Orientador: Roger Josef Zemp / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Química / Made available in DSpace on 2018-08-23T09:46:30Z (GMT). No. of bitstreams: 1 Santos_RaphaelRibeiroCruz_M.pdf: 3339099 bytes, checksum: 221059e3eab5645623f894afee5431de (MD5) Previous issue date: 2013 / Resumo: Neste trabalho é apresentada uma metodologia de desenvolvimento de um sistema supervisório utilizando Microsoft Excel®. O sistema foi desenvolvido para operar uma planta de polimerização em solução de estireno. A aplicação de códigos computacionais na em VBA (Visual Basic for Applications) na planilha torna possível a criação de um cliente OPC para comunicação entre o computador e hardware. A planilha possui, portanto condições de receber, enviar, armazenar e tratar informações vindas do processo. A polimerização de estireno é realizada utilizando tolueno como solvente e BPO (Peroxido de Benzoíla) como iniciador. Aquisição da temperatura do reator, temperatura de entrada e saída da camisa e massa especifica do meio reacional foram realizada. A partir dos dados experimentais extraídos da reação de polimerização, foi gerado um modelo empírico utilizando Redes Neurais Artificiais (RNA). As RNAs são implementadas no Excel de forma simples, usando a própria planilha, por operar matrizes, tornando-se oportunas para o desenvolvimento de controladores preditivos. O Microsoft Excel® mostrou-se uma interessante ferramenta para aplicação em automação de protótipos experimentais / Abstract: This research work presents a methodology for the development of a supervisory system using Microsoft Excel®. The system was designed to operate a styrene polymerization plant. The development of computational codes in VBA (Visual Basic for Applications) allows for the creation of an OPC client for communication between computer and hardware. The spreadsheet has porting is able to receive, send, store and process information from the process. The polymerization of styrene is carried out using toluene as solvent and BPO (benzoyl peroxide) as an initiator. Acquisition of reactor temperature, inlet temperature and outlet shirt and specific mass of the reaction medium were performed. From the experimental data extracted from the polymerization reaction, an empirical model was generated using Artificial Neural Networks (RNA). RNAs are easily implemented in Excel simply, using operations arrays, making it appropriate for the development of predictive controllers. The use of Microsoft Excel® proved to be an interesting tool for application in automation experimental prototypes / Mestrado / Engenharia Química / Mestre em Engenharia Química

Orientador: Alberto Luiz Serpa / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecânica / Made available in DSpace on 2018-08-23T23:04:02Z (GMT). No. of bitstreams: 1 ArizaZambrano_WilliamCamilo_M.pdf: 5789145 bytes, checksum: 151f5e3ef1780d5448a7073b85b4715f (MD5) Previous issue date: 2013 / Resumo: Este trabalho tem como objetivo principal o estudo de um método de controle baseado no uso de redes neurais artificiais aplicado ao problema de controle de vibrações em estruturas flexíveis. Este trabalho centra-se no estudo do esquema de controle inverso-direto, que consiste em identificar a dinâmica inversa da planta através de uma rede neural artificial para ser usada como controlador. Três exemplos de aplicação foram resolvidos utilizando-se controladores projetados com o método inverso-direto. A primeira aplicação é o controle de vibrações em uma estrutura mecânica de parâmetros concentrados. O segundo exemplo de aplicação é o controle de vibrações de uma placa engastada em uma de suas extremidades. Neste caso, a placa engastada foi modelada utilizando-se o método de elementos finitos. No seguinte exemplo, o modelo da placa usado no exemplo anterior foi reduzido, deixando apenas os primeiros modos de vibração. No último exemplo tratou-se o problema de controle não colocado das vibrações em uma placa engastada. Os resultados foram analisados a partir da resposta temporal e da resposta em frequência do sistema em malha fechada. Para comparar os resultados obtidos utilizando-se o método de controle baseado em redes neurais artificiais, os exemplos citados anteriormente foram também resolvidos utilizando-se o método de controle ??. Os resultados obtidos demonstram que o método de controle baseado em modelo inverso usando redes neurais foi eficaz na resolução deste tipo de problema / Abstract: The goal of this work is to study a control method based on artificial neural networks applied to the vibration control of flexible structures problem. This work focuses in the direct-inverse control scheme which consists of identifing the inverse dynamics of the plant through an artificial neural network to be used as the controller. Three application examples using the direct-inverse method were solved. The first application is the vibration control in a mechanical structure of concentrated parameters. The second application example is the vibration control of a cantilever plate. The cantilever plate was modeled using the finite elements method. In the third example, a reduction of the cantilever plate model was made. In the last example a non-collocated control problem of vibration in a cantilever plate was treated. The results of the scheme were evaluated according to the temporal response and the frequency response of the closed-loop system. In order to compare the results obtained using the control method based on artificial neural networks, the previous examples were also solved using the ?? control method. The obtained results show that the control method based on inverse model using neural networks was effective in solving this kind of problem / Mestrado / Mecanica dos Sólidos e Projeto Mecanico / Mestre em Engenharia Mecânica

Orientadores: Ana Maria Frattini Fileti, Florival Rodrigues de Carvalho / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Quimica / Made available in DSpace on 2018-08-03T02:39:01Z (GMT). No. of bitstreams: 1 Vieira_WilliamGoncalves_D.pdf: 7050309 bytes, checksum: c1e288d81c3eebed686e86e3bcc5edda (MD5) Previous issue date: 2002 / Resumo: A unidade de Craqueamento Catalítico em Leito Fluido - FCC, modelo Kellogg Orthoflow F., representa um processo de refino de petróleo apresentando característica altamente não linear, possuindo fortes interaçães entre as variáveis de produção, e condições de operação extremamente severas. Essas unidades são constituídas basicamente de duas seções: uma de reação catalítica na qual ocorrem as reações de quebra de cadeia hidrocarbônica e também há formação de coque, desativando o catalisador; e outra seção onde ocorre a regeneração do catalisador desativado. O objetivo dessa unidade é transformar produtos de elevado peso molecu1ar, que apresentam baixo valor agregado, em compostos de elevado valor comercial. As unidades FCC, devido às condições severas de operação, necessitam de um controle rigoroso de determinadas variáveis operacionais. Apesar de existirem instalados controladores avançados baseados em modelos de convolução, fteqüentemente essas unidades são reguladas por meio de controladores PID padrões e também através de controle manual baseado no conhecimento de operadores das refinarias. O presente estudo tem como objetivo desenvolver um controlador preditivo multivariável (Multivariable Predictive Control - MPC) para ser implementado na unidade FCC, utilizando Redes Neurais Artificiais (RNA) como modelo interno do controlador. Inicialmente é previsto realizar a identificação do processo da FCC em RNA, obedecendo a seguinte estratégia: usando um modelo fenomenológico que representa a unidade industrial, e partindo de um estado inicial são aplicados diversos degraus nas variáveis manipuladas analisando as respostas nas variáveis controladas do processo. A partir destas simulações são gerados diversos conjuntos de dados divididos em grupos de treinamento, validação e teste. Diversas redes neurais do tipo multicamada feedforward são então criadas para representar o modelo fenomenológico, sendo selecionada aquela que apresenta melhor desempenho, quando comparada com o modelo. A configuração da RNA escolhida como modelo interno foi 8x15x4 (camadas de entrada, escondida e de saída, respectivamente) apresentando um erro relativo máximo de 1% quando comparado com os resultados do modelo rigoroso. Posteriormente, foi previsto desenvolver um controlador preditivo multivariável usando como modelo interno esta rede selecionada. Este controlador foi implementado dentro da rotina do modelo fenomenológico, sendo então realizados testes para verificar seu desempenho, comparando o resultado com o sistema aberto e também com o controlador DMC (Dynamic Matrix Contro!) existente. Diversos horizontes de predição e controle foram analisados, sendo selecionados aqueles que apresentaram melhor desempenho. Foi introduzido um ruído nos sinais do modelo fenomenológico para testar a robustez do controlador proposto. O controlador apresentou bom desempenho mesmo na presença de ruídos de 1,5%, levando sempre as variáveis controladas para seus valores de referência, o que comprova sua robustez. Baseados nestes resultados, conclui-se que um controlador preditivo multivariável baseado em RNA é perfeitamente capaz de controlar um sistema não linear de porte do FCC, onde elevada interação entre suas variáveis operacionais e fortes restrições estão presentes. Isto nos permite extrapolar que são boas as expectativas para uma futura utilização na unidade industrial, principalmente devido à sua simplicidade, robustez e facilidade de implementação, a despeito da dificuldade de sintonia do controlador / Abstract: The Fluid Cracking Catalytic unit - FCC, Kellogg Orthotlow F. model, represents a very strong nonlinear process, with severe interactions among the process variables, and extremely severe operation conditions. The unit is composed of two sections: one is the catalytic reaction, where the hydrocarbon breaks chain reactions and coke deposition take place becoming the catalyst inactive, and the other where the catalyst regeneration happens. The objective is to transform products derived ITom petroleum, with high molecular weight and low added value, into products with higher profit. Due to the severe operation conditions, rigorous control of some variable is needed. In spite of the existence of advanced control based on a convolution model, in practice, FCC units are ftequently regulated by standard PID controllers, and also through manual control actions based on the knowledge of the refinery operators. The objective of this study is to develop a Multivariable Predictive Control (MPC) to be implemented in the FCC unit, using the Artificial Neural Networks (ANN) as internal model. Initially, the process identification in ANN of the FCC was done by the following strategy: an initial state was fust achieved using numerical simulations based on the phenomenological mo deI. Then, several steps changes were applied to the manipulated variables and the response in the controlled variables were monitored and recorded. From these simulations, several groups of data were generated for training, validation and testing. The Neural Network of multilayer feedforward type were created to represent the phenomenological model, being selected the one that better represents the phenomenological model. The ANN configuration chosen to be the internal model was 8x15x4 (Input x Hidden x Output) architecture, with a maximum relative error below 1 % when comparing the results with the phenomenological model results. Later on, it was developed a multivariable predictive control based on this internal model. This control was implemented inside the routine of the phenomenological model. The performance tests were evaluated comparing the results with the open system and with the Dynamics Matrix Control (DMC). Several prediction and control horizons were analyzed. The ANN control presented good performance even in the presence of noise of 1,5% of intensity, taking back the controlled variables to its setpoints, proving its robustness. Based on these results, a multivariable predictive control based on ANN showed be perfectly able to control a nonlinear system like a FCC unit, where high interactions among process variables, and strong restriction conditions exists. This allows us to have good expectations for a future use in the industrial unit, mainly due to its simplicity, robustness and facility ofuse, in spite ofthe difliculty oftune control / Doutorado / Sistemas de Processos Quimicos e Informatica / Doutor em Engenharia Química

Orientador: Douglas Eduardo Zampieri / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Mecanica / Made available in DSpace on 2018-08-04T03:12:52Z (GMT). No. of bitstreams: 1 Goncalves_JoaoBosco_D.pdf: 9689870 bytes, checksum: db2fb279a1080765ab5bddf9068a356d (MD5) Previous issue date: 2004 / Resumo: Este trabalho concebeu um robô bípede composto por uma sucessão de elos rígidos interconectados por 12 articulações rotativas, permitindo movimentos tridimensionais. O robô bípede é constituído por dois subsistemas: tronco e membros inferiores. A modelagem matemática foi realizada em separado para cada um dos subsistemas, que são integrados pelas forças reativas de vínculo. Nossa proposta permite ao robô bípede executar a andadura dinâmica utilizando o tronco para fornecer o balanço dinâmico (estabilidade postural). De forma inédita, foi desenvolvido um gerador automático de trajetória para o tronco que processa as informações de posições e acelerações impostas aos membros inferiores, dotando o robô bípede de reflexos. Foi desenvolvido um gerador de marcha que utiliza a capacidade do robô bípede de executar movimentos tridimensionais, implicando andadura dinamicamente estável sem a efetiva utilização do tronco. O gerador automático de trajetória para o tronco entra em ação se a marcha gerada não mantiver o balanço dinâmico, restabelecendo uma marcha estável. Foi projetado um sistema de controle adaptativo por modelo de referência que utiliza redes neurais artificiais. A avaliação de estabilidade é feita segundo o critério de Lyapunov. O sistema de controle e o gerador automático de trajetórias para o tronco são integrados, compondo os mecanismos adaptativos desenvolvidos para solucionar o modo de andar dinâmico / Abstract: The main objective of this work is to project a biped robot machine with a trunk. The mathematical model was realized by considering two sub-systems: the legs and the trunk. The trajectories of the trunk are planned to compensate torques inherent to the dynamic gait, permitting to preserve the dynamic balance of the biped robot. An automatic generator of trajectory for the trunk was developed that processes the infonnation of positions and accelerations imposed to the legs. A gait generator was developed that uses the capacity of the biped robot to execute three-dimensional movements, causing a steady dynamic gait without the effective use of the trunk. The automatic generator of trajectory for the trunk actuates, if the generated do not keep the dynamic balance, reestablishing he steady dynamic gait. A neural network reference model for the adaptive control was projected, which utilizes an RBF neural network and a stability evaluation is based on the criterion of Lyapunov. The system of control and the automatic generator of trajectories for the trunk are integrated, composing the adaptive mechanisms developed to solve the way of dynamic walking / Doutorado / Mecanica dos Sólidos e Projeto Mecanico / Doutor em Engenharia Mecânica

Orientador: Saide Jorge Calil / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação / Made available in DSpace on 2018-08-04T18:29:09Z (GMT). No. of bitstreams: 1 FerreyraRamirez_ErnestoFernando_D.pdf: 2533123 bytes, checksum: 4ec959f90e4b7a78bc857107a1a5d844 (MD5) Previous issue date: 2005 / Resumo: No Brasil, existe uma parcela significativa de equipamentos médico-hospitalares inoperantes devido à condução inadequada, feita por pessoas despreparadas, do processo de aquisição desses equipamentos. Visando uma futura solução para esse problema, nesta tese foi desenvolvido um estudo para mostrar a possibilidade de representar (através da utilização de redes neurais artificiais) o processo cognitivo utilizado por engenheiros clínicos experientes durante a fase de ponderação dos critérios para julgamento de propostas de fornecimento de equipamentos médicos. Para isso, as respostas fornecidas, a uma pesquisa com engenheiros clínicos de várias regiões do país, foram usadas para construir exemplos para treinamento de diversas arquiteturas de redes neurais. Os melhores resultados (maior correlação com as respostas originais e menor erro quadrático de teste) foram obtidos para a composição (ensemble) de 100 redes neurais de duas camadas escondidas treinadas com o algoritmo back-propagation. Isso mostrou a viabilidade de representar o conhecimento dos especialistas na forma de um modelo conexionista não-linear, cujas saídas fornecem a importância de diversos fatores (clínico, financeiro, qualidade, segurança e técnico) envolvidos no processo de julgamento de propostas para aquisição de um equipamento médico / Abstract: Most recently, in Brazil, there are evidences of a great number of useless medical equipment, due to the absence of experienced professionals to conduct an effective purchasing plan by the healthcare institutions. In order to search a future solution to this problem it was developed a study to verify the liability of representing (trought artificial neural networks) the cognitive process used by clinical engineering experts, during the evaluation phase of purchasing proposals for medical equipment. An inquiry (using electronic mail) to clinical engineers from several brazilian regions was conducted, using an electronic chart that contained a list of parameters commonly used for this evaluation phase. Data from the filled charts were used to train, and to test, diverse types of artificial neural networks. The best results (major correlation and minor quadratic errors with respect to the original entries) were encountered for an ensemble of 100 two-hidden-layers perceptrons trained with the backpropagation algorithm. It was then showed that the knowlegde of clinical engineers (for the evaluation process of purchasing proposals) can be represented by a non-linear connectionist model, whose entries would be the phisical risk, cost and strategic importance of the medical equipment. The model's outputs are the importance given by clinical engineers for five factors (clinical, financial, quality, safety and technical) for the evaluation of a medical equipment / Doutorado / Engenharia Biomedica / Doutor em Engenharia Elétrica

In this thesis we present a theoretical investigation of the feasibility of using a problem specific inductive bias for back-propagated neural networks. We argue that if a learning algorithm is biased towards optimizing a certain performance measure, it is plausible to assume that it will generate a higher performance score when evaluated using that particular measure. We use the term measure function for a multi-criteria evaluation function that can also be used as an inherent function in learning algorithms, in order to customize the bias of a learning algorithm for a specific problem. Hence, the term measure-based learning algorithms. We discuss different characteristics of the most commonly used performance measures and establish similarities among them. The characteristics of individual measures and the established similarities are then correlated to the characteristics of the backpropagation algorithm, in order to explore the applicability of introducing a measure function to backpropagated neural networks. Our study shows that there are certain characteristics of the error back-propagation mechanism and the inherent gradient search method that limit the set of measures that can be used for the measure function. Also, we highlight the significance of taking the representational bias of the neural network into account when developing methods for measure-based learning. The overall analysis of the research shows that measure-based learning is a promising area of research with potential for further exploration. We suggest directions for future research that might help realize measure-based neural networks. / The study is an investigation on the feasibility of using a generic inductive bias for backpropagation artificial neural networks, which could incorporate any one or a combination of problem specific performance metrics to be optimized. We have identified several limitations of both the standard error backpropagation mechanism as well the inherent gradient search approach. These limitations suggest exploration of methods other than backpropagation, as well use of global search methods instead of gradient search. Also, we emphasize the importance of taking the representational bias of the neural network in consideration, since only a combination of both procedural and representational bias can provide highly optimal solutions.

This dissertation evaluates evolutionary methods for evolving cooperative teams of robots. Cooperative robotics is a challenging research area in the field of artificial intelligence. Individual and autonomous robots may by cooperation enhance their performance compared to what they can achieve separately. The challenge of cooperative robotics is that performance relies on interactions between robots. The interactions are not always fully understood, which makes the designing process of hardware and software systems complex. Robotic soccer, such as the RoboCup competitions, offers an unpredictable dynamical environment for competing robot teams that encourages research of these complexities. Instead of trying to solve these problems by designing and implement the behavior, the robots can learn how to behave by evolutionary methods. For this reason, this dissertation evaluates evolution of neural controllers for a team of two robots in a competitive soccer environment. The idea is that evolutionary methods may be a solution to the complexities of creating cooperative robots. The methods used in the experiments are influenced by research of evolutionary algorithms with single autonomous robots and on robotic soccer. The results show that robot teams can evolve to a form of cooperative behavior with simple reactive behavior by relying on self-adaptation with little supervision and human interference.

The emerging need for dynamically scheduled real-time systems requires methods for handling transient overloads. Current methods have in common that they deal with transient overloads as they occur, which gives the real-time system limited time to react to the overload. In this work we enable new approaches to overload management. Our work shows that artificial neural networks (ANNs) can predict future transient overloads. This way the real-time system can prepare for a transient overload before it actually occurs. Even though the artificial neural network is not yet integrated into any system, the results show that ANNs are able to satisfactory distinguish different workload scenarios into those that cause future overloads from those that do not. Two ANN architectures have been evaluated, one standard feed-forward ANN and one recurrent ANN. These ANNs were trained and tested on sporadic workloads with different average arrival rates. At best the ANNs are able to predict up to 85% of the transient overloads in the test workload, while causing around 10% false alarms.

Admission controllers in dynamic real-time systems perform traditional schedulability tests in order to determine whether incoming tasks will meet their deadlines. These tests are computationally expensive and typically run in n * log n time where n is the number of tasks in the system. An incoming task might therefore miss its deadline while the schedulability test is being performed, when there is a heavy load on the system. In our work we evaluate a new approach for admission control in firm real-time systems. Our work shows that ANNs can be used to perform a schedulability test in order to work as an admission controller in firm real-time systems. By integrating the ANN admission controller to a real-time simulator we show that our approach provides feasible performance compared to a traditional approach. The ANNs are able to make up to 86% correct admission decisions in our simulations and the computational cost of our ANN schedulability test has a constant value independent of the load of the system. Our results also show that the computational cost of a traditional approach increases as a function of n log n where n is the number of tasks in the system.

Developing artificial agents able to autonomously discover new goals, to select them and learn the related skills is an important challenge for robotics. This becomes even crucial if we want robots to interact with real environments where they have to face many unpredictable problems and where it is not clear which skills will be the more suitable to solve them. The ability to learn and store multiple skills in order to use them when required is one of the main characteristics of biological agents: forming ample repertoires of actions is important to widen the possibility for an agent to better adapt to different environments and to improve its chance of survival and reproduction. Moreover, humans and other mammals explore the environment and learn new skills not only on the basis of reward-related stimuli but also on the basis of novel or unexpected neutral stimuli. The mechanisms related to this kind of learning processes have been studied under the heading of “Intrinsic Motivations” (IMs), and in the last decades the concept of IMs have been used in developmental and autonomous robotics to foster an artificial curiosity that can improve the autonomy and versatility of artificial agents. In the research presented in this thesis I focus on the development of open-ended learning robots able to autonomously discover interesting events in the environment and autonomously learn the skills necessary to reproduce those events. In particular, this research focuses on the role that IMs can play in fostering those processes and in improving the autonomy and versatility of artificial agents. Taking inspiration from recent and past research in this field, I tackle some of the interesting open challenges related to IMs and to the implementation of intrinsically motivated robots. I first focus on the neurophysiology underlying IM learning signals, and in particular on the relations between IMs and phasic dopamine (DA). With the support of a first computational model, I propose a new hypothesis that addresses the dispute over the nature and the functions of phasic DA activations: reconciling two contrasting theories in the literature and taking xi into account the different experimental data, I suggest that phasic DA can be considered as a reinforcement prediction error learning signal determined by both unexpected changes in the environment (temporary, intrinsic reinforcements) and biological rewards (permanent, extrinsic reinforcements). The results obtained with my computational model support the presented hypothesis, showing how such a learning signal can serve two important functions: driving both the discovery and acquisition of novel actions and the maximisation of rewards. Moreover, those results provide a first example of the power of IMs to guide artificial agents in the cumulative learning of complex behaviours that would not be learnt simply providing a direct reward for the final tasks. In a second work, I move to investigate the issues related to the implementation of IMs signal in robots. Since the literature still lacks a specific analysis of which is the best IM signal to drive skill acquisition, I compare in a robotic setup different typologies of IMs, as well as the different mechanisms used to implement them. The results provide two important contributions: 1) they show how IM signals based on the competence of the system are able to generate a better guidance for skill acquisition with respect to the signals based on the knowledge of the agent; 2) they identify a proper mechanism to generate a competence-based IM signal, showing that the stronger the link between the IM signal and the competence of the system, the better the performance. Following the aim of widening the autonomy and the versatility of artificial agents, in a third work I focus on the improvement of the control architecture of the robot. I build a new 3-level architecture that allows the system to select the goals to pursue, to search for the best way to achieve them, and acquire the related skills. I implement this architecture in a simulated iCub robot and test it in a 3D experimental scenario where the agent has to learn, on the basis of IMs, a reaching task where it is not clear which arm of the robot is the most suitable to reach the different targets. The performance of the system is compared to the one of my previous 2-level architecture system, where tasks and computational resources are associated at design time. The better performance of the system endowed with the new 3-level architecture highlights the importance of developing robots with different levels of autonomy, and in particular both the high-level of goal selection and the low-level of motor control. Finally, I focus on a crucial issue for autonomous robotics: the development of a system that is able not only to select its own goals, but also to discover them through the interaction with the environment. In the last work I present GRAIL, a Goal-discovering Robotic Architecture for Intrisically-motivated Learning. Building on the insights provided by my previous research, GRAIL is a 4-level hierarchical architecture that for the first time assembles in unique system different features necessary for the development of truly autonomous robots. GRAIL is able to autonomously 1) discover new goals, 2) create and store representations of the events associated to those goals, 3) select the goal to pursue, 4) select the computational resources to learn to achieve the desired goal, and 5) self-generate its own learning signals on the basis of the achievement of the selected goals. I implement GRAIL in a simulated iCub and test it in three different 3D experimental setup, comparing its performance to my previous systems, showing its capacity to generate new goals in unknown scenarios, and testing its ability to cope with stochastic environments. The experiments highlight on the one hand the importance of an appropriate hierarchical architecture for supporting the development of autonomous robots, and on the other hand how IMs (together with goals) can play a crucial role in the autonomous learning of multiple skills.