Among the renewable energies, wind energy presents the highest growth in installed capacity and penetration in modern power systems. This is why reliability of wind turbines becomes an important topic in research and industry. To this end, condition monitoring (or health monitoring) systems are needed for wind turbines. The core of any condition monitoring system (CMS) are fault diagnosis algorithms whose task is to provide early warnings upon the occurrence of incipient (small magnitude) faults. Thanks to the use of CMS we can avoid premature breakdowns and reduce significatively maintenance costs.<p><p>The present thesis deals with fault diagnosis in sensors of a doubly-fed induction generator (DFIG) for wind turbine (WT) applications. In particular we are interested in performing fault detection and isolation (FDI) of incipient faults affecting the measurements of the three-phase signals (currents and voltages) in a controlled DFIG. Although different authors have dealt with FDI for sensors in induction machines and in DFIGs, most of them rely on the machine model with<p>constant parameters. However, the parameter uncertainties due to changes in the operating conditions will produce degradation in the performance of such FDI systems.<p><p>In this work we propose a systematic methodology for the design of sensor FDI systems with the following characteristics: i) capable of detecting and isolating incipient additive (bias, drifts) and multiplicative (changes in the sensor<p>gain) faults, ii) robust against changes in the references/disturbances affecting the controlled DFIG as well as modelling/parametric uncertainties, iii) residual generation system based on a multi-observer strategy to enhance the isolation process, iv) decision system based on statistical-change detection algorithms to treat the entire residual and perform fault detection and isolation at once.<p><p>Three novel sensor FDI approaches are proposed. The first is a signal-based approach, that uses the model of the balanced three-phase signals (currents or voltages) for residual generation purposes. The second is a model-based approach<p>that accounts for variation in the parameters. Finally, a third approach that combines the benefits of both the signal- and the model-based approaches is proposed. The designed sensor FDI systems have been validated using measured voltages, as well as simulated data from a controlled DFIG and a speed-controlled induction<p>motor. <p><p>In addition, in this work we propose a discrete-time multiple input multiple output (MIMO) regulator for each power converter, namely for the rotor side converter (RSC) and for the grid side converter (GSC). In particular, for RSC<p>control, we propose a modified feedback linearization technique to obtain a linear time invariant (LTI) model dynamics for the compensated DFIG. The novelty of this approach is that the compensation does not depend on highly uncertain parameters such as the rotor resistance. For GSC control, a LTI model dynamics<p>is derived using the ideas behind feedback linearization. The obtained LTI model dynamics are used to design Linear Quadratic Gaussian (LQG) regulators. A single design is needed for all the possible operating conditions. / Doctorat en Sciences de l'ingénieur / info:eu-repo/semantics/nonPublished

This work proposes a hybrid system for alarm processing and fault diagnosis in electrical networks which use two methods of computational intelligence: Generalized Regression Neural Network and Genetic Algorithms. The neural network has the function of processing the set of received alarms and present as a response the characteristic(s) event(s), using for this, an elaborated knowledge based on the functional diagrams for protection and interviews with operators. Six modules were implemented for different neural components of a test system, according to their protection schemes. The output of these modules is used as input to the GA which has to do a combined analysis along with its database and provide the operator with the main protective components involved in the incident, as well as the probable causes of defects and actions to be taken in order to return the system in the shortest possible time and greater safety. For average inserted random errors of 0%, 7,73%, 15,46% and 23,19% in the received alarms, the system was able to diagnoses correctly in 100%, 93,60%, 74,26% and 48,07% of the cases respectively. It was found that the genetic algorithm improved the results obtained by neural network with good capability of generalization and condition to present multiple solutions, and the response time of the hybrid system was acceptable to the under consideration problem. / O presente trabalho propõe um sistema híbrido para processamento de alarmes e diagnóstico de faltas em redes elétricas com a utilização de dois métodos de inteligência computacional: Generalized Regression Neural Network e Algoritmos Genéticos. A rede neural tem a função de processar o conjunto de alarmes reportados e apresentar como resposta o evento(s) característico(s), utilizando-se, para isso, de um conhecimento elaborado com base nos diagramas funcionais da proteção e entrevista com operadores. Foram implementados seis módulos neurais para diferentes componentes de um sistema teste, de acordo com os seus respectivos esquemas de proteção. A saída destes módulos é utilizada como entrada para o AG que deve fazer uma análise combinatória juntamente com sua base de dados e apresentar ao operador os principais componentes de proteção envolvidos na incidência, bem como as prováveis causas do defeito e ações a serem tomadas de forma a restabelecer o sistema no menor tempo possível e com maior segurança. Para erros aleatórios médios inseridos de 0%, 7,73%, 15,46% e 23,19% nos alarmes reportados, o sistema se mostrou capaz de diagnosticar corretamente em respectivamente 100%, 93,60%, 74,26% e 48,07% dos casos. Verificou-se que o algoritmo genético melhorou os resultados obtidos pela rede neural, apresentando boa capacidade de generalização e condições de apresentação de múltiplas soluções, sendo o tempo de resposta do sistema híbrido aceitável para o problema tratado.

Automated start–up and shutdown procedures increase the profitability and safety of a process, but are difficult to implement due to the complex nature of the concepts that must be incorporated. Generic components used specifically for the implementation of automated startup and shutdown procedures were defined to streamline the implementation process. The generic components developed are based on Sequential Function Charts and were applied to the startup of a fixed–bed gasification unit, for which a dynamic simulation model was developed. The application showed that the automated startup can be defined by a few generic components and that the flexibility of the startup procedure is increased through the incorporation of a fault accommodation module. The use of a visual–based definition of sequential processes increases the understanding of the complex scheduling procedures as well as the efficiency of the development of these automated procedures. In addition, iterative learning was incorporated into the generic definition to optimise controller performance during the non–linear phases of operation. / Dissertation (MEng (Control Engineering))--University of Pretoria, 2006. / Chemical Engineering / MEng / Unrestricted

The induction motors are used worldwide in various industries. Several maintenance techniques are applied to increase the operating time and the lifespan of these motors. Among these, the predictive maintenance techniques such as Motor Current Signature Analysis (MCSA), Motor Square Current Signature Analysis (MSCSA), Park's Vector Approach (PVA) and Park's Vector Square Modulus (PVSM) are used to detect and diagnose faults in electric motors, characterized by patterns in the stator current frequency spectrum. In this work, these techniques are applied and compared on real motors, which have the faults of eccentricity in the air-gap, inter-turn short circuit and broken bars. It was used a theoretical model of an electric induction motor without fault and with the same voltage supply in order to assist comparison between the stator current frequency spectrum patterns with and without faults. Metrics were purposed and applied to evaluate the sensitivity of each technique fault detection. The results presented here show that the above techniques are suitable for the faults above mentioned. / Os motores elétricos de indução são utilizados em todo o mundo nos mais variados ramos industriais. Diversas técnicas de manutenção são aplicadas para aumentar o tempo de operação e a vida útil destes motores. No contexto da manutenção preditiva, técnicas como Motor Current Signature Analysis (MCSA), Motor Square Current Signature Analysis (MSCSA), Park’s Vector Approach (PVA) e Park’s Vector Square Modulus (PVSM) são utilizadas para detectar e diagnosticar falhas em motores elétricos, caracterizadas por padrões no espectro de frequência da corrente estatórica. Neste trabalho, estas técnicas são aplicadas e comparadas em motores reais, os quais apresentam as falhas de excentricidade no entreferro, curto circuito entre espiras e barras quebradas. Utilizou-se um modelo teórico de um motor elétrico de indução sem falhas, com a mesma tensão de suprimento, com o objetivo de auxiliar a comparação entre os padrões do espectro de frequência de corrente estatórica com e sem falhas. Foram propostas e aplicadas métricas que avaliam a sensibilidade de cada técnica na detecção da falha. Os resultados apresentados neste trabalho mostraram que as técnicas acima mencionadas foram adequadas para as falhas supracitadas, cuja comparação entre estas evidenciou a adequabilidade de cada uma.

Ce travail concerne l’étude d’un réacteur chimique continu afin de construire un modèle pour la phase d’apprentissage de méthode et localisation et détection de pannes. Un dispositif expérimental a été conçu pour disposer de données expérimentales significatives. Pour le diagnostique et la localisation des méthodes orientées données ont été retenues, principalement les réseaux Bayésiens et les réseaux de neurones à Fonctions Radiales de Base (RBF) couplés à un algorithme génétique auto adaptatif à ajustement local (GAAPA). Les données collectées à partir du dispositif expérimental ont servi à l’apprentissage et à la validation du modèle. / The aim is to study a continuous chemical process, and then analyze the hold process of the reactor and build the models which could be trained to realize the fault diagnosis and localization in the process. An experimental system has been built to be the research base. That includes experiment part and record system. To the diagnosis and localization methods, the work presented the methods with the data-based approach, mainly the Bayesian network and RBF network based on GAAPA (Genetic Algorithm with Auto-adapted of Partial Adjustment). The data collected from the experimental system are used to train and test the models.

Le procédé de distillation est largement utilisé dans de nombreuses applications telles que la production pétrochimique, le traitement du gaz naturel, les raffineries de pétrole, etc. Généralement, la maintenance des réacteurs chimiques est très coûteuse et perturbe la production pendant de longues périodes. Tous ces facteurs démontrent réellement la nécessité de stratégies efficaces de diagnostic et de pronostic des défauts pour pouvoir réduire et éviter le plus grand nombre de ces problèmes catastrophiques. La première partie de nos travaux vise à proposer une méthode de diagnostic fiable pouvant être utilisée dans le régime permanent d’une procédure non linéaire. De plus, nous proposons une procédure modifiée de la méthode MFCM permettant de calculer la variation en pourcentage entre deux classes. L’utilisation de MFCM a pour objectif de réduire le temps de calcul et d’accroître les performances du classifieur. Les résultats de la méthode proposée confirment la capacité de classifier entre les différentes classes de défaillances considérées. Le calcul de la durée de vie du système est extrêmement important pour éviter les pannes catastrophiques. Notre deuxième objectif est de proposer une méthode fiable de pronostic permettant d’estimer le chemin de dégradation d’une colonne de distillation et de calculer le pourcentage de durée de vie de ce système. Le travail présente une approche basée sur le système d’inférence neuro-fuzzy adaptatif (ANFIS) combiné avec (FCM) pour prédire la trajectoire future et calculer le pourcentage de durée de vie du système. Les résultats obtenus démontrent la validité de la technique proposée pour atteindre les objectifs requis avec une précision de haut niveau. Pour améliorer les performances d’ANFIS, nous proposons la distribution de Parzen comme nouvelle fonction d’appartenance de l’algorithme ANFIS. Les résultats ont démontré l’importance de la technique proposée car elle s’est avérée efficace pour réduire le temps de calcul. En outre, la distribution de Parzen présentait la plus petite erreur quadratique moyenne (RMSE). La dernière partie de cette thèse se concentrait sur la proposition d’un nouvel algorithme pouvant être appliqué pour obtenir un système de surveillance en temps réel s’appuyant sur la prédiction de défauts ; cela signifie que cette méthode permet de prédire l’état futur du système, puis de diagnostiquer quelle est la source d’erreur probable. Elle permet d’évaluer la dégradation d’une colonne de distillation et de diagnostiquer par la suite les défauts ou accidents pouvant survenir à la suite de la dégradation estimée. Cette nouvelle approche combine les avantages d’ANFIS à ceux de RNA permettant d’atteindre un haut niveau de précision. / The distillation process is largely used in many applications such a petrochemical production, natural gas processing, and petroleum refineries, etc. Usually, maintenance of the chemical reactors is very costly and it disrupts production for long periods of time. All these factors really demonstrate the fundamental need for effective fault diagnosis and prognostic strategies that they are able to reduce and avoid the greatest number of thes problems and disasters. The first part of our work aims to propose a reliable diagnostic method that can be used in the steady-state regime of a nonlinear procedure. Moreover, we propose a modified procedure of the fuzzy c-means clustering method (MFCM) where MFCM calculates the percentage variation between the two clustered classes. The purpose of using MFCM is to reduce the computing time and increase the performance of the classifier. The results of the proposed method confirm the ability to classify between normal mode and eight abnormal modes of faults. Our second goal aims to propose a prognosis reliable method used to estimate the degradation path of a distillation column and calculate the lifetime percentage of this system. The work presents an approach based on adaptive neuro-fuzzy inference system (ANFIS) combined with (FCM) to predict the future path and calculate the lifetime percentage of the system. The results obtained demonstrate the validity of the proposed technique to achieve the needed objectives with a high-level accuracy. To improve ANFIS performance we propose Parzen windows distribution as a new membership function for ANFIS algorithm. Results demonstrated the importance of the proposed technique since it proved to be highly successful in terms of reducing the time consumed. Additionally, Parzen windows had the smallest Root Mean Square Error (RMSE). The last part of this thesis was focusing on the proposing of new algorithm which can be applied to obtain real-time monitoring system which relies on the fault production module to reach the diagnosis module in contrast to the previous strategies ; this means this method predict the future state of the system then diagnosis what is the probable fault source. This proposed method has proven to be a reliable process that can evaluate the degradation of a distillation column and subsequently diagnose the possible faults or accidents that can emerge as a result of the estimated degradation. This new approach combines the benefits of ANFIS with the benefits of feedforward ANN. The results were demonstrated that the technique achieved with a high level of accuracy, the objective of prediction and diagnosis especially when applied to the data obtained from automated distillation process in the chemical industry.

Orientador: Fábio Bertequini Leão / Resumo: Este trabalho apresenta modelos de programação não linear e linear inteira binária como novos métodos para resolver o problema de diagnóstico de faltas em sistemas de transmissão de energia elétrica. Os modelos de otimização são desenvolvidos com base no conjunto de coberturas mínimas e possui como restrições as equações que descrevem a lógica e a filosofia de proteção empregadas por empresas de energia elétrica. As equações de restrições modelam a associação dos alarmes dos relés de proteção informados pelo sistema de supervisão e aquisição de dados (SCADA) com os estados esperados das funções dos relés de proteção. Os modelos de programação matemática realizam o diagnóstico de falta em uma única etapa, identificando a seção em falta através da análise dos estados dos disjuntores e das funções de proteção associadas a cada equipamento do sistema elétrico. O modelo proposto é um problema muito complexo de programação não linear inteira binária, portanto é reformulado como outro problema, em que algumas expressões são linearizadas, o que resulta em um modelo matemático de programação linear inteiro binário. A solução ótima obtida pelo modelo proposto é encontrada utilizando solvers comerciais de programação matemática. Os resultados obtidos mostram eficiência e robustez do modelo matemático. Na literatura, o problema de diagnóstico de falta é resolvido principalmente por técnicas heurísticas, portanto, o método proposto é inovador. / Doutor

<div>The electric power system of a deep space vehicle is mission-critical, and needs to operate autonomously because of high latency in communicating with ground-based mission control. Key tasks to be automated include managing loads under various physical constraints, continuously monitoring the system state to detect and locate faults, and efficiently responding to those faults. </div><div><br></div><div>This work focuses on three aspects for achieving autonomous, fault-tolerant operation in the dc power system of a spacecraft. First, a sequential procedure is proposed to estimate the node voltages and branch currents in the power system from erroneous sensor measurements. An optimal design for the sensor network is also put forth to enable reliable sensor fault detection and identification. Secondly, a machine-learning based approach that utilizes power-spectrum based features of the current signal is suggested to identify component faults in power electronic converters in the system. Finally, an optimization algorithm is set</div><div>forth that decides how to operate the power system under both normal and faulted conditions. Operational decisions include shedding loads, switching lines, and controlling battery charging. Results of case studies considering various faults in the system are presented.</div>

No description available.

Indiana University-Purdue University Indianapolis (IUPUI) / Electronic Fuel Control (EFC) valve regulates fuel flow to the injector fuel supply line in the Cummins Pressure Time (PT) fuel system. The EFC system controls the fuel flow by means of a variable orifice that is electrically actuated. The supplier of the EFC valves inspects all parts before they are sent out. Their inspection test results provide a characteristic curve which shows the relationship between pressure and current provided to the EFC valve. This curve documents the steady state characteristics of the valve but does not adequately capture its dynamic response. A dynamic test procedure is developed in order to evaluate the performance of the EFC valves. The test itself helps to understand the effects that proposed design changes will have on the stability of the overall engine system. A by product of this test is the ability to evaluate returned EFC valves that have experienced stability issues. The test determines whether an EFC valve is faulted or not before it goes out to prime time use. The characteristics of a good valve and bad valve can be observed after the dynamic test. In this thesis, a mathematical model has been combined with experimental research to investigate and understand the behavior of the characteristics of different types of EFC valves. The model takes into account the dynamics of the electrical and mechanical portions of the EFC valves. System Identification has been addressed to determine the transfer functions of the different types of EFC valves that were experimented. Methods have been used both in frequency domain as well as time domain. Also, based on the characteristic patterns exhibited by the EFC valves, fuzzy logic has been implemented for the use of pattern classification.