O presente trabalho desenvolve uma topologia de controle para o gerador de indução duplamente alimentado (GIDA) em condições normais e em condições de falta monofásica. O sistema de controle é dividido em três partes principais: sistema de identificação de faltas, controle em condições normais e controle em condições de falta monofásica. A primeira parte, o sistema de identificação (SI) de faltas, é responsável pela seleção da topologia de controle da máquina. O SI é composto por uma combinação entre redes neurais artificiais (RNA) e a Fast Fourier Transform (FFT). As RNA são responsáveis pela identificação do estado atual da rede, se possui falta ou não. Os dados de entrada das RNA são as correntes de linha do estator que passam por um pré-processamento por meio da FFT. Alguns conteúdos harmônicos de saída da FFT irrelevantes no processo de identificação são eliminados por um método similar ao Principal Components Analysis (PCA). A segunda parte do trabalho é o controle em condições normais, sendo ativado quando o SI aponta a ausência de faltas. A topologia de controle vetorial é utilizada nesta condição para manter a tensão e frequência constante com a velocidade mecânica do eixo variável. A última parte do trabalho é o controle em condições adversas, que é ativado quando o SI detecta uma falta monofásica. A topologia de controle nesta condição utiliza as transformações ortogonais para reduzir o fluxo concatenado no enrolamento do estator com falta. A utilização deste novo controle reduz a corrente do estator quando comparado com o controle vetorial em condições de falta, sendo que a tensão do estator nas fases sem falta é mantida dentro de uma faixa de operação. O trabalho possui resultados de simulação das três principais partes do sistema de controle. Primeiramente, resultados do controle vetorial de tensão e frequência do GIDA sob condições de velocidade do eixo variável e cortes de carga são apresentados. Logo após, apresenta-se os resultados do SI na identificação de faltas monofásicas na fase B e o seu comportamento sob condições adversas como desequilíbrio de carga e cortes de cargas. Finalmente, alguns resultados do controle em condições de falta sobre uma falta fase-neutro na fase B são apresentados. / This paper presents a control topology for doubly fed induction generator (DFIG) in normal and single fault conditions. The control system is divided into three main parts: fault identification system, control in normal condition and control in single fault conditions. In the first part, the system of identification (SI) is responsible for selecting the topology of the control. The SI is composed by a combination of artificial neural networks (ANN) and Fast Fourier Transform (FFT). The ANN is responsible for identifying the current state of the grid, if has fault or not. The inputs of the ANN are stator currents line through of a pre-processing by means of FFT. Some harmonic contents are irrelevant in the identification process and they are eliminated by a method similar to Principal Components Analysis (PCA). The second part of the paper is the control under normal conditions, activated when the SI indicates the absence of faults. The topology of vector control in this condition is used to maintain the voltage and frequency constant, where the speed of the mechanical axis variable. The last part of the work is the control in adverse conditions, which is activated when the SI detects a singlephase fault. The control topology in this condition uses the orthogonal transformations to reduce the mutual flux in the stator winding with fault. The use of this new control reduces the stator current as compared to vector control in fault conditions, and the stator voltage in the stages without fault is maintained within an operating range. The paper has simulation results of three main parts of the control system. First, the results of the vector control voltage and frequency of DFIG under conditions of variable shaft speed and load sections are provided. Soon after, the results of the SI in identifying faults in the phase B under conditions such as load imbalance and cutting loads are shown. Finally, some results of control in fault condition in the phase B are shown.

O presente trabalho desenvolve uma topologia de controle para o gerador de indução duplamente alimentado (GIDA) em condições normais e em condições de falta monofásica. O sistema de controle é dividido em três partes principais: sistema de identificação de faltas, controle em condições normais e controle em condições de falta monofásica. A primeira parte, o sistema de identificação (SI) de faltas, é responsável pela seleção da topologia de controle da máquina. O SI é composto por uma combinação entre redes neurais artificiais (RNA) e a Fast Fourier Transform (FFT). As RNA são responsáveis pela identificação do estado atual da rede, se possui falta ou não. Os dados de entrada das RNA são as correntes de linha do estator que passam por um pré-processamento por meio da FFT. Alguns conteúdos harmônicos de saída da FFT irrelevantes no processo de identificação são eliminados por um método similar ao Principal Components Analysis (PCA). A segunda parte do trabalho é o controle em condições normais, sendo ativado quando o SI aponta a ausência de faltas. A topologia de controle vetorial é utilizada nesta condição para manter a tensão e frequência constante com a velocidade mecânica do eixo variável. A última parte do trabalho é o controle em condições adversas, que é ativado quando o SI detecta uma falta monofásica. A topologia de controle nesta condição utiliza as transformações ortogonais para reduzir o fluxo concatenado no enrolamento do estator com falta. A utilização deste novo controle reduz a corrente do estator quando comparado com o controle vetorial em condições de falta, sendo que a tensão do estator nas fases sem falta é mantida dentro de uma faixa de operação. O trabalho possui resultados de simulação das três principais partes do sistema de controle. Primeiramente, resultados do controle vetorial de tensão e frequência do GIDA sob condições de velocidade do eixo variável e cortes de carga são apresentados. Logo após, apresenta-se os resultados do SI na identificação de faltas monofásicas na fase B e o seu comportamento sob condições adversas como desequilíbrio de carga e cortes de cargas. Finalmente, alguns resultados do controle em condições de falta sobre uma falta fase-neutro na fase B são apresentados. / This paper presents a control topology for doubly fed induction generator (DFIG) in normal and single fault conditions. The control system is divided into three main parts: fault identification system, control in normal condition and control in single fault conditions. In the first part, the system of identification (SI) is responsible for selecting the topology of the control. The SI is composed by a combination of artificial neural networks (ANN) and Fast Fourier Transform (FFT). The ANN is responsible for identifying the current state of the grid, if has fault or not. The inputs of the ANN are stator currents line through of a pre-processing by means of FFT. Some harmonic contents are irrelevant in the identification process and they are eliminated by a method similar to Principal Components Analysis (PCA). The second part of the paper is the control under normal conditions, activated when the SI indicates the absence of faults. The topology of vector control in this condition is used to maintain the voltage and frequency constant, where the speed of the mechanical axis variable. The last part of the work is the control in adverse conditions, which is activated when the SI detects a singlephase fault. The control topology in this condition uses the orthogonal transformations to reduce the mutual flux in the stator winding with fault. The use of this new control reduces the stator current as compared to vector control in fault conditions, and the stator voltage in the stages without fault is maintained within an operating range. The paper has simulation results of three main parts of the control system. First, the results of the vector control voltage and frequency of DFIG under conditions of variable shaft speed and load sections are provided. Soon after, the results of the SI in identifying faults in the phase B under conditions such as load imbalance and cutting loads are shown. Finally, some results of control in fault condition in the phase B are shown.

This thesis describes an investigation into the influences of arcing and conductor deflection due to magnetic forces on the accuracy of fault locator algorithms in electrical distribution networks. The work also explores the possibilities of using the properties of an arc to identify two specific types of faults that may occur on an overhead distribution line. A new technique using the convolution operator is introduced for deriving differential equation algorithms. The first algorithm was derived by estimating the voltage as an array of impulse functions while the second algorithm was derived using a piecewise linear voltage signal. These algorithms were tested on a simulated single-phase circuit using a PI-model line. It was shown that the second algorithm gave identical results as the existing dynamic integration operator type algorithm. The first algorithm used a transformation to a three-phase circuit that did not require any matrix calculations as an equivalent sequence component circuit is utilised for a single-phase to ground fault. A simulated arc was used to test the influence of the non-linearity of an arc on the accuracy of this algorithm. The simulations showed that the variation in the resistance due to arcing causes large oscillations of the algorithm output and a 40th order mean filter was used to increase the accuracy and stability of the algorithm. The same tests were performed on a previously developed fault locator algorithm that includes a square-wave power frequency proximation of the fault arc. This algorithm gave more accurate and stable results even with large arc length variations. During phase-to-phase fault conditions, two opposing magnetic fields force the conductors outwards away from each other and this movement causes a change in the total inductance of the line. A three dimensional finite element line model based on standard wave equations but incorporating magnetic forces was used to evaluate this phenomenon. The results show that appreciable errors in the distance estimations can be expected especially on poorly tensioned di stribution lines.New techniques were also explored that are based on identification of the fault arc. Two methods were successfully tested on simulated networks to identify a breakingconductor. The methods are based on the rate of increase in arc length during the breaking of the conductor. The first method uses arc voltage increase as the basis of the detection while the second method make use of the increase in the non-linearity of the network resistance to identify a breaking conductor. An unsuccessful attempt was made to identifying conductor clashing caused by high winds: it was found that too many parameters influence the separation speed of the two conductors. No unique characteristic could be found to identify the conductor clashing using the speed of conductor separation. The existing algorithm was also used to estimate the voltage in a distribution network during a fault for power quality monitoring purposes.

Made available in DSpace on 2015-03-03T15:08:47Z (GMT). No. of bitstreams: 1 BrunoSJC_TESE.pdf: 2605632 bytes, checksum: cc7fdbd9d8d7dfe3adac23f17fab1ae2 (MD5) Previous issue date: 2014-05-13 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior / In this work, we propose a two-stage algorithm for real-time fault detection and identification of industrial plants. Our proposal is based on the analysis of selected features using recursive density estimation and a new evolving classifier algorithm. More specifically, the proposed approach for the detection stage is based on the concept of density in the data space, which is not the same as probability density function, but is a very useful measure for abnormality/outliers detection. This density can be expressed by a Cauchy function and can be calculated recursively, which makes it memory and computational power efficient and, therefore, suitable for on-line applications. The identification/diagnosis stage is based on a self-developing (evolving) fuzzy rule-based classifier system proposed in this work, called AutoClass. An important property of AutoClass is that it can start learning from scratch". Not only do the fuzzy rules not need to be prespecified, but neither do the number of classes for AutoClass (the number may grow, with new class labels being added by the on-line learning process), in a fully unsupervised manner. In the event that an initial rule base exists, AutoClass can evolve/develop it further based on the newly arrived faulty state data. In order to validate our proposal, we present experimental results from a level control didactic process, where control and error signals are used as features for the fault detection and identification systems, but the approach is generic and the number of features can be significant due to the computationally lean methodology, since covariance or more complex calculations, as well as storage of old data, are not required. The obtained results are significantly better than the traditional approaches used for comparison / Este trabalho prop?e um algoritmo de dois estagios para detec??o e identifica??o de falhas, em tempo real, em plantas industriais. A proposta baseia-se na analise de caracter?sticas selecionadas utilizando estimativa de densidade recursiva e um novo algoritmo evolutivo de classifica??o. Mais especificamente, a abordagem proposta para detec??o e baseada no conceito de densidade no espa?o de dados, o que difere da tradicional fun??o densidade de probabilidade, porem, sendo uma medida bastante util na detec??o de anormalidades/outliers. Tal densidade pode ser expressa por uma fun??o de Cauchy e calculada recursivamente, o que torna o algoritmo computacionalmente eficiente, em termos de processamento e memoria, e, dessa maneira, apropriado para aplica??es on-line. O estagio de identifica??o/diagnostico e realizado por um classificador baseado em regras fuzzy capaz de se auto-desenvolver (evolutivo), chamado de AutoClass, e introduzido neste trabalho. Uma propriedade importante do AutoClass e que ele e capaz de aprender a partir do zero". Tanto as regras fuzzy, quanto o numero de classes para o algoritmo n?o necessitam de pre-especifica??o (o numero de classes pode crescer, com os rotulos de classe sendo adicionados pelo processo de aprendizagem on-line), de maneira n~ao-supervisionada. Nos casos em que uma base de regras inicial existe, AutoClass pode evoluir/desenvolver-se a partir dela, baseado nos dados adquiridos posteriormente. De modo a validar a proposta, o trabalho apresenta resultados experimentais de simula??o e de aplica??es industriais reais, onde o sinal de controle e erro s?o utilizados como caracter?sticas para os estagios de detec??o e identifica??o, porem a abordagem e generica, e o numero de caracter?sticas selecionadas pode ser significativamente maior, devido ? metodologia computacionalmente eficiente adotada, uma vez que calculos mais complexos e armazenamento de dados antigos n?o s?o necess?rios. Os resultados obtidos s?o signifificativamente melhores que os gerados pelas abordagens tradicionais utilizadas para compara??o

Artificial Intelligence in Automotive Industry / Cost reduction is one of the main concerns in industry. Companies invest considerably for better performance in end-of-line fault diagnosis systems. A common strategy is to use data obtained from existing instrumentation. This research investigates the challenge of learning from historical data that have already been collected by companies. Machine learning is basically one of the most common and powerful techniques of artificial intelligence that can learn from data and identify fault features with no need for human interaction. In this research, labeled sound and vibration measurements are processed into fault signatures for vehicle starter motors and alternators. A fault detection and identification system has been developed to identify fault types for end-of-line testing of motors. However, labels are relatively difficult to obtain, expensive, time consuming and require experienced humans, while unlabeled samples needs less effort to collect. Thus, learning from unlabeled data together with the guidance of few labels would be a better solution. Furthermore, in this research, learning from unlabeled data with absolutely no human intervention is also implemented and discussed as well. / Thesis / Master of Applied Science (MASc)

This thesis presents the development of a fault detection and diagnosis (FDD) procedure capable of capturing, isolating and identifying multiple abrupt parametric faults. The proposed method relies on parameter estimation deployed in a set membership framework. This approach presupposes the utilization of a linearly parametrizable model and the a priori knowledge of bounded noise errors and parameter perturbations. Under these assumptions, a data-hyperspace is generated at every time instant. The goal of set membership identification (SMI) is the determination of the parametric set, formed as an orthotope or ellipsoid, within which the nominal parameter vector resides and intersects with the data-hyperspace. The fault detection mechanism is activated when the normal operation of the SMI procedure is interrupted due to an empty intersection of the data-hyperspace and the estimated parametric set. At the detection instant, a resetting procedure is performed in order to compute the parameter set and the data-hyperspace that contain the varied nominal parameter vector, allowing the SMI algorithm to continue its operation. During the fault isolation, consistency tests are executed, relying on the projections of the worst case parametric sets and the ones arisen from the normal operation of SMI. A faulty component is indicated when these projections do not intersect, while the distance of their centers is used for fault identification. In case of the ellipsoidal SMI-based FDD and under the assumption of a time invariant parameter vector, a new fault detection criterion is defined based on the intersection of support orthotopes of ellipsoids. A more accurate estimation of the time instant of fault occurrence is proposed based on the application of a backward-in-time procedure starting from the fault detection instant, while the conditions under which a fault will never be detected by the orthotopic and ellipsoidal SMI based FDD are provided. This dissertation explores the efficiency of the proposed FDD methodology for capturing failure modes of two microelectromechanical systems; an electrostatic parallel-plate microactuator and a torsionally resonant atomic force microscope. From an engineering point of view, failure modes appeared in the microcomponents of the microactuator and the TR-AFM are encountered as parameter variations and are captured, isolated and identified by the proposed FDD methodology. / Σε αυτή την διατριβή, παρουσιάζεται η ανάπτυξη μιας διαδικασίας Ανίχνευσης και Διάγνωσης Σφαλμάτων, η οποία είναι ικανή να εντοπίζει, απομονώνει και αναγνωρίζει πολλαπλά, απότομα παραμετρικά σφάλματα. H προτεινόμενη μέθοδος βασίζεται στην αναγνώριση του συνόλου συμμετοχής των παραμέτρων. Ο στόχος της Αναγνώρισης Συνόλου Συμμετοχής είναι ο καθορισμός του παραμετρικού συνόλου εντός του οποίου κείται το ονομαστικό διάνυσμα παραμέτρων, δεδομένου ότι το ονομαστικό διάνυσμα παραμέτρων ανήκει επίσης σε έναν υπερχώρο δεδομένων. Το παραμετρικό σύνολο απεικονίζεται ως ένα ορθότοπο ή ένα ελλειψοειδές, λόγω της εύκολης μαθηματικής τους περιγραφής. Έτσι, η διαδικασία Αναγνώρισης Συνόλου Συμμετοχής αντιστοιχεί σε ένα πρόβλημα βελτιστοποίησης, το οποίο αποσκοπεί στον υπολογισμό του ορθοτόπου ή ελλειψοειδούς το οποίο περιέχει το ονομαστικό διάνυσμα παραμέτρων και τέμνεται με τον υπερχώρο δεδομένων. Ο μηχανισμός Ανίχνευσης Σφαλμάτων ενεργοποιείται όταν διακόπτεται η φυσιολογική λειτουργία της Αναγνώρισης Συνόλου Συμμετοχής, λόγω της κενής τομής μεταξύ των εκτιμώμενου παραμετρικού συνόλου και του υπερχώρου δεδομένων. Τη χρονική στιγμή ανίχνευσης ενός σφάλματος, εφαρμόζεται μια διαδικασία επαναρύθμισης που σκοπεύει στον υπολογισμό του νέου παραμετρικού συνόλου, το οποίο περιέχει το μεταβεβλημένο ονομαστικό διάνυσμα παραμέτρων και τέμνεται με το υπερχώρο δεδομένων. Κατά τη διάρκεια της διαδικασίας απομόνωσης του σφάλματος, εκτελούνται τεστ συμβατότητας, τα οποία βασίζονται στις προβολές των νέων παραμετρικών συνόλων και στις προβολές των παραμετρικών συνόλων χείριστης περίπτωσης, ενώ η απόσταση των κέντρων των προβολών χρησιμοποιείται για αναγνώριση σφάλματος. Στην περίπτωση που η Ανίχνευση και Διάγνωση Σφαλμάτων πραγματοποιείται βασιζόμενη στην Αναγνώριση Συνόλου Συμμετοχής με ελλειψοειδή και θεωρώντας το ονομαστικό διάνυσμα παραμέτρων χρονικά αμετάβλητο, ορίζεται ένα νέο κριτήριο ανίχνευσης σφαλμάτων, χρησιμοποιώντας την τομή των περιβαλλόντων ορθοτόπων των ελλειψοειδών. Σε αυτή την περίπτωση, ένα σφάλμα ανιχνεύεται όταν η τομή αυτή είναι κενή. Ακόμη, προτείνεται μια πιο ακριβής εκτίμηση της χρονικής στιγμής εμφάνισης του σφάλματος, ενώ παρατίθενται οι συνθήκες υπό τις οποίες ένα σφάλμα μπορεί να μην ανιχνευθεί ποτέ με την εφαρμογή των προτεινόμενων μεθόδων. Η συγκεκριμένη διατριβή επίσης ερευνά την αποτελεσματικότητα της προτεινόμενης μεθοδολογίας Ανίχνευσης και Διάγνωσης Σφαλμάτων για τον εντοπισμό των τρόπων εκδήλωσης σφαλμάτων σε δύο μικροηλεκτρομηχανικά συστήματα (ΜΗΜΣ), έναν ηλεκτροστατικό μικροεπενεργητή παράλληλων πλακών και ένα ατομικό μικροσκόπιο συντονισμού στρέψης. Από πλευράς μηχανικής, οι τρόποι εκδήλωσης σφαλμάτων στα δομικά στοιχεία του μικροεπενεργητή ή του ατομικού μικροσκοποίου αντιμετωπίζονται ως απότομες παραμετρικές, οι οποίες εντοπίζονται και διαγιγνώσκονται από τις προτεινόμενες μεθόδους.