O diagnóstico de eventos que afetam a qualidade da energia elétrica tem se tornado preocupação de magnitude mundial, em especial em dois temas importantes que são: a localização relativa da origem do evento (LROE) e a classificação automática da causa fundamental de eventos (CACFE). O primeiro está relacionado com a identificação da fonte do evento, isto é, a montante ou a jusante do medidor de qualidade de energia (MQE). O segundo pode ser dividido em dois grupos: a classificação das causas internas e das causas externas. As causas internas estão relacionadas a eventos produzidos pela operação do sistema elétrico (energização ou desenergização do sistema, energização de transformador, chaveamento de capacitores dentre outros), e as causas externas estão vinculadas a eventos produzidos por faltas externas ao sistema elétrico (contato com galhos de árvore, animais, descargas atmosféricas, dentre outros). Ambos os temas, LROE e CACFE, são abordados nesta tese de doutorado. Para classificar eventos por causas internas ou externas é necessário antes definir se realmente trata-se ou não de um evento, para o qual é imprescindível conhecer a LROE. Este último necessita de um processo de segmentação das formas de onda de tensão e corrente para funcionar de forma correta. A segmentação identifica segmentos transitórios e não transitórios nas formas de onda e contribui também na extração de características para os diferentes algoritmos de classificação. Neste sentido, neste trabalho de pesquisa é proposta uma metodologia de diagnóstico da qualidade de eventos, focada em LROE e CACFE. Para isto foram desenvolvidos diferentes algoritmos de segmentação, extração de características e classificação, sendo criada uma ferramenta computacional em MatLab® que inclui pré-processamento de sinais de tensão e corrente de um banco de dados real fornecido por uma concessionária do Estado de São Paulo. Além disto, foram propostos novos algoritmos de LROE com resultados satisfatórios quando comparados com outros dois disponíveis na literatura científica. Para as causas internas, dois novos índices são propostos para separar eventos produzidos por faltas e energização de transformadores. Finalmente, são propostos novos algoritmos de extração de características baseados na energia dos coeficientes de decomposição da transformada wavelet bem como o algoritmo à trous modificado. São propostos dois novos vetores de descritores de energia (VDE) baseados no primeiro segmento transitório do evento. Para a classificação destes eventos foi utilizado um algoritmo de indução de regras de decisão (CN2), que gera regras de simples implementação. Todos os métodos de classificação utilizados nesta tese estão baseados em regras, sendo seu desempenho avaliado por meio da matriz de confusão. / Diagnosing events that affect power quality have become a worldwide concern, especially with respect to two important issues related to the relative location of the event origin (RLEO) and automatic cause classification of events (ACCE). The first one is related to the identification of the event source, i.e. either upstream or downstream in relation to the power quality meter (PQM). The second one can be subdivided into two groups, namely the classification of internal causes and of external causes. Internal causes are related to events produced by power system operation (connection or disconnection of feeders, power transformer inrush, capacitor switching, amongst others) and external causes that are related to events produced by external faults to the power system (network contacts to tree branches, animals contact, atmospheric discharges, amongst others). Both topics, RLEO and ACCE, are herein considered. In order to classify events due to internal or external causes, one should first define whether it is an actual event, what demands the RLEO. This makes use of a segmentation process applied to the voltage and current waveforms. The segmentation identifies the transient and stationary segments within the waveforms, contributing also to the feature extraction for different classification algorithms. Based on the aforementioned, this research proposes a methodology to diagnose power quality events, focusing on RLEO and ACCE. Different algorithms of segmentation, feature extraction and classification were then developed by the use of a computational tool implemented in MatLab®, that considers also the preprocessing of voltage and current signals in a real data base which was made available by a distribution company in Sao Paulo State. Besides that, new RLEO algorithms have shown satisfactory results when compared to algorithms published in the scientific literature. As for the internal causes, two new indices were proposed in order to separate events produced by faults or by the connection of power transformers. New algorithms for feature extraction are proposed, which are based on the energy of decomposition coefficients of the wavelet transform as well as the modified à trous algorithm. Two vectors of energy descriptors are proposed, which are based on the first transient segment of the event. The classification of such events was carried out by an induction algorithm of decision rules (CN2), that generates easily implementable rules. All classification methods utilized in this thesis are based on rules and their performances are assessed by the confusion matrix.

Este trabalho tem a finalidade de analisar algumas técnicas de processamento do sinal mioelétrico, de forma a possibilitar uma posterior implementação de um circuito, que reconheça este sinal e apresente como saída um sinal de controle a ser utilizado em sistemas de reabilitação. Foram simuladas e avaliadas três técnicas de filtragem para o sinal mioelétrico, a fim de atenuar a interferência dos principais ruídos que corrompem este sinal. As técnicas avaliadas foram: filtragem digital clássica; cancelamento de ruído adaptativo e reconstrução do sinal por meio das componentes obtidas pela transformada wavelet. Também foi implementado e analisado um sistema simplificado de reconhecimento dos padrões para este sinal, realizado por meio de uma rede neural artificial, em que foi aplicado em sua entrada o próprio sinal mioelétrico e não suas características obtidas por processamentos matemáticos. Diante dos resultados obtidos os canceladores de ruído adaptativos apresentaram melhores resultados com relação às outras técnicas de filtragem. Apesar de não ter sido adequada para a filtragem, a transformada wavelet mostrou-se uma poderosa ferramenta de análise de sinais, em virtude da sua característica multiresolução. A técnica utilizada para reconhecer os padrões do sinal mostrou bons resultados com os sinais analisados. / This work has the purpose to analyze some techniques for myoelectric signal processing, towards a subsequent implementation of a circuit which can recognize this signal and present as output a control signal to be used in rehabilitation systems. Simulation and evaluation of three filtering techniques for the myoelectric signal were done in order to attenuate the main interferences of noises which corrupt this signal. The evaluated techniques were: classic digital filtering; adaptive noise cancelling and the signal reconstruction through the obtained components by the wavelet transform. A simplified system of pattern recognition for this signal also was implemented and analyzed, accomplished through an artificial neural network. The myoelectric signal itself was applied to the input instead of its characteristics obtained by mathematical processing. According to the results obtained the adaptive noise cancelling presented better results in comparison to the other filtering techniques. Despite not being adequate for filtering, the wavelet transform proved to be a powerful tool for signal analysis, by virtue of its multiresolution characteristics. The technique used to recognize the signal patterns has shown good results with the analyzed signals.

Este trabalho estuda a possibilidade de se adicionar funções de detecção e registro de eventos, que afetam a qualidade de energia elétrica, aos medidores eletrônicos de faturamento. A presença destes medidores na grande maioria dos consumidores ligados à média tensão transforma este recurso em uma poderosa ferramenta de análise. A existência de canais de comunicação remota nestes medidores facilita a coleta dos dados e seu armazenamento. Estes dados transferidos para sistemas com alta capacidade de processamento permitem uma análise mais precisa dos eventos que ocorrem em toda a rede. Foram consideradas as limitações dos medidores de faturamento quanto à sua capacidade de processamento, memória e taxa de amostragem. Os eventos que podem ser detectados com estes medidores ficam restritos a variações de tensão de curta duração (VTCDs) e transitórios oscilatórios de baixa freqüência. As funções criadas para registro de eventos podem ser usadas para registro de sinais em regime permanente, aumentando a capacidade de diagnóstico e análise da qualidade da energia elétrica em locais da rede. São utilizadas como técnicas de detecção de eventos: as variações dos valores eficazes, a decomposição de sinais, a Norma Euclidiana Instantânea (NEI) e a transformada wavelet (TW). / This work shows the possibility to add new functions for detection and registration of disturbances that affect power quality to electronic billing meters. The widespread installation of these meters in the power network makes this resource a powerful analysis tool. Remote communication channels in these meters create an easy way for reading and registering such power quality events. Data transferred to powerful processing systems allow accurate analysis of events occurring in the power grid. The limitations on billing meters, namely processing resources, memory availability and sampling rate, were taken into account. Despite these limitations, the electronic meters can handle short duration voltage events and low-frequency oscillatory transients. Those new functions can also be used for registering steady state phenomena, extending the ability to diagnose power quality problems throughout the power network. Techniques like RMS variations, signal decomposition, instantaneous Euclidian norm and wavelet transform were used for detection of the considered power quality events.

The big data era poses great challenges as well as opportunities for researchers to develop efficient statistical approaches to analyze massive data. Sufficient dimension reduction is such an important tool in modern data analysis and has received extensive attention in both academia and industry. In this dissertation, we introduce inverse regression estimators using Fourier transforms, which is superior to the existing SDR methods in two folds, (1) it avoids the slicing of the response variable, (2) it can be readily extended to solve the high dimensional data problem. For the ultra-high dimensional problem, we investigate both eigenvalue decomposition and minimum discrepancy approaches to achieve optimal solutions and also develop a novel and efficient optimization algorithm to obtain the sparse estimates. We derive asymptotic properties of the proposed estimators and demonstrate its efficiency gains compared to the traditional estimators. The oracle properties of the sparse estimates are derived. Simulation studies and real data examples are used to illustrate the effectiveness of the proposed methods. Wavelet transform is another tool that effectively detects information from time-localization of high frequency. Parallel to our proposed Fourier transform methods, we also develop a wavelet transform version approach and derive the asymptotic properties of the resulting estimators.

Over the last decade, data has become a highly valuable resource. Electrical power grids deal with large quantities of data, and continuously collect this for analytical purposes. Anomalies that occur within this data is important to identify since they could cause nonoptimal performance within the substations, or in worse cases damage to the substations themselves. However, large datasets in the order of millions are hard or even impossible to gain a reasonable overview of the data manually. When collecting data from electrical power grids, predefined triggering criteria are often used to indicate that an event has occurred within the specific system. This makes it difficult to search for events that are unknown to the operator of the deployed acquisition system. Clustering, an unsupervised machine learning method, can be utilised for fault prediction within systems generating large amounts of multivariate time-series data without labels and can group data more efficiently and without the bias of a human operator. A large number of clustering techniques exist, as well as methods for extracting information from the data itself, and identification of these was of utmost importance. This thesis work presents a study of the methods involved in the creation of such a clustering system which is suitable for the specific type of data. The objective of the study was to identify methods that enables finding the underlying structures of the data and cluster the data based on these. The signals were split into multiple frequency sub-bands and from these features could be extracted and evaluated. Using suitable combinations of features the data was clustered with two different clustering algorithms, CLARA and CLARANS, and evaluated with established quality analysis methods. The results indicate that CLARA performed overall best on all the tested feature sets. The formed clusters hold valuable information such as indications of unknown events within the system, and if similar events are clustered together this can assist a human operator further to investigate the importance of the clusters themselves. A further conclusion from the results is that research into the use of more optimised clustering algorithms is necessary so that expansion into larger datasets can be considered.

Currently available methods for managing glaucoma, e.g. the planimetry on stereo disc photographs, involve a subjective component either by the patient or examiner. In addition, a few structures may overlap together on the essential 2-D images, which can decrease reproducibility. Spectral domain optical coherence tomography (SD-OCT) provides a 3-D, cross-sectional, microscale depiction of biological tissues. Given the wealth of volumetric information at microscale resolution available with SD-OCT, it is likely that better parameters can be obtained for measuring glaucoma changes that move beyond what is possible using fundus photography etc. The neural canal opening (NCO) is a 3-D single anatomic structure in SD-OCT volumes. It is proposed as a basis for a stable reference plane from which various optic nerve morphometric parameters can be derived. The overall aim of this Ph.D. project is to develop a framework to segment the 3-D NCO and its related structure retinal vessels using information from SD-OCT volumes and/or fundus photographs to aid the management of glaucoma changes. Based on the mutual positional relationship of the NCO and vessels, a multimodal 3-D scale-learning-based framework is developed to iteratively identify them in SD-OCT volumes by incorporating each other's pre-identified positional information. The algorithm first applies a 3-D wavelet-transform-learning-based layer segmentation and pre-segments the NCO using graph search. To aid a better NCO detection, the vessels are identified either using a SD-OCT segmentation approach incorporating the presegmented NCO positional information to the vessel classification or a multimodal approach combining the complementary features from SD-OCT volumes and fundus photographs (or a registered-fundus approach based on the original fundus vessel segmentation). The obtained vessel positional information is then used to help enhance the NCO segmentation by incorporating that to the cost function of graph search. Note that the 3-D wavelet transform via lifting scheme has been used to remove high frequency noises and extract texture properties in SD-OCT volumes etc. The graph search has been used for finding the optimal solution of 3-D multiple surfaces using edge and additionally regional information. In this work, the use of the 3-D wavelet-transform-learning-based cost function for the graph search is a further extension of the 3-D wavelet transform and graph search. The major contributions of this work include: 1) extending the 3-D graph theoretic segmentation to the use of 3-D scale-learning-based cost function, 2) developing a graph theoretic approach for segmenting the NCO in SD-OCT volumes, 3) developing a 3-D wavelet-transform-learning-based graph theoretic approach for segmenting the NCO in SD-OCT volumes by iteratively utilizing the pre-identified NCO and vessel positional information (from 4 or 5), 4) developing a vessel classification approach in SD-OCT volumes by incorporating the pre-segmented NCO positional information to the vessel classification to suppress the NCO false positives, and 5) developing a multimodal concurrent classification and a registered-fundus approach for better identifying vessels in SD-OCT volumes using additional fundus information.

Im Rahmen des TERA-Projektes (Turbo Evaluation and Rapid Algorithms) wurde ein neuartiger, hochgradig effizienter probabilistischer Algorithmus zum Lösen polynomialer Gleichungssysteme entwickelt und für den komplexen Fall implementiert. Die Geometrie polarer Varietäten gestattet es, diesen Algorithmus zu einem Verfahren zur Charakterisierung der reellen Lösungsmengen polynomialer Gleichungssysteme zu erweitern. Ziel dieser Arbeit ist es, eine Implementierung dieses Verfahrens zur Bestimmung reeller Lösungen auf eine Klasse von Beispielproblemen anzuwenden. Dabei wurde Wert darauf gelegt, dass diese Beispiele reale, praxisbezogene Anwendungen besitzen. Diese Anforderung ist z.B. für polynomiale Gleichungssysteme erfüllt, die sich aus dem Entwurf von schnellen Wavelet-Transformationen ergeben. Die hier betrachteten Wavelet-Transformationen sollen die praktisch wichtigen Eigenschaften der Orthogonalität und Symmetrie besitzen. Die Konstruktion einer solchen Wavelet-Transformation hängt von endlich vielen reellen Parametern ab. Diese Parameter müssen gewisse polynomiale Gleichungen erfüllen. In der veröffentlichten Literatur zu diesem Thema wurden bisher ausschließlich Beispiele mit endlichen Lösungsmengen behandelt. Zur Berechnung dieser Beispiele war es dabei ausreichend, quadratische Gleichungen in einer oder zwei Variablen zu lösen. Zur Charakterisierung der reellen Lösungsmenge eines polynomialen Gleichungssystems ist es ein erster Schritt, in jeder reellen Zusammenhangskomponente mindestens einen Punkt aufzufinden. Schon dies ist ein intrinsisch schweres Problem. Es stellt sich heraus, dass der Algorithmus des TERA-Projektes zur Lösung dieser Aufgabe bestens geeignet ist und daher eine größere Anzahl von Beispielproblemen lösen kann als die besten kommerziell erhältlichen Lösungsverfahren. / As a result of the TERA-project on Turbo Evaluation and Rapid Algorithms a new type, highly efficient probabilistic algorithm for the solution of systems of polynomial equations was developed and implemented for the complex case. The geometry of polar varieties allows to extend this algorithm to a method for the characterization of the real solution set of systems of polynomial equations. The aim of this work is to apply an implementation of this method for the determination of real solutions to a class of example problems. Special emphasis was placed on the fact that those example problems possess real-life, practical applications. This requirement is satisfied for the systems of polynomial equations that result from the design of fast wavelet transforms. The wavelet transforms considered here shall possess the practical important properties of symmetry and orthogonality. The specification of such a wavelet transform depends on a finite number of real parameters. Those parameters have to obey certain polynomial equations. In the literature published on this topic, only example problems with a finite solution set were presented. For the computation of those examples it was sufficient to solve quadratic equations in one or two variables. To characterize the set of real solutions of a system of polynomial equations it is a first step to find at least one point in each connected component. Already this is an intrinsically hard problem. It turns out that the algorithm of the TERA-project performes very well with this task and is able to solve a larger number of examples than the best known commercial polynomial solvers.

3D seismic data are widely used for many different purposes. Despite different objectives, a common goal in almost all 3D seismic programs is to attain better understanding of the subsurface features. In gas injection projects, which are mainly for Enhanced Oil Recovery (EOR) and recently for environmental purposes, seismic data have an important role in the gas monitoring phase. This thesis deals with a 3D seismic investigation at the CO2 injection site at Ketzin, Germany. I focus on two critical aspects of the project: the internal architecture of the heterogeneous Stuttgart reservoir and the detectability of the CO2 response from surface seismic data. Conventional seismic methods are not able to conclusively map the internal reservoir architecture due to their limited seismic resolution. In order to overcome this limitation, I use the Continuous Wavelet Transform (CWT) decomposition technique, which provides frequency spectra with high temporal resolution without the disadvantages of the windowing process associated with the other techniques. Results from applying this technique reveal more of the details of sand bodies within the Stuttgart Formation. The CWT technique also helps to detect and map remnant gas on the top of the structure. In addition to this method, I also show that the pre-stack spectral blueing method, which is presented for the first time in this research, has an ability to enhance seismic resolution with fewer artifacts in comparison with the post-stack spectral blueing method. The second objective of this research is to evaluate the CO2 response on surface seismic data as a feasibility study for CO2 monitoring. I build a rock physics model to estimate changes in elastic properties and seismic velocities caused by injected CO2. Based on this model, I study the seismic responses for different CO2 injection geometries and saturations using one dimensional (1D) elastic modeling and two dimensional (2D) acoustic finite-difference modeling. Results show that, in spite of random and coherent noises and reservoir heterogeneity, the CO2 seismic response should be strong enough to be detectable on surface seismic data. I use a similarity-based image registration method to isolate amplitude changes due to the reservoir from amplitude changes caused by time shifts below the reservoir. In support of seismic monitoring using surface seismic data, I also show that acoustic impedance versus Poisson’s ratio cross-plot is a suitable attribute for distinguishing gas-bearing sands from brine-bearing sands. / CO2SINK Project

Los Radares de Apertura Sintética, o sistemas SAR, representan el mejorejemplo de sistemas activos de teledetección por microondas. Debido a su naturaleza coherente, un sistema SAR es capaz de adquirir información dedispersión electromagnética con una alta resolución espacial, pero por otro lado, esta naturaleza coherente provoca también la aparición de speckle.A pesar de que el speckle es una medida electromagnética, sólo puede ser analizada como una componente de ruido debido a la complejidad asociadacon el proceso de dispersión electromagnética.Para eliminar los efectos del ruido speckle adecuadamente, es necesario un modelo de ruido, capaz de identificar las fuentes de ruido y como éstasdegradan la información útil. Mientras que este modelo existe para sistemasSAR unidimensionales, conocido como modelo de ruido speckle multiplicativo,éste no existe en el caso de sistemas SAR multidimensionales.El trabajo presentado en esta tesis presenta la definición y completa validación de nuevos modelos de ruido speckle para sistemas SAR multidimensionales,junto con su aplicación para la reducción de ruido speckle y la extracción de información.En esta tesis, los datos SAR multidimensionales, se consideran bajo una formulación basada en la matriz de covarianza, ya que permite el análisisde datos sobre la base del producto complejo Hermítico de pares de imágenesSAR. Debido a que el mantenimiento de la resolución especial es un aspectoimportante del procesado de imágenes SAR, la reducción de ruido speckleestá basada, en este trabajo, en la teoría de análisis wavelet.

This study focuses on developing structural health monitoring techniques to detect damage in deepwater risers subjected to vortex-induced vibration (VIV), and studying vibration control strategies to extend the service life of offshore structures. Vibration-based damage detection needs both responses from the undamaged and damaged deepwater risers. Because no experimental data for damaged deepwater risers is available, a model to predict the VIV responses of deepwater risers with given conditions is needed, which is the forward problem. In this study, a new three dimensional (3D) analytical model is proposed considering coupled VIV (in-line and cross-flow) for top-tensioned riser (TTR) with wake oscillators. The model is verified by direct numerical simulations and experimental data. The inverse problem is to detect damage using VIV responses from the analytical models with/without damage, where the change between dynamic properties obtained from riser responses represents damage. The inverse problem is performed in two steps: blind identification and damage detection. For blind identification, a wavelet modified second order blind identification (WMSOBI) method and a complex WMSOBI (CWMSOBI) method are proposed to extract modal properties from output only responses for standing and traveling wave vibration, respectively. Numerical simulations and experiments validate the effectiveness of proposed methods. For damage detection, a novel weighted distribution force change (WDFC) index (for standing wave) and a phase angle change (PAC) index (for traveling wave) are proposed and proven numerically. Experiments confirm that WDFC can accurately locate damage and estimate damage severity. Furthermore, a new fatigue damage estimation method involving WMSOBI, S-N curve and Miner's rule is proposed and proven to be effective using field test data. Vibration control is essential to extend the service life and enhance the safety of offshore structures. Literature review shows that semi-active control devices are potentially a good solution. A novel semi-active control strategy is proposed to tune the damper properties to match the dominant frequency of the structural response in real-time. The effectiveness of proposed strategy in vibration reduction for deepwater risers and offshore floating wind turbines is also validated through numerical studies.