Vibration-based fault diagnosis (FD) with a simple spectrum can be complex, especially when considering FD of rotating machinery with multiple bearings like a multi-stage turbine. Various studies have sought to better interpret fault spectra, but the process remains equivocal. Consequently, it has been accepted that the simple spectra requires support from additional techniques, such as orbit analysis. But even orbit analysis can be inconclusive. Though promising, attempts at developing viable methods that rival the failure coverage of spectrum analysis without gaining computational complexity remain protracted. Interestingly, few researchers have developed FD methods for transient machine operation, however, these have proven to be involved. Current practices limit vibration data to a single machine, which usually requires a large unique data history. However, if sharing of data between similar machines with different foundations was possible, the need for unique histories would be mitigated. From readily available works, this has not been encountered. Therefore, a simple but robust vibration-based approach is warranted. In light of this, a novel on-bearing vibration response integration approach for condition monitoring of shaft-related faults irrespective of speed and foundation type is proposed in the present study. Vibration data are acquired at different speeds for: a baseline, unbalance, bow, crack, looseness, misalignment, and rub conditions on three laboratory rigs with dynamically different foundations, namely: rigid, flexible support 1 (FS1) and flexible support 2 (FS2). Testing is done on the rigid rig set up first, then FS1, and afterwards FS2. Common vibration features are computed from the measured data to be input to the proposed approach for further processing. First, the proposed approach is developed through its application to a machine at a steady speed in a novel Single-speed FD technique which exploits a single vibration sensor per bearing and fusion of features from different bearings for FD. Initially, vibration features are supplemented with bearing temperature readings with improved classification compared to vibration features alone. However, it is observed that temperature readings are insensitive to faults on the FS1 and FS2 rigs, when compared to vibration features, which are standardised for consistent classification on the different rigs tested. Thus, temperature is not included as a final feature. The observed fault classifications on the different rigs at different speeds with the standardised vibration features are encouraging. Thereafter, a novel Unified Multi-speed FD technique that is based on the initial proposed approach and which works by fusion of vibration features from different bearings at different speeds in a single analysis step for FD is proposed. Experiments on the different rigs repeatedly show the novel Multi-speed technique to be suitable for transient machine operation. Then, a novel generic Multi-foundation Technique (also based on the proposed approach) that allows sharing of vibration data of a wide range of fault conditions between two similarly configured machines with similar speed operation but different foundations is implemented to further mitigate data requirements in the FD process. Observations made with the rigs during steady and transient speed tests show this technique is applicable in situations where data history is available on one machine but lacking on the other. Comparison of experimental results with results obtained from theoretical simulations indicates the approach is consistent. Thus, the proposed approach has the potential for practical considerations.

Actuellement, l'inconfort intérieur dans les bâtiments est l'une des questions cruciales, ainsi que la consommation énergétique du bâtiment. En effet, les gens passent 60 à 90% de leur vie dans les bâtiments. Le confort intérieur est indispensable en termes de bienfaits sur la santé, la productivité et le bien-être des occupants. C'est pourquoi diverses stratégies d'optimisation fondées sur des règles, anticipatives ou prédictives ont été proposées pour atteindre le confort perçu en tenant compte de la consommation d'énergie. Dans la pratique, il existe un écart entre l'anticipation et la réalité. Habituellement, les plans anticipatifs sont synchronisés avec une période d'anticipation d'une heure et ne tiennent pas compte des différentes sources de conflits ainsi que des configurations d'enveloppes des bâtiments possibles. Il peut en résulter des conséquences négatives sur le coût et le confort. Pour résoudre ce problème, le système de gestion du bâtiment doit être conçu aussi comme réactif, de sorte à ce qu'il puisse répondre à toutes sortes de divergences par rapport au plan anticipatif, de manière réactive. Dans ce but, un système multi-échelle d'analyse de diagnostic réactif anticipatif (ARD- BMS) est proposé dans cette thèse. ARD-BMS est une gestion interne qui effectue trois actions importantes, c'est-à-dire la détection de défaut, l'isolement des causes et, enfin, les actions correctives. ARD-BMS permet ainsi une gestion réactive à court terme, à savoir 10 minutes pour analyser les tendances des défauts et la mise à jour de la dynamique du bâtiment et prendre ainsi les mesures correctives nécessaires pour maintenir le niveau de confort désiré. Cette thèse propose donc un modèle réactif à dynamique rapide simplifiée qui peut être utilisé pour estimer l'état actuel du bâtiment. Les bâtiments modernes sont un système très sophistiqué avec un grand nombre de capteurs, de contrôleurs et de CVC. La plupart des installations de construction utilisent des services prévus de maintenance préventive provenant des opérations périodiques des bâtiments. Ces problèmes imprévus puce causer des répercussions inexplicables sur le confort de l'occupant pendant le cycle de fonctionnement de 24 heures. Ces problèmes ne sont pas inadmissibles tels que les situations imprévues, les pannes de prévisions météorologiques. Le diagnostic des causes d'inconfort à court terme est encore un problème difficile au niveau de l'opération de construction intégrale. En outre, pour analyser cette situation, proposez une méthodologie diagnostique pour la détection et l'isolement des causes (fautes) dans les bâtiments. La méthodologie proposée comprend une HAZOP fondée sur les règles (analyse des risques et de l'optimisation) et une approche basée sur un modèle. / Currently, indoor discomfort in dwellings is one of the crucial issues along with the building energy consumption. Indeed, people spend 60-90% of their lives in buildings. Indoor comfort plays a vital role in occupants health, productivity, and well-being. However, various optimization and rule-based anticipative or predictive building strategies have been proposed to achieve the perceived comfort taking into account the energy consumption. However, in practice, anticipation or plans are far from the reality. Usually, anticipative plans are synchronized with one-hour anticipation period and do not consider the various sources of discrepancies as well as current envelope configurations. Unbeknownst to many, discrepancies from different sources could cause big penalty over cost and comfort. To tackle this issue, building management system needs to be designed as reactive or almost with no planning so that it can respond to all discrepancies re-actively. To address this problem, a multi-scale Anticipative Reactive Diagnosing- Building Management System (ARD-BMS) is proposed in this dissertation. ARD-BMS is an internal management and performs three important actions i.e., Discrepancydetection, Cause isolation, and finally Corrective actions. ARD-BMS follow the short-time resolution i.e., 10-minute to analyze the fault trends and current the building dynamics and take necessary corrective actions to maintain the desired level of comfort. This thesis proposes a fast dynamics simplified reactive model that can be used to estimate the current status of the building. Modern buildings are very sophisticated system with a large number of sensors, controllers, and HVACs. Most of the building facilities are using a scheduled preventive maintenance services derived from periodic operations of the buildings. These preventive actions do not take into account the other inadmissible issues such as unplanned situations, weather prediction failures etc. These unplanned issues could cause unaccountable impacts over occupant’s comfort during the 24-hour operation cycle. Diagnosability of short-term discomfort causes is still a challenging job at whole building operation level. Furthermore, to analyze this situation the thesis proposes a diagnostic methodology for detection and isolation of cause (faults) in buildings. The proposed methodology includes a rule-based HAZOP (Hazard and Operability analysis) and model-based approach. Further, in order to oversee unplanned discomforts, a short-term reactive optimization has been proposed.

Field-Programmable Gate Arrays (FPGAs) são largamente utilizadas em sistemas digitais por características como flexibilidade, baixo custo e alta densidade. Estas características advém do uso de células de SRAM na memória de configuração, o que torna estes dispositivos suscetíveis a erros induzidos por radiação, tais como SEUs. TMR é o método de mitigação mais utilizado, no entanto, possui um elevado custo tanto em área como em energia, restringindo seu uso em aplicações de baixo custo e/ou baixo consumo. Como alternativa a TMR, propõe-se utilizar DMR associado a um mecanismo de reparo da memória de configuração da FPGA chamado scrubbing. O reparo de FPGAs em sistemas em tempo real apresenta desafios específicos. Além da garantia da computação correta dos dados, esta computação deve se dar completamente dentro do tempo disponível (time-slot), devendo ser finalizada antes do tempo limite (deadline). A diferença entre o tempo de computação dos dados e a deadline é chamado de slack e é o tempo disponível para reparo do sistema. Este trabalho faz uso de scrubbing deslocado dinâmico, que busca maximizar a probabilidade de reparo da memória de configuração de FPGAs dentro do slack disponível, baseado em um diagnóstico do erro. O scrubbing deslocado já foi utilizado com técnicas de diagnóstico de grão fino (NAZAR, 2015). Este trabalho propõe o uso de técnicas de diagnóstico de grão grosso para o scrubbing deslocado, evitando as penalidades de desempenho e custos em área associados a técnicas de grão fino. Circuitos do conjunto MCNC foram protegidos com as técnicas propostas e submetidos a seções de injeção de erros (NAZAR; CARRO, 2012a). Os dados obtidos foram analisados e foram calculadas as melhores posição iniciais do scrubbing para cada um dos circuitos. Calculou-se a taxa de Failure-in-Time (FIT) para comparação entre as diferentes técnicas de diagnóstico propostas. Os resultados obtidos confirmaram a hipótese inicial deste trabalho que a redução do número de bits sensíveis e uma baixa degradação do período do ciclo de relógio permitiram reduzir a taxa de FIT quando comparadas com técnicas de grão fino. Por fim, uma comparação entre as três técnicas propostas é feita, analisando o desempenho e custos em área associados a cada uma. / Field-Programmable Gate Arrays (FPGAs) are widely used in digital systems due to characteristics such as flexibility, low cost and high density. These characteristics are due to the use of SRAM memory cells in the configuration memory, which make these devices susceptible to radiation-induced errors, such as SEUs. TMR is the most used mitigation technique, but it has an elevated cost both in area as well as in energy, restricting its use in low cost/low energy applications. As an alternative to TMR, we propose the use of DMR associated with a repair mechanism of the FPGA configuration memory called scrubbing. The repair of FPGA in real-time systems present a specific set of challenges. Besides guaranteeing the correct computation of data, this computation must be completely carried out within the available time (time-slot), being finalized before a time limit (deadline). The difference between the computation time and the deadline is called the slack and is the time available to repair the system. This work uses a dynamic shifted scrubbing that aims to maximize the repair probability of the configuration memory of the FPGA within the available slack based on error diagnostic. The shifted scrubbing was already proposed with fine-grained diagnostic techniques (NAZAR, 2015). This work proposes the use of coarse-grained diagnostic technique as a way to avoid the performance penalties and area costs associated to fine-grained techniques. Circuits of the MCNC suite were protected by the proposed techniques and subject to error-injection campaigns (NAZAR; CARRO, 2012a). The obtained data was analyzed and the best scrubbing starting positions for each circuit were calculated. The Failure-in-Time (FIT) rates were calculated to compare the different proposed diagnostic techniques. The obtained results validated the initial hypothesis of this work that the reduction of the number of sensitive bits and a low degradation of the clock cycle allowed a reduced FIT rate when compared with fine-grained diagnostic techniques. Finally, a comparison is made between the proposed techniques, considering performance and area costs associated to each one.

It is shown that the path delay fault (PDF) model may not be very effective in guiding post silicon debug. It is also shown that the multiple transition fault (MTF) model allows for significant reduction of the initial suspect set. However the number of faults is much higher than the number of PDFs. A Monte Carlo approach is presented that uses multiple transition faults with appropriately assigned weights to identify defective embedded segments. It is experimentally verified that the approach guides diagnosis more efficiently than the path delay fault model. Fault-implicit algorithms are presented to cope with fault-related scalability challenges. Our results in ISCAS '89, ISCAS'85, ITC '99 benchmarks show a huge reduction in the suspect set using the proposed fault model and algorithms. It is shown that the proposed method guides effectively fault diagnosis.

O escopo desta dissertação é o desenvolvimento de uma metodologia para a detecção e diagnóstico de falhas em processos industriais baseado em modelos empíricos no subespaço das variáveis do processo com expansão não linear das bases. A detecção e o diagnóstico de falhas são fundamentais para aumentar a segurança, confiabilidade e lucratividade de processos industriais. Métodos qualitativos, quantitativos e baseados em dados históricos do processo têm sido estudados amplamente. Para demonstrar as vantagens da metodologia proposta, ela será comparada com duas metodologias consideradas padrão, uma baseada em Análise de Componentes Principais (PCA) e a outra baseada em Mínimos Quadrados Parciais (PLS). Dois estudos de casos são empregados nessa comparação. O primeiro consiste em um tanque de aquecimento com mistura e o segundo contempla o estudo de caso do processo da Tennessee Eastman. As vantagens da metodologia proposta consistem na redução da dimensionalidade dos dados a serem usados para um diagnóstico adequado, além de detectar efetivamente a anormalidade e identificar as variáveis mais relacionadas à falha, permitindo um melhor diagnóstico. Além disso, devido à expansão das bases dos modelos é possível trabalhar efetivamente com sistemas não lineares, através de funções polinomiais e exponenciais dentro do modelo. Adicionalmente o trabalho contém uma metodologia de validação dos resultados da metodologia proposta, que consiste na eliminação das variáveis do melhor modelo obtido pelos Modelos Empíricos, através do método Backward Elimination. A metodologia proposta forneceu bons resultados na área do diagnóstico de falhas: conseguiu-se uma grande diminuição da dimensionalidade nos sistemas estudados em até 93,55%, bem como uma correta detecção de anormalidades e permitiu a determinação das variáveis mais relacionadas às anormalidades do processo. As comparações feitas com as metodologias padrões permitiram demonstrar que a metodologia proposta tem resultados superiores, pois consegue detectar as anormalidades em um espaço dimensional reduzido, detectando comportamentos não lineares e diminuindo incertezas. / Fault detection and diagnosis are critical to increasing the safety, reliability, and profitability of industrial processes. Qualitative and quantitative methods and process historical data have been extensively studied. This article proposes a methodology for fault detection and diagnosis, based on historical data of processes and the creation of empirical models with the expansion of nonlinear bases (polynomial and exponential bases) and regularization techniques. To demonstrate the advantages of the proposed approach, it is compared with two standard methodologies: Principal Components Analysis (PCA) and the Partial Least Squares (PLS), performed in two case studies: a mixed heating tank and the Tennessee Eastman Process. The advantages of the proposed methodology are the reduction of the dimensionality of the data used, in addition to the effective detection of abnormalities, identifying the variables most related to the fault. Furthermore, the work contains a methodology to validate the diagnosis results consisting of variable elimination from the best empirical models with the Backward Elimination algorithm. The proposed methodology achieved a promising performance, since it can decrease the dimensionality of the studied systems up to 93.55%, reducing uncertainties, and capturing nonlinear behaviors.

Orientador: Jose Roberto Sanches Mantovani / Banca: Rubén Augusto Romero Lázaro / Banca: Carlos Roberto Minussi / Banca: Oriane Magela Neto / Banca: Julio Cesar Stacchini de Souza / Resumo: Neste trabalho é proposta uma metodologia para a análise e interpretação de alarmes em tempo real em sistemas de distribuição de energia elétrica, considerando o diagnóstico em nível de subestações e redes. A metodologia busca superar as dificuldades e desvantagens dos métodos já propostos na literatura especializada para resolver o diagnóstico de faltas em sistemas de potência. O método proposto emprega um modelo matemático original bem como um novo algoritmo genético para efetuar o diagnóstico dos alarmes de maneira eficiente e rápida. O modelo matemático é dividido em duas partes fundamentais: (1) modelo de operação do sistema de proteção; e (2) modelo de Programação Binária Irrestrita (PBI). A parte (1) é composta por um conjunto de equações de estados esperados das funções de proteção dos relés do sistema, modeladas com base na lógica de operação de funções de proteção tais como sobrecorrente, diferencial e distância, bem como na filosofia de proteção de sistemas de potência. A parte (2) é estabelecida através de uma função objetivo formulada com base na teoria de cobertura parcimoniosa (parcimonious set covering theory), e busca a associação ou "match" entre os relatórios de alarmes informados pelo sistema SCADA (Supervisory Control and Data Acquisition) e os estados esperados das funções de proteção formuladas na parte (1) do modelo. O novo algoritmo genético proposto é empregado para minimizar o modelo de PBI e possui como característica a utilização de dois parâmetros de controle. O algoritmo possui taxas de recombinação e mutação automática e dinamicamente calibradas, baseadas na saturação da população corrente, possuindo uma imediata resposta à possível convergência prematura para ótimos locais. A metodologia desenvolvida para o diagnóstico... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: This work proposes a methodology for the analysis and interpretation of real-time alarms in electric power distribution systems in the substation level and network level. The methodology seeks to overcome the difficulties and disadvantages of the methods already proposed in the literature to solve the fault diagnosis in power systems. The proposed method employs a novel mathematical model and a genetic algorithm to carry out the diagnosis of alarms efficiently and quickly. The model is divided into two main parts: (1) a protection system operation model; and (2) Unconstrained Binary Programming (UBP) model. Part (1) provides a set of expected state equations of the protective relay functions established based on the protection operation logic such as overcurrent, differential and distance as well as the protection philosophy. Part (2) is established through an objective function formulated based on parsimonious set covering theory for associating the alarms reported by SCADA (Supervisory Control and Data Acquisition) system with the expected states of the protective relay functions. The novel genetic algorithm use only two control parameters and is employed to minimize the UBP model. In addition the algorithm has recombination and mutation rates automatically and dynamically calibrated based on the saturation of the current population and it presents an immediate response to possible premature convergence to local optima. The methodology developed for the diagnosis of substations is extended to distribution networks considering that the network has sufficient level of automation for remote monitoring of the primary feeders. In this way a new paradigm for protection of distribution networks developed based on Smart Grid concept is proposed. Extensive tests are performed with the methodology applied to distribution... (Complete abstract click electronic access below) / Doutor

O escopo desta dissertação é o desenvolvimento de uma metodologia para a detecção e diagnóstico de falhas em processos industriais baseado em modelos empíricos no subespaço das variáveis do processo com expansão não linear das bases. A detecção e o diagnóstico de falhas são fundamentais para aumentar a segurança, confiabilidade e lucratividade de processos industriais. Métodos qualitativos, quantitativos e baseados em dados históricos do processo têm sido estudados amplamente. Para demonstrar as vantagens da metodologia proposta, ela será comparada com duas metodologias consideradas padrão, uma baseada em Análise de Componentes Principais (PCA) e a outra baseada em Mínimos Quadrados Parciais (PLS). Dois estudos de casos são empregados nessa comparação. O primeiro consiste em um tanque de aquecimento com mistura e o segundo contempla o estudo de caso do processo da Tennessee Eastman. As vantagens da metodologia proposta consistem na redução da dimensionalidade dos dados a serem usados para um diagnóstico adequado, além de detectar efetivamente a anormalidade e identificar as variáveis mais relacionadas à falha, permitindo um melhor diagnóstico. Além disso, devido à expansão das bases dos modelos é possível trabalhar efetivamente com sistemas não lineares, através de funções polinomiais e exponenciais dentro do modelo. Adicionalmente o trabalho contém uma metodologia de validação dos resultados da metodologia proposta, que consiste na eliminação das variáveis do melhor modelo obtido pelos Modelos Empíricos, através do método Backward Elimination. A metodologia proposta forneceu bons resultados na área do diagnóstico de falhas: conseguiu-se uma grande diminuição da dimensionalidade nos sistemas estudados em até 93,55%, bem como uma correta detecção de anormalidades e permitiu a determinação das variáveis mais relacionadas às anormalidades do processo. As comparações feitas com as metodologias padrões permitiram demonstrar que a metodologia proposta tem resultados superiores, pois consegue detectar as anormalidades em um espaço dimensional reduzido, detectando comportamentos não lineares e diminuindo incertezas. / Fault detection and diagnosis are critical to increasing the safety, reliability, and profitability of industrial processes. Qualitative and quantitative methods and process historical data have been extensively studied. This article proposes a methodology for fault detection and diagnosis, based on historical data of processes and the creation of empirical models with the expansion of nonlinear bases (polynomial and exponential bases) and regularization techniques. To demonstrate the advantages of the proposed approach, it is compared with two standard methodologies: Principal Components Analysis (PCA) and the Partial Least Squares (PLS), performed in two case studies: a mixed heating tank and the Tennessee Eastman Process. The advantages of the proposed methodology are the reduction of the dimensionality of the data used, in addition to the effective detection of abnormalities, identifying the variables most related to the fault. Furthermore, the work contains a methodology to validate the diagnosis results consisting of variable elimination from the best empirical models with the Backward Elimination algorithm. The proposed methodology achieved a promising performance, since it can decrease the dimensionality of the studied systems up to 93.55%, reducing uncertainties, and capturing nonlinear behaviors.

Field-Programmable Gate Arrays (FPGAs) são largamente utilizadas em sistemas digitais por características como flexibilidade, baixo custo e alta densidade. Estas características advém do uso de células de SRAM na memória de configuração, o que torna estes dispositivos suscetíveis a erros induzidos por radiação, tais como SEUs. TMR é o método de mitigação mais utilizado, no entanto, possui um elevado custo tanto em área como em energia, restringindo seu uso em aplicações de baixo custo e/ou baixo consumo. Como alternativa a TMR, propõe-se utilizar DMR associado a um mecanismo de reparo da memória de configuração da FPGA chamado scrubbing. O reparo de FPGAs em sistemas em tempo real apresenta desafios específicos. Além da garantia da computação correta dos dados, esta computação deve se dar completamente dentro do tempo disponível (time-slot), devendo ser finalizada antes do tempo limite (deadline). A diferença entre o tempo de computação dos dados e a deadline é chamado de slack e é o tempo disponível para reparo do sistema. Este trabalho faz uso de scrubbing deslocado dinâmico, que busca maximizar a probabilidade de reparo da memória de configuração de FPGAs dentro do slack disponível, baseado em um diagnóstico do erro. O scrubbing deslocado já foi utilizado com técnicas de diagnóstico de grão fino (NAZAR, 2015). Este trabalho propõe o uso de técnicas de diagnóstico de grão grosso para o scrubbing deslocado, evitando as penalidades de desempenho e custos em área associados a técnicas de grão fino. Circuitos do conjunto MCNC foram protegidos com as técnicas propostas e submetidos a seções de injeção de erros (NAZAR; CARRO, 2012a). Os dados obtidos foram analisados e foram calculadas as melhores posição iniciais do scrubbing para cada um dos circuitos. Calculou-se a taxa de Failure-in-Time (FIT) para comparação entre as diferentes técnicas de diagnóstico propostas. Os resultados obtidos confirmaram a hipótese inicial deste trabalho que a redução do número de bits sensíveis e uma baixa degradação do período do ciclo de relógio permitiram reduzir a taxa de FIT quando comparadas com técnicas de grão fino. Por fim, uma comparação entre as três técnicas propostas é feita, analisando o desempenho e custos em área associados a cada uma. / Field-Programmable Gate Arrays (FPGAs) are widely used in digital systems due to characteristics such as flexibility, low cost and high density. These characteristics are due to the use of SRAM memory cells in the configuration memory, which make these devices susceptible to radiation-induced errors, such as SEUs. TMR is the most used mitigation technique, but it has an elevated cost both in area as well as in energy, restricting its use in low cost/low energy applications. As an alternative to TMR, we propose the use of DMR associated with a repair mechanism of the FPGA configuration memory called scrubbing. The repair of FPGA in real-time systems present a specific set of challenges. Besides guaranteeing the correct computation of data, this computation must be completely carried out within the available time (time-slot), being finalized before a time limit (deadline). The difference between the computation time and the deadline is called the slack and is the time available to repair the system. This work uses a dynamic shifted scrubbing that aims to maximize the repair probability of the configuration memory of the FPGA within the available slack based on error diagnostic. The shifted scrubbing was already proposed with fine-grained diagnostic techniques (NAZAR, 2015). This work proposes the use of coarse-grained diagnostic technique as a way to avoid the performance penalties and area costs associated to fine-grained techniques. Circuits of the MCNC suite were protected by the proposed techniques and subject to error-injection campaigns (NAZAR; CARRO, 2012a). The obtained data was analyzed and the best scrubbing starting positions for each circuit were calculated. The Failure-in-Time (FIT) rates were calculated to compare the different proposed diagnostic techniques. The obtained results validated the initial hypothesis of this work that the reduction of the number of sensitive bits and a low degradation of the clock cycle allowed a reduced FIT rate when compared with fine-grained diagnostic techniques. Finally, a comparison is made between the proposed techniques, considering performance and area costs associated to each one.

Este trabalho propÃe um novo mÃtodo de diagnÃstico de falhas incipientes em equipamentos de potÃncia com isolamento a Ãleo baseado em regras interpretÃveis extraÃdas a partir das concentraÃÃes dos gases presentes e dissolvidos no Ãleo isolante e de relaÃÃes entre essas concentraÃÃes nÃo exploradas por outros mÃtodos. Utilizando uma Ãrvore de decisÃo (Decision Tree - DT) com modificaÃÃes na dimensÃo dos atributos de entrada para a obtenÃÃo de regras, o mÃtodo proposto: (i) permite que o usuÃrio conheÃa as regras que geraram o diagnÃstico, (ii) contorna limitaÃÃes comuns a outros mÃtodos (ex.: existÃncia de regiÃes de ânÃo-diagnÃsticoâ, existÃncia de mais de um diagnÃstico possÃvel) e (iii) apresenta percentual elevado de acerto no diagnÃstico de dados de DGA (Dissolved Gas Analysis) disponÃveis na literatura. O mÃtodo proposto à comparado com outros mÃtodos de DGA descritos em normas: GÃs Chave, Rogers, DÃrnenburg e TriÃngulo de Duval. Na comparaÃÃo dos mÃtodos foram utilizados bancos de dados com informaÃÃes publicadas, entre eles os dados do IEC TC 10 que apresentam informaÃÃes e diagnÃstico de um comità de especialistas. Os Ãndices de acerto observados com o mÃtodo proposto estiveram acima de 94%. Comparativamente, o MÃtodo do TriÃngulo de Duval apresenta Ãndices de acerto entre 83% e 88%. O mÃtodo tambÃm foi avaliado com relaÃÃo à sensibilidade na ocorrÃncia de erros nas medidas de laboratÃrio, apresentando resultados comparativamente superiores (Ãndice de acerto de 83% em comparaÃÃo com 78% do MÃtodo do TriÃngulo de Duval). TambÃm foi desenvolvido neste trabalho um mÃtodo para o diagnÃstico simplificado de equipamentos de potÃncia com isolamento a Ãleo utilizando dados fÃsico-quÃmicos â os Ãndices de acerto obtidos foram inferiores aos alcanÃados com o DGA proposto, porÃm prÃximos aos observados em mÃtodos de DGA citados na literatura, como Rogers IEEE C57.104 â 2008, com a vantagem dos dados nÃo dependerem da disponibilidade de um cromatÃgrafo. / This work proposes a new method for diagnosis of incipient faults in power equipment with insulation oil. It consists on a set of interpretable rules extracted from the concentrations of the dissolved gases in insulating oil, and the relationships between these concentrations. Using a Decision Tree (DT) with changes in the size of the input attributes for obtaining rules, the proposed method: (i) allows the user to know the rules that generated the diagnosis, (ii) bypasses limitations common to other methods (e.g. the existence of "no diagnosis" regions, and the occurrence of more than one possible diagnosis) and (iii) the accuracy in diagnosis of DGA (Dissolved Gas Analysis) data available in the literature. The proposed method is compared with other DGA standard methods: Gas Key, Rogers, Dornenburg and DuvalÂs Triangle. The performed comparison considered databases available in technical literature, including data from the IEC TC 10. The observed accuracy of the proposed method is above 94%. Comparatively, the DuvalÂs Triangle method presents accuracy between 83% and 88% for the considered databases. The method was also evaluated with respect to sensitivity to measurements errors of laboratory â the results are very promising, the proposed method shows a hit rate of 83% in comparison with a hit rate of 78% for the DuvalÂs Triangle method. This work has also been developed a method for simplified diagnostic of power equipments with insulated oil using physico-chemical data - the hit rates obtained were lower than those achieved with the proposed DGA, however close to those observed in DGA methods cited in the literature, as Rogers IEEE C57.104 - 2008, with the advantage of the data not depend on the availability of a gas chromatograph.

This thesis investigates the possibility of improving black box fault diagnosis by a process called manifold learning, which simply stated is a way of finding patterns in recorded sensor data. The idea is that there is more information in the data than is exploited when using simple classification algorithms such as k-Nearest Neighbor and Support Vector Machines, and that this additional information can be found by using manifold learning methods. To test the idea of using manifold learning, data from two different fault diagnosis scenarios is used: A Scania truck engine and an electrical system called Adapt. Two linear and one non-linear manifold learning methods are used: Principal Component Analysis and Linear Discriminant Analysis (linear) and Laplacian Eigenmaps (non-linear).Some improvements are achieved given certain conditions on the diagnosis scenarios. The improvements for different methods correspond to the systems in which they are achieved in terms of linearity. The positive results for the relatively linear electrical system are achieved mainly by the linear methods Principal Component Analysis and Linear Discriminant Analysis and the positive results for the non-linear Scania system are achieved by the non-linear method Laplacian Eigenmaps.The results for scenarios without these special conditions are not improved however, and it is uncertain wether the improvements in special condition scenarios are due to gained information or to the nature of the cases themselves.