As máquinas elétricas rotativas são hoje a principal forma de transformação da energia elétrica em mecânica motriz e os motores de indução trifásicos têm grande relevância dentro do setor produtivo. A garantia de um correto funcionamento torna-se vital para eficácia e competitividade da empresa dentro do setor fabril. Assim sendo, um correto diagnóstico e classificação de falhas de funcionamento dos motores em operação pode fornecer maior segurança no processo de tomada de decisão sobre a manutenção, aumentar a produtividade e eliminar os riscos e os danos aos processos como um todo. A proposição deste trabalho baseia-se na análise das correntes de estator no domínio da frequência com sinais injetados na máquina juntamente com a modulação de alimentação para o diagnóstico do motor sem defeitos, com falhas de curtocircuito nos enrolamentos do estator e com falhas de rotor. A proposta é validada numa ampla faixa de frequências de operação bem como de regimes de conjugado de carga. São analisados os desempenhos individuais de cinco técnicas de classificadores de padrões, sendo proposta a utilização de: i) Perceptron Multicamadas, ii) Máquina de Vetores de Suporte, iii) k-Vizinhos Próximos, iv) Árvore de Decisão C 4.5 e v) Naive Bayes. Complementarmente, é desenvolvido um comparativo dos métodos de classificação de padrões para avaliar a precisão de classificação frente aos diversos níveis de severidade das falhas. Resultados experimentais com motor de 1 cv são apresentados para validar a proposta. / Rotating electric machines are today the main form of transformation of electrical energy in motor mechanics and three-phase induction motors have great relevance within the productive sector. Thus a correct diagnosis and classification of failures of the engines in operation can provide security in the decision making process on maintenance, increase productivity and eliminate risks and damages to processes as a whole. The purpose of this paper is based on the analysis of the stator currents in the frequency domain with signals injected into the machine together with the power modulation for the diagnosis of motor faultless, stator winding short-circuit faults and rotor faults. Considering also, for validation of the proposal is validated on a broad range frequency of operation as well as load torque. We analyze the individual performances of five standard classifier techniques, proposing the use of: i) Multilayers Perceptron, ii) Support Vector Machine, iii) k-Nearest Neighbor, iv) C 4.5 Decision Tree and v) Naive Bayes. Complementarily, a comparison of the methods of classification of standards is developed to evaluate the accuracy of classification against the different levels of severity of the failures. Experimental results with 735.5 w and 1.471 w engines are presented to validate the proposal.

With emergence of biomedical informatics, Web intelligence, and E-business, new challenges are coming for knowledge discovery and data mining modeling problems. In this dissertation work, a framework named Granular Support Vector Machines (GSVM) is proposed to systematically and formally combine statistical learning theory, granular computing theory and soft computing theory to address challenging predictive data modeling problems effectively and/or efficiently, with specific focus on binary classification problems. In general, GSVM works in 3 steps. Step 1 is granulation to build a sequence of information granules from the original dataset or from the original feature space. Step 2 is modeling Support Vector Machines (SVM) in some of these information granules when necessary. Finally, step 3 is aggregation to consolidate information in these granules at suitable abstract level. A good granulation method to find suitable granules is crucial for modeling a good GSVM. Under this framework, many different granulation algorithms including the GSVM-CMW (cumulative margin width) algorithm, the GSVM-AR (association rule mining) algorithm, a family of GSVM-RFE (recursive feature elimination) algorithms, the GSVM-DC (data cleaning) algorithm and the GSVM-RU (repetitive undersampling) algorithm are designed for binary classification problems with different characteristics. The empirical studies in biomedical domain and many other application domains demonstrate that the framework is promising. As a preliminary step, this dissertation work will be extended in the future to build a Granular Computing based Predictive Data Modeling framework (GrC-PDM) with which we can create hybrid adaptive intelligent data mining systems for high quality prediction.

Due to complexity of biomedical problems, adaptive and intelligent knowledge discovery and data mining systems are highly needed to help humans to understand the inherent mechanism of diseases. For biomedical classification problems, typically it is impossible to build a perfect classifier with 100% prediction accuracy. Hence a more realistic target is to build an effective Decision Support System (DSS). In this dissertation, a novel adaptive Fuzzy Association Rules (FARs) mining algorithm, named FARM-DS, is proposed to build such a DSS for binary classification problems in the biomedical domain. Empirical studies show that FARM-DS is competitive to state-of-the-art classifiers in terms of prediction accuracy. More importantly, FARs can provide strong decision support on disease diagnoses due to their easy interpretability. This dissertation also proposes a fuzzy-granular method to select informative and discriminative genes from huge microarray gene expression data. With fuzzy granulation, information loss in the process of gene selection is decreased. As a result, more informative genes for cancer classification are selected and more accurate classifiers can be modeled. Empirical studies show that the proposed method is more accurate than traditional algorithms for cancer classification. And hence we expect that genes being selected can be more helpful for further biological studies.

Computational intelligence techniques, such as artificial neural networks (ANNs), have been widely used to improve the performance of power system monitoring and control. Although inspired by the neurons in the brain, ANNs are largely different from living neuron networks (LNNs) in many aspects. Due to the oversimplification, the huge computational potential of LNNs cannot be realized by ANNs. Therefore, a more brain-like artificial neural network is highly desired to bridge the gap between ANNs and LNNs. The focus of this research is to develop a biologically inspired artificial neural network (BIANN), which is not only biologically meaningful, but also computationally powerful. The BIANN can serve as a novel computational intelligence tool in monitoring, modeling and control of the power systems. A comprehensive survey of ANNs applications in power system is presented. It is shown that novel types of reservoir-computing-based ANNs, such as echo state networks (ESNs) and liquid state machines (LSMs), have stronger modeling capability than conventional ANNs. The feasibility of using ESNs as modeling and control tools is further investigated in two specific power system applications, namely, power system nonlinear load modeling for true load harmonic prediction and the closed-loop control of active filters for power quality assessment and enhancement. It is shown that in both applications, ESNs are capable of providing satisfactory performances with low computational requirements. A novel, more brain-like artificial neural network, i.e. biologically inspired artificial neural network (BIANN), is proposed in this dissertation to bridge the gap between ANNs and LNNs and provide a novel tool for monitoring and control in power systems. A comprehensive survey of the spiking models of living neurons as well as the coding approaches is presented to review the state-of-the-art in BIANN research. The proposed BIANNs are based on spiking models of living neurons with adoption of reservoir-computing approaches. It is shown that the proposed BIANNs have strong modeling capability and low computational requirements, which makes it a perfect candidate for online monitoring and control applications in power systems. BIANN-based modeling and control techniques are also proposed for power system applications. The proposed modeling and control schemes are validated for the modeling and control of a generator in a single-machine infinite-bus system under various operating conditions and disturbances. It is shown that the proposed BIANN-based technique can provide better control of the power system to enhance its reliability and tolerance to disturbances. To sum up, a novel, more brain-like artificial neural network, i.e. biologically inspired artificial neural network (BIANN), is proposed in this dissertation to bridge the gap between ANNs and LNNs and provide a novel tool for monitoring and control in power systems. It is clearly shown that the proposed BIANN-based modeling and control schemes can provide faster and more accurate control for power system applications. The conclusions, the recommendations for future research, as well as the major contributions of this research are presented at the end.

Στη συγκεκριμένη διπλωματική εργασία αναλύεται το πρόβλημα του εντοπισμού ομάδων σε σύνολα δεδομένων (ομαδοποίηση δεδομένων). Δίνεται μια σύντομη ανασκόπηση των μεθόδων που χρησιμοποιούνται σήμερα στην ομαδοποίηση δεδομένων και ιδιαίτερα στην ολοένα και αυξανόμενη χρήση Εξελικτικών Αλγόριθμων (ΕΑ) στην ομαδοποίηση. Οι ΕΑ έχουν αποδειχθεί ιδιαίτερα αποτελεσματικοί σε μια πληθώρα προβλημάτων βελτιστοποίησης. Η χρήση ΕΑ είναι αναμενόμενη, καθώς η ομαδοποίηση δεδομένων μπορεί να εκφραστεί και ως πρόβλημα βελτιστοποίησης. Επιπρόσθετα, παρουσιάζεται μια μέθοδος αντιμετώπισης της (συνήθως) μεγάλης διάστασης των προβλημάτων ομαδοποίησης, κάτι που επιβαρύνει ιδιαίτερα τους ΕΑ. Αναλυτικότερα, το πρώτο μέρος της διπλωματικής εργασίας παρέχει μια σφαιρική εικόνα του προβλήματος της ομαδοποίησης καθώς και των κατηγοριών των αλγορίθμων, που έχουν προταθεί για τον εντοπισμό ομάδων. Επιπλέον, παρουσιάζονται δομές δεδομένων που χρησιμοποιούνται από αλγόριθμους ομαδοποίησης για την επιτάχυνσή τους, όπως είναι τα Range Trees και τα BBD Trees. Εν συνεχεία, παρουσιάζονται αναλυτικά οι ΕΑ και ο τρόπος εφαρμογής τους σε προβλήματα ομαδοποίησης δεδομένων, αναλύοντας τρόπους αναπαράστασης του προβλήματος ομαδοποίησης, έτσι ώστε να είναι δυνατή η χρήση ΕΑ καθώς επίσης και οι μορφές των αντικειμενικών συναρτήσεων. Εισάγεται μια νέα προσέγγιση της εφαρμογής των ΕΑ σε προβλήματα ομαδοποίησης με σκοπό την πλήρη αποδέσμευση της διαδικασίας από εκτιμήσεις του πλήθους των ομάδων. Η διπλωματική εργασία κλείνει με τη σύγκριση υπάρχοντων αλγορίθμων ομαδοποίησης, που εφαρμόζουν την καθιερωμένη προσέγγιση της εφαρμογής των ΕΑ σε προβλήματα ομαδοποίησης, ένα νέο τρόπο εφαρμογής των ΕΑ, καθώς και κλασικούς αλγόριθμους όπως ο k-means και ο DBSCAN. Η σύγκριση γίνεται σε τεχνητά σύνολα δεδομένων, το κάθε ένα με διαφορετικές ιδιαιτερότητες. / In this master thesis, the problem of finding groups in data sets (data clustering) is analyzed. Data clustering methods in general and, more specifically, Evolutionary Algorithms (EA) based methods are shortly reviewed. EA's have proven to be effective in a extensive number of optimization problems. Since data clustering can be formulated as an optimization problem, EA can be utilized. Additionally, a method of reducing the (usually) large dimensionality of clustering problems is presented, since this hinders the performance and stability of EAs. The first part of this thesis provides an introduction to clustering as well as to existing clustering algorithms. Additionally, data structures used by clustering algorithms such as Range trees and BBD trees are described. After that, EA is described thoroughly as well as approaches of applying them on clustering problems, by analyzing forms of presenting a clustering problem in a way than an EA can be used, as well as and possible objective functions. A new approach of applying EAs on clustering problems is introduced, in an attempt to automatically determine the number of clusters present in a data set. Finally, an existing EA-based method and well known clustering algorithms such as k-means and DBSCAN are compared to the proposed approach. This comparison is made on artificial data sets, each one with its own characteristics.

Many methods have been developed and studied to detect damage through the change of dynamic response of a structure. Due to its capability to recognize pattern and to correlate non-linear and non-unique problem, Artificial Neural Networks (ANN) have received increasing attention for use in detecting damage in structures based on vibration modal parameters. Most successful works reported in the application of ANN for damage detection are limited to numerical examples and small controlled experimental examples only. This is because of the two main constraints for its practical application in detecting damage in real structures. They are: 1) the inevitable existence of uncertainties in vibration measurement data and finite element modeling of the structure, which may lead to erroneous prediction of structural conditions; and 2) enormous computational effort required to reliably train an ANN model when it involves structures with many degrees of freedom. Therefore, most applications of ANN in damage detection are limited to structure systems with a small number of degrees of freedom and quite significant damage levels. In this thesis, a probabilistic ANN model is proposed to include into consideration the uncertainties in finite element model and measured data. Rossenblueth's point estimate method is used to reduce the calculations in training and testing the probabilistic ANN model. The accuracy of the probabilistic model is verified by Monte Carlo simulations. Using the probabilistic ANN model, the statistics of the stiffness parameters can be predicted which are used to calculate the probability of damage existence (PDE) in each structural member. The reliability and efficiency of this method is demonstrated using both numerical and experimental examples. In addition, a parametric study is carried out to investigate the sensitivity of the proposed method to different damage levels and to different uncertainty levels. As an ANN model requires enormous computational effort in training the ANN model when the number of degrees of freedom is relatively large, a substructuring approach employing multi-stage ANN is proposed to tackle the problem. Through this method, a structure is divided to several substructures and each substructure is assessed separately with independently trained ANN model for the substructure. Once the damaged substructures are identified, second-stage ANN models are trained for these substructures to identify the damage locations and severities of the structural ii element in the substructures. Both the numerical and experimental examples are used to demonstrate the probabilistic multi-stage ANN methods. It is found that this substructuring ANN approach greatly reduces the computational effort while increasing the damage detectability because fine element mesh can be used. It is also found that the probabilistic model gives better damage identification than the deterministic approach. A sensitivity analysis is also conducted to investigate the effect of substructure size, support condition and different uncertainty levels on the damage detectability of the proposed method. The results demonstrated that the detectibility level of the proposed method is independent of the structure type, but dependent on the boundary condition, substructure size and uncertainty level.

Thesis (M.Eng. (Computer Enginnering)) -- University of Pretoria, 2007. / Includes bibliographical references (leaves 105-109)

Thesis (Ph. D.)--University of Washington, 2007. / Vita. Includes bibliographical references (leaves 77-82).

Σκοπός της παρούσας διπλωματικής εργασίας είναι η ταξινόμηση κλινικών περιπτώσεων κοιλιακών αλγών και συγκεκριμένα περιπτώσεων σκωληκοειδίτιδας σε παιδιά ηλικίας μέχρι 14 ετών μέσω ενός εργαλείου που υλοποιούμε. Βασικός λόγος για τη κατασκευή αυτού του εργαλείου αποτέλεσε η δυσκολία στη πρόβλεψη της ασθένειας από τους ειδικούς (κατά μέσο όρο γίνονται 20% - 30% αχρείαστες εγχειρήσεις), η συχνή σύγχυσή της με άλλες περιπτώσεις κοιλιακών αλγών ενώ το ποσοστό θνησιμότητας στα παιδιά με σκωληκοειδίτιδα ποικίλλει από 0,1% - 1%. Βασισμένοι σε ένα σύνολο δεδομένων από τη Παιδοχειρουργική Κλινική του Πανεπιστημιακού Νοσοκομείου της Αλεξανδρούπολης, διεξάγουμε αναζήτηση των καλύτερων παραμέτρων για τη κατασκευή μοντέλων ταξινομητών βασισμένων στις τρεις παρακάτω τεχνικές Υπολογιστικής Νοημοσύνης: α) τα Τεχνητά Νευρωνικά Δίκτυα, β) τις Μηχανές Διανυσμάτων Υποστήριξης και γ) τα Τυχαία Δάση. Χρησιμοποιώντας ένα σύνολο 14 κλινικών και εργαστηριακών παραγόντων, υλοποιούμε μοντέλα ταξινομητών. Η βασική ιδέα για την υλοποίηση τους είναι η αντιμετώπιση των παρακάτω προβλημάτων: : α) έχει το παιδί σκωληκοειδίτιδα ή όχι; β) Αν έχει σκωληκοειδίτιδα, ποιος τρόπος αντιμετώπισής της ενδείκνυται: χειρουργική επέμβαση ή συντηρητική αγωγή; Μετά την εύρεση των βέλτιστων μοντέλων από κάθε μία από τις μεθόδους Υπολογιστικής Νοημοσύνης που χρησιμοποιήθηκαν, υλοποιήθηκε ένα εργαλείο εύχρηστης διεπαφής χρήστη στο προγραμματιστικό περιβάλλον της Matlab 2012a το οποίο ευελπιστούμε ότι θα υποβοηθήσει τους ειδικούς στη λήψη απόφασης για τη πορεία ενός νεαρού ασθενούς που εισέρχεται στο νοσοκομείο παραπονούμενος για σκωληκοειδίτιδα. Το εργαλείο αυτό ελέγχθηκε με καινούργια πραγματικά κλινικά δεδομένα από το Καραμανδάνειο Νοσοκομείο Παίδων Πατρών και η απόδοσή του ήταν ενθαρρυντική. / The purpose of this paper is the classification of clinical cases of abdominal pain and, to be more precise, the prediction of cases with acute appendicitis at children aged up to 14 years old through a tool that we implement. The main reasons for the construction of this tool are: a) the difficulty in the prediction of the appendicitis since the 20%-30% of the operations made from the experts for this disease are gratuitous, b) the frequent confusion that there is with other diseases that cause abdominal pain and c) the mortality rate at children with appendicitis varies from 0,1% to 1%. Based on a data set from the Department of the Child Surgery of the Hospital of the University of Alexandroupolis, we conduct a search of the best parameters for the construction of model classifiers based on the three following techniques of the Computational Intelligence: a) the Artificial Neural Networks, b) the Support Vector Machines and c) the Random Forests. The basic idea for the implementation of these models is, based on a sum of 14 clinical and laboratory factors, facing the following questions: a) if a child has appendicitis or not?, b) and if it does have appendicitis, which way should we follow to cure it: operational surgery or medication? After finding these best models, we implement a tool which is actually a Graphical User Interface of Matlab 2012a which we hope that will assist the experts in making the correct decision about a young patient that goes to the hospital complaining for appendicitis. This tool was tested on new real clinical data of patients of the Child Hospital of Patras and its performance was found really encouraging.

Made available in DSpace on 2014-12-17T15:47:47Z (GMT). No. of bitstreams: 1 JoseFVS.pdf: 404321 bytes, checksum: 4e0ffed231c4c23b63bb8f6830619c82 (MD5) Previous issue date: 2007-11-30 / This work presents JFLoat, a software implementation of IEEE-754 standard for binary floating point arithmetic. JFloat was built to provide some features not implemented in Java, specifically directed rounding support. That feature is important for Java-XSC, a project developed in this Department. Also, Java programs should have same portability when using floating point operations, mainly because IEEE-754 specifies that programs should have exactly same behavior on every configuration. However, it was noted that programs using Java native floating point types may be machine and operating system dependent. Also, JFloat is a possible solution to that problem / Este trabalho apresenta JFloat, uma implementa??o de software do padr?o IEEE-754 de aritm?tica de ponto flutuante bin?ria. JFloat foi constru?da para prover algumas caracter?sticas n?o implementadas em Java, especificamente o suporte ao arredondamento direcionado. Esta caracter?stica ? de grande import?ncia para o prosseguimento do projeto Java-XSC, em desenvolvimento por esta linha de pesquisa. Apesar de programas escritos em Java, a princ?pio, serem port?veis para qualquer arquitetura ao usar opera??es de ponto flutuante, principalmente porque IEEE-754 especifica que programas deveriam ter precisamente o mesmo comportamento em toda configura??o, observou-se que programas que usam tipos de ponto flutuantes nativos de Java podem ser dependentes da m?quina e do sistema operacional. JFloat tamb?m se apresenta como uma poss?vel solu??o para este problema