Neste trabalho, um novo enfoque para detecção e diagnóstico de falhas (DDF) em robôs manipuladores é apresentado. Um robô com falhas pode causar sérios danos e pode colocar em risco o pessoal presente no ambiente de trabalho. Geralmente, os pesquisadores têm proposto esquemas de DDF baseados no modelo matemático do sistema. Contudo, erros de modelagem podem ocultar os efeitos das falhas e podem ser uma fonte de alarmes falsos. Aqui, duas redes neurais artificiais são utilizadas em um sistema de DDF para robôs manipuladores. Um perceptron multicamadas treinado por retropropagação do erro é usado para reproduzir o comportamento dinâmico do manipulador. As saídas do perceptron são comparadas com as variáveis medidas, gerando o vetor de resíduos. Em seguida, uma rede com função de base radial é usada para classificar os resíduos, gerando a isolação das falhas. Quatro algoritmos diferentes são empregados para treinar esta rede. O primeiro utiliza regularização para reduzir a flexibilidade do modelo. O segundo emprega regularização também, mas ao invés de um único termo de penalidade, cada unidade radial tem um regularização individual. O terceiro algoritmo emprega seleção de subconjuntos para selecionar as unidades radiais a partir dos padrões de treinamento. O quarto emprega o mapa auto-organizável de Kohonen para fixar os centros das unidades radiais próximos aos centros dos aglomerados de padrões. Simulações usando um manipulador com dois graus de liberdade e um Puma 560 são apresentadas, demostrando que o sistema consegue detectar e diagnosticar corretamente falhas que ocorrem em conjuntos de padrões não-treinados. / In this work, a new approach for fault detection and diagnosis in robotic manipulators is presented. A faulty robot could cause serious damages and put in risk the people involved. Usually, researchers have proposed fault detection and diagnosis schemes based on the mathematical model of the system. However, modeling errors could obscure the fault effects and could be a false alarm source. In this work, two artificial neural networks are employed in a fault detection and diagnosis system to robotic manipulators. A multilayer perceptron trained with backpropagation algorithm is employed to reproduce the robotic manipulator dynamical behavior. The perceptron outputs are compared with the real measurements, generating the residual vector. A radial basis function network is utilized to classify the residual vector, generating the fault isolation. Four different algorithms have been employed to train this network. The first utilizes regularization to reduce the flexibility of the model. The second employs regularization too, but instead of only one penalty term, each radial unit has a individual penalty term. The third employs subset selection to choose the radial units from the training patterns. The forth algorithm employs the Kohonen’s self-organizing map to fix the radial unit center near to the cluster centers. Simulations employing a two link manipulator and a Puma 560 manipulator are presented, demonstrating that the system can detect and isolate correctly faults that occur in nontrained pattern sets.

Countries world-over have thousands of critical structures and bridges which have been built decades back when strength-based designs were the order of the day. Over the years, magnitude and frequency of loadings on these have increased. Also, these structures have been exposed to environmental degradation during their service life. Hence, structural health monitoring (SHM) has attracted the attention of researchers, world over. Structural health monitoring is recommended both for vulnerable old bridges and structures as well as for new important structures. Structural health monitoring as a principle is derived from condition monitoring of machinery, where the day-to-day recordings of sound and vibration from machinery is compared and sudden changes in their features is reported for inspection and trouble-shooting. With the availability of funds for repair and retrofitting being limited, it has become imperative to rank buildings and bridges that require rehabilitation for prioritization. Visual inspection and expert judgment continues to rule the roost. Non-destructive testing techniques though have come of age and are providing excellent inputs for judgment cannot be carried out indiscriminately. They are best suited for evaluating local damage when restricted areas are investigated in detail. A few modern bridges, particularly long-span bridges have been provided with sophisticated instrumentation for health monitoring. It is necessary to identify local damages existing in normal bridges. The methodology adopted for such identification should be simple, both in terms of investigations involved and the instrumentation. Researchers have proposed various methodologies including damage identification from mode shapes, wavelet-based formulations and optimization-based damage identification and instrumentation schemes and so on. These are technically involved but may be difficult to be applied for all critical bridges, where the sheer volume of number of bridges to be investigated is enormous. Ideally, structural health monitoring has to be carried out in two stages: (a) Stage-1: Remote monitoring of global damage indicators and inference of the health of the structure. Instrumentation for this stage should be less, simple, but at critical locations to capture the global damage in a reasonable sense. (b) Stage -2: If global indicators show deviation beyond a specified threshold, then a detailed and localized instrumentation and monitoring, with controlled application of static and dynamic loads is to be carried out to infer the health of the structure and take a decision on the repair and retrofit strategies. The thesis proposes the first stage structural health monitoring methodology using natural frequencies and static deflections as damage indicators. The idea is that the stage-1 monitoring has to be done for a large number of bridges and vulnerable structures in a remote and wire-less way and a centralized control and processing unit should be able to number-crunch the in-coming data automatically and the features extracted from the data should help in determining whether any particular bridge warrants second stage detailed investigation. Hence, simple and robust strategies are required for estimating the health of the structure using some of the globally available response data. Identification methodology developed in this thesis is applicable to distributed smeared damage, which is typical of reinforced concrete structures. Simplified expressions and methodologies are proposed in the thesis and numerically and experimentally validated towards damage estimation of typical structures and elements from measured natural frequencies and static deflections. The first-order perturbation equation for a dynamical system is used to derive the relevant expressions for damage identification. The sensitivity of Eigen-value-cumvector pair to damage, modeled as reduction in flexural rigidity (EI for beams, AE for axial rods and Et 12(1 2 )3− μ for plates) is derived. The forward equation relating the changes in EI to changes in frequencies is derived for typical structural elements like simply-supported beams, plates and axial rods (along with position and extent of damage as the other controlling parameters). A distributed damage is uniquely defined with its position, extent and magnitude of EI reduction. A methodology is proposed for the inverse problem, making use of the linear relationship between the reductions in EI (in a smeared sense) to Eigen-values, such that multiple damages could be estimated using changes in natural frequencies. The methodology is applied to beams, plates and axial rods. The performance of this inverse methodology under influence of measurement errors is investigated for typical error profiles. For a discrete three dimensional structure, computationally derived sensitivity matrix is used to solve the damages in each floor levels, simulating the post-earthquake damage scenario. An artificial neural network (ANN) based Radial basis function network (RBFN) is also used to solve the multivariate interpolation problem, with appropriate training sets involving a number of pairs of damage and Eigen-value-change vectors. The acclaimed Cawley-Adams criteria (1979) states that, “the ratio of changes in natural frequencies between two modes is independent of the damage magnitude” and is governed only by the position (or location) and extent of damage. This criterion is applied to a multiple damage problem and contours with equal frequency change ratios, termed as Iso_Eigen_value_change contours are developed. Intersection of these contours for different pairs of frequencies shows the position and extent of damage. Experimental and analytical verification of damage identification methodology using Cawley-Adams criteria is successfully demonstrated. Sensitivity expressions relating the damages to changes in static deflections are derived and numerically and experimentally proved. It is seen that this process of damage identification from static deflections is prone to more errors if not cautiously exercised. Engineering and physics based intuition is adopted in setting the guidelines for efficient damage detection using static deflections. In lines of Cawley-Adams criteria for frequencies, an invariant factor based on static deflections measured at pairs of symmetrical points on a simply supported beam is developed and established. The power of the factor is such that it is governed only by the position of damage and invariant with reference to extent and magnitude of damage. Such a revelation is one step ahead of Caddemi and Morassi’s (2007) recent paper, dealing with static deflection based damage identification for concentrated damage. The invariant factor makes it an ideal candidate for base-line-free measurement, if the quality and resolution of instrumentation is good. A moving damage problem is innovatively introduced in the experiment. An attempt is made to examine wave-propagation techniques for damage identification and a guideline for modeling wave propagation as a transient dynamic problem is done. The reflected-wave response velocity (peak particle velocity) as a ratio of incident wave response is proposed as a damage indicator for an axial rod (representing an end-supported pile foundation). Suitable modifications are incorporated in the classical expressions to correct for damping and partial-enveloping of advancing wave in the damage zone. The experimental results on axial dynamic response of free-free beams suggest that vibration frequency based damage identification is a viable complementary tool to wave propagation. Wavelet-multi-resolution analysis as a feature extraction tool for damage identification is also investigated and structural slope (rotation) and curvatures are found to be the better indicators of damage coupled with wavelet analysis. An adaptive excitation scheme for maximizing the curvature at any arbitrary point of interest is also proposed. However more work is to be done to establish the efficiency of wavelets on experimentally derived parameters, where large noise-ingression may affect the analysis. The application of time-period based damage identification methodology for post-seismic damage estimation is investigated. Seismic damage is postulated by an index based on its plastic displacement excursion and the cumulative energy dissipated. Damage index is a convenient tool for decision making on immediate-occupancy, life-safety after repair and demolition of the structure. Damage sensitive soft storey structure and a weak story structure are used in the non-linear dynamic analysis and the DiPasquale-Cakmak (1987) damage index is calibrated with Park-Ang (1985) damage index. The exponent of the time-period ratio of DiPasquale-Cakmak model is modified to have consistency of damage index with Park-Ang (1985) model.

In the Canadian Prairies recurring droughts are one of the realities which can have significant economical, environmental, and social impacts. For example, droughts in 1997 and 2001 cost over $100 million on different sectors. Drought frequency analysis is a technique for analyzing how frequently a drought event of a given magnitude may be expected to occur. In this study the state of the science related to frequency analysis of droughts is reviewed and studied. The main contributions of this thesis include development of a model in Matlab which uses the qualities of Fuzzy C-Means (FCMs) clustering and corrects the formed regions to meet the criteria of effective hydrological regions. In FCM each site has a degree of membership in each of the clusters. The algorithm developed is flexible to get number of regions and return period as inputs and show the final corrected clusters as output for most case scenarios. While drought is considered a bivariate phenomena with two statistical variables of duration and severity to be analyzed simultaneously, an important step in this study is increasing the complexity of the initial model in Matlab to correct regions based on L-comoments statistics (as apposed to L-moments). Implementing a reasonably straightforward approach for bivariate drought frequency analysis using bivariate L-comoments and copula is another contribution of this study. Quantile estimation at ungauged sites for return periods of interest is studied by introducing two new classes of neural network and machine learning: Radial Basis Function (RBF) and Support Vector Machine Regression (SVM-R). These two techniques are selected based on their good reviews in literature in function estimation and nonparametric regression. The functionalities of RBF and SVM-R are compared with traditional nonlinear regression (NLR) method. As well, a nonlinear regression with regionalization method in which catchments are first regionalized using FCMs is applied and its results are compared with the other three models. Drought data from 36 natural catchments in the Canadian Prairies are used in this study. This study provides a methodology for bivariate drought frequency analysis that can be practiced in any part of the world.

This thesis discusses the modeling of a micro-grid with photovoltaic (PV)-wind-fuel cell (FC) hybrid energy system and its operations. The system consists of the PV power, wind power, FC power, static var compensator (SVC) and an intelligent power controller. Wind and PV are primary power sources of the system, and an FC-electrolyzer combination is used as a backup and a long-term storage system. A simulation model for the micro-grid control of hybrid energy system has been developed using MATLAB/Simulink. A SVC was used to supply reactive power and regulate the voltage of the hybrid system. To achieve a fast and stable response for the real power control, the intelligent controller consists of a Radial Basis Function Network-Sliding Mode Control (RBFNSM) and a General Regression Neural Network (GRNN) for maximum power point tracking (MPPT). The pitch angle of wind turbine is controlled by RBFNSM, and the PV system uses GRNN, where the output signal is used to control the DC/DC boost converters to achieve the MPPT.

Processor architectures are becoming increasingly complex and hence architects have to evaluate a large design space consisting of several parameters, each with a number of potential settings. In order to assist in guiding design decisions we develop simple and accurate models of the superscalar processor design space using a detailed and validated superscalar processor simulator. Firstly, we obtain precise estimates of all significant micro-architectural parameters and their interactions by building linear regression models using simulation based experiments. We obtain good approximate models at low simulation costs using an iterative process in which Akaike’s Information Criteria is used to extract a good linear model from a small set of simulations, and limited further simulation is guided by the model using D-optimal experimental designs. The iterative process is repeated until desired error bounds are achieved. We use this procedure for model construction and show that it provides a cost effective scheme to experiment with all relevant parameters. We also obtain accurate predictors of the processors performance response across the entire design-space, by constructing radial basis function networks from sampled simulation experiments. We construct these models, by simulating at limited design points selected by latin hypercube sampling, and then deriving the radial neural networks from the results. We show that these predictors provide accurate approximations to the simulator’s performance response, and hence provide a cheap alternative to simulation while searching for optimal processor design points.

The present work focuses on the system identification method of aerodynamic parameter estimation which is used to calculate the stability and control derivatives required for aircraft flight mechanics. A new rotorcraft parameter estimation technique is proposed which uses a type of artificial neural network (ANN) called radial basis function network (RBFN). Rotorcraft parameter estimation using ANN is an unexplored research topic and the earlier works in this area have used the output error, equation error and filter error methods which are conventional parameter estimation methods. However, the conventional methods require an accurate non-linear rotorcraft simulation model which is not required by the ANN based method. The application of RBFN overcomes the drawbacks of multilayer perceptron (MLP) based delta method of parameter estimation and gives satisfactory results at either end of the ordered set of estimates. This makes the RBFN based delta method for parameter estimation suitable for rotorcraft studies, as both transition and high speed flight regime characteristics can be studied. The RBFN based delta method for parameter estimation is used for computation of aerodynamic parameters from both simulated and real time flight data. The simulated data is generated from an 8-DoF non-linear simulation model based on the Level-1 criteria of rotorcraft simulation modeling. The generated simulated data is used for computation of the quasi-steady and the time-variant stability and control parameters for different flight conditions using the RBFN based delta method. The performance of RBFN based delta method is also analyzed in the presence of state and measurement noise as well as outliers. The established methodology is then applied to compute parameters directly from real time flight test data for a BO 105 S123 helicopter obtained from DLR (German Aerospace Center). The parameters identified using the RBFN based delta method are compared with the identified values for the BO 105 helicopter from published literature which have used conventional parameter estimation techniques for parameter estimation using a 6-DoF and a 9-DoF rotorcraft simulation model. Finally, the estimated parameters are verified from the flight data generated by a frequency sweep pilot control input for assessing the predictive capability of the RBFN based delta method. Since the approach directly computes the parameters from flight data, it can be used for a reliable description of the higher frequency range, which is needed for high bandwidth flight control and in-flight simulation.

In the Canadian Prairies recurring droughts are one of the realities which can have significant economical, environmental, and social impacts. For example, droughts in 1997 and 2001 cost over $100 million on different sectors. Drought frequency analysis is a technique for analyzing how frequently a drought event of a given magnitude may be expected to occur. In this study the state of the science related to frequency analysis of droughts is reviewed and studied. The main contributions of this thesis include development of a model in Matlab which uses the qualities of Fuzzy C-Means (FCMs) clustering and corrects the formed regions to meet the criteria of effective hydrological regions. In FCM each site has a degree of membership in each of the clusters. The algorithm developed is flexible to get number of regions and return period as inputs and show the final corrected clusters as output for most case scenarios. While drought is considered a bivariate phenomena with two statistical variables of duration and severity to be analyzed simultaneously, an important step in this study is increasing the complexity of the initial model in Matlab to correct regions based on L-comoments statistics (as apposed to L-moments). Implementing a reasonably straightforward approach for bivariate drought frequency analysis using bivariate L-comoments and copula is another contribution of this study. Quantile estimation at ungauged sites for return periods of interest is studied by introducing two new classes of neural network and machine learning: Radial Basis Function (RBF) and Support Vector Machine Regression (SVM-R). These two techniques are selected based on their good reviews in literature in function estimation and nonparametric regression. The functionalities of RBF and SVM-R are compared with traditional nonlinear regression (NLR) method. As well, a nonlinear regression with regionalization method in which catchments are first regionalized using FCMs is applied and its results are compared with the other three models. Drought data from 36 natural catchments in the Canadian Prairies are used in this study. This study provides a methodology for bivariate drought frequency analysis that can be practiced in any part of the world.

Στην παρούσα διπλωματική εργασία θα γίνει μελέτη και εφαρμογή μεθόδων επίλυσης του προβλήματος αναγνώρισης γεωμετρικών χαρακτηριστικών ανθρώπινων ερυθρών αιμοσφαιρίων από προσομοιωμένες εικόνες σκέδασης ΗΜ ακτινοβολίας ενός He-Ne laser 632.8 μm. Στο πρώτο κεφάλαιο γίνεται μια εισαγωγή στις ιδιότητες και τα χαρακτηριστικά του ερυθροκυττάρου καθώς, επίσης, παρουσιάζονται διάφορες ανωμαλίες των ερυθροκυττάρων και οι μέχρι στιγμής χρησιμοποιούμενοι τρόποι ανίχνευσής των. Στο δεύτερο κεφάλαιο της εργασίας γίνεται μια εισαγωγή στις ιδιότητες της ΗΜ ακτινοβολίας, περιγράφεται το φαινόμενο της σκέδασης και παρουσιάζεται το ευθύ πρόβλημα σκέδασης ΗΜ ακτινοβολίας ανθρώπινων ερυθροκυττάρων. Το τρίτο κεφάλαιο αποτελείται από δύο μέρη. Στο πρώτο μέρος γίνεται εκτενής ανάλυση της θεωρίας των τεχνητών νευρωνικών δικτύων και περιγράφονται τα νευρωνικά δίκτυα ακτινικών συναρτήσεων RBF. Στη συνέχεια, αναφέρονται οι μέθοδοι εξαγωγής παραμέτρων και, πιο συγκεκριμένα, δίνεται το θεωρητικό και μαθηματικό υπόβαθρο των μεθόδων που χρησιμοποιήθηκαν οι οποίες είναι ο αλογόριθμος Singular Value Decomposition (SVD), o Angular Radial μετασχηματισμός (ART) και φίλτρα Gabor. Στο δεύτερο μέρος περιγράφεται η επίλυση του αντίστροφου προβλήματος σκέδασης. Παρουσιάζεται η μεθοδολογία της διαδικασίας επίλυσης όπου εφαρμόστηκαν ο αλογόριθμος συμπίεσης εικόνας SVD, o περιγραφέας σχήματος ART και ο περιγραφέας υφής με φίλτρα Gabor για την εύρεση των γεωμετρικών χαρακτηριστικών και νευρωνικό δίκτυο ακτινικών συναρτήσεων RBF για την ταξινόμηση των ερυθροκυττάρων. Στο τέταρτο και τελευταίο κεφάλαιο γίνεται δοκιμή και αξιολόγηση της μεθόδου και συνοψίζονται τα αποτελέσματα και τα συμπεράσματα που εξήχθησαν κατά τη διάρκεια της εκπόνησης αυτής της διπλωματικής. / In this thesis we study and implement methods of estimating the geometrical features of the human red blood cell from a set of simulated light scattering images produced by a He-Ne laser beam at 632.8 μm. Ιn first chapter an introduction to the properties and the characteristics of red blood cells are presented. Furthermore, we describe various abnormalities of erythrocytes and the until now used ways of detection. In second chapter the properties of electromagnetic radiation and the light scattering problem of EM radiation from human erythrocytes are presented. The third chapter consists of two parts. In first part we analyse the theory of neural networks and we describe the radial basis function neural network. Then, we describe the theoritical and mathematical background of the methods that we use for feature extraction which are Singular Value Decomposition (SVD), Angular Radial Transform and Gabor filters. In second part the solution of the inverse problem of light scattering is described. We present the methodology of the solution process in which we implement a Singular Value Decomposition approach, a shape descriptor with Angular Radial Transform and a homogenous texture descriptor which uses Gabor filters for the estimation of the geometrical characteristics and a RBF neural network for the classification of the erythrocytes. In the forth and last chapter the described methods are evaluated and we summarise the experimental results and conclusions that were extracted from this thesis.

Neste trabalho, um novo enfoque para detecção e diagnóstico de falhas (DDF) em robôs manipuladores é apresentado. Um robô com falhas pode causar sérios danos e pode colocar em risco o pessoal presente no ambiente de trabalho. Geralmente, os pesquisadores têm proposto esquemas de DDF baseados no modelo matemático do sistema. Contudo, erros de modelagem podem ocultar os efeitos das falhas e podem ser uma fonte de alarmes falsos. Aqui, duas redes neurais artificiais são utilizadas em um sistema de DDF para robôs manipuladores. Um perceptron multicamadas treinado por retropropagação do erro é usado para reproduzir o comportamento dinâmico do manipulador. As saídas do perceptron são comparadas com as variáveis medidas, gerando o vetor de resíduos. Em seguida, uma rede com função de base radial é usada para classificar os resíduos, gerando a isolação das falhas. Quatro algoritmos diferentes são empregados para treinar esta rede. O primeiro utiliza regularização para reduzir a flexibilidade do modelo. O segundo emprega regularização também, mas ao invés de um único termo de penalidade, cada unidade radial tem um regularização individual. O terceiro algoritmo emprega seleção de subconjuntos para selecionar as unidades radiais a partir dos padrões de treinamento. O quarto emprega o mapa auto-organizável de Kohonen para fixar os centros das unidades radiais próximos aos centros dos aglomerados de padrões. Simulações usando um manipulador com dois graus de liberdade e um Puma 560 são apresentadas, demostrando que o sistema consegue detectar e diagnosticar corretamente falhas que ocorrem em conjuntos de padrões não-treinados. / In this work, a new approach for fault detection and diagnosis in robotic manipulators is presented. A faulty robot could cause serious damages and put in risk the people involved. Usually, researchers have proposed fault detection and diagnosis schemes based on the mathematical model of the system. However, modeling errors could obscure the fault effects and could be a false alarm source. In this work, two artificial neural networks are employed in a fault detection and diagnosis system to robotic manipulators. A multilayer perceptron trained with backpropagation algorithm is employed to reproduce the robotic manipulator dynamical behavior. The perceptron outputs are compared with the real measurements, generating the residual vector. A radial basis function network is utilized to classify the residual vector, generating the fault isolation. Four different algorithms have been employed to train this network. The first utilizes regularization to reduce the flexibility of the model. The second employs regularization too, but instead of only one penalty term, each radial unit has a individual penalty term. The third employs subset selection to choose the radial units from the training patterns. The forth algorithm employs the Kohonen’s self-organizing map to fix the radial unit center near to the cluster centers. Simulations employing a two link manipulator and a Puma 560 manipulator are presented, demonstrating that the system can detect and isolate correctly faults that occur in nontrained pattern sets.

Made available in DSpace on 2016-08-17T14:53:28Z (GMT). No. of bitstreams: 1 Dissertacao Gustavo Araujo.pdf: 2606649 bytes, checksum: efb1a5ded768b058f25d23ee8967bd38 (MD5) Previous issue date: 2014-04-28 / In this work the main objective is to present the development of learning algorithms for online application for the solution of algebraic Hamilton-Jacobi-Bellman equation. The concepts covered are focused on developing the methodology for control systems, through techniques that aims to design online adaptive controllers to reject noise sensors, parametric variations and modeling errors. Concepts of neurodynamic programming and reinforcement learning are are discussed to design algorithms where the context of a given operating point causes the control system to adapt and thus present the performance according to specifications design. Are designed methods for online estimation of adaptive critic focusing efforts on techniques for gradient estimating of the environment value function. / Neste trabalho o principal objetivo é apresentar o desenvolvimento de algoritmos de aprendizagem para execução online para a solução da equação algébrica de Hamilton-Jacobi-Bellman. Os conceitos abordados se concentram no desenvolvimento da metodologia para sistemas de controle, por meio de técnicas que tem como objetivo o projeto online de controladores adaptativos são projetados para rejeitar ruídos de sensores, variações paramétricas e erros de modelagem. Conceitos de programação neurodinâmica e aprendizagem por reforço são abordados para desenvolver algoritmos onde a contextualização de determinado ponto de operação faz com que o sistema de controle se adapte e, dessa forma, apresente o desempenho de acordo com as especificações de projeto. Desenvolve-se métodos para a estimação online do crítico adaptativo concentrando os esforços em técnicas de estimação do gradiente da função valor do ambiente.