectory, the vehicle is able to autonomously recognize which behavior best suits a portion of the trajectory. The USV uses a combination of a supervisory switching control structure and a reinforcement learning algorithm to create a hybrid deliberative and reactive approach to switch between controllers and actions. Reinforcement learning provides a deliberative method to create a controller switching policy, while supervisory switching control acts reactively to instantaneous changes in the environment. Each action is restricted to one controller. Due to the nonlinear effects in these behaviors, two underactuated backstepping controllers and a fully-actuated backstepping controller are proposed for each transiting, reversing and station-keeping behavior, respectively, restricted to three degrees of freedom. Field experiments are presented to validate this system on the water with a physical USV platform under Sea State 1 conditions. Main outcomes of this work are that the proposed system provides better performance than a comparable gain-scheduled nonlinear controller in terms of an Integral of Absolute Error metric. Additionally, the deliberative component allows the system to identify dynamically infeasible trajectories and properly accommodate them. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2016. / FAU Electronic Theses and Dissertations Collection