Wireless Sensor Network routing protocols rely on the estimation of the quality of the links between nodes to determine a suitable path from the data source nodes to a data-collecting node. Several link estimators have been proposed, but most of these use only one link property. Fuzzy logic based link quality estimators have been recently proposed which consider a number of link quality metrics. The fuzzification of crisp values to fuzzy values is done through membership functions. The shape of the fuzzy link quality estimator membership functions is primarily performed leveraging qualitative knowledge and an improper assignment of fuzzy membership functions can lead to poor route selection and hence to unacceptable packet losses. This thesis evaluated the Channel Quality membership function of, an existing fuzzy link quality estimator and it was seen that this membership function didn???t perform as well as expected. This thesis presents an experimental approach to determine a suitable Channel Quality fuzzy membership function based on varying the shape of the fuzzy set for a multipath wireless sensor network scenario and choosing an optimum shape that maximizes the Packet Delivery Ratio of the network. The computed fuzzy set membership functions were evaluated against an existing fuzzy link quality estimator under more complex scenarios and it is shown the performance of the experimental refined membership function was better in terms of packet reception ratio and end to end delay.The fuzzy link quality estimator was applied in WiseRoute (a simple converge cast based routing protocol) and shown that this SNR based fuzzy link estimator performed better than the original implemented RSSI based link quality used in WiseRoute.

Wikipedia is a free, web-based, collaborative, multilingual encyclopedia project supported by the non-profit Wikimedia Foundation. Due to the free nature of Wikipedia and allowing open access to everyone to edit articles the quality of articles may be affected. As all people don’t have equal level of knowledge and also different people have different opinions about a topic so there may be difference between the contributions made by different authors. To overcome this situation it is very important to classify the articles so that the articles of good quality can be separated from the poor quality articles and should be removed from the database. The aim of this study is to classify the articles of Wikipedia into two classes class 0 (poor quality) and class 1(good quality) using the Adaptive Neuro Fuzzy Inference System (ANFIS) and data mining techniques. Two ANFIS are built using the Fuzzy Logic Toolbox [1] available in Matlab. The first ANFIS is based on the rules obtained from J48 classifier in WEKA while the other one was built by using the expert’s knowledge. The data used for this research work contains 226 article’s records taken from the German version of Wikipedia. The dataset consists of 19 inputs and one output. The data was preprocessed to remove any similar attributes. The input variables are related to the editors, contributors, length of articles and the lifecycle of articles. In the end analysis of different methods implemented in this research is made to analyze the performance of each classification method used.

No description available.

No description available.

No description available.

L’industrie aéronautique fait face à trois enjeux majeurs : la réduction de son empreinte environnementale, l’absorption de l’augmentation du trafic et le maintien d’un haut niveau de sécurité pour des systèmes de plus en plus complexes, à coûts équivalents. La maintenance prédictive permet de répondre en partie à ces trois enjeux. Les systèmes, dont on peut prédire la durée de vie peuvent être utilisés plus longtemps, ce qui diminue le nombre de composants utilisés sur la vie d’un avion. Prédire les pannes permet également d’augmenter la disponibilité des aéronefs en évitant les arrêts non planifiés. Enfin, le suivi de l’état de vieillissement de l’avion permet d’optimiser la maintenance et ainsi de réduire les coûts. Dans les années 2000, ces méthodologies ont été appliquées sur les moteurs. Elles commencent maintenant à se généraliser aux autres systèmes avioniques. Ainsi cette thèse concerne la mise au point de méthodologies amenant à la maintenance prédictive d’actionneurs électro – mécaniques (EMA) de commande de vol. Les problématiques et les contraintes (temps de calculs, quantités de mémoire…) liées à cette thématique sont détaillées. Dans un premier temps, le calcul de signatures de défauts est abordé. Une méthode pour les systèmes visécrou, basée sur l’identification d’un modèle de frottement, est proposée. Une deuxième méthode, reposant sur l’analyse des courants à partir d’une combinaison de décomposition modale empirique ensembliste complète et d’analyse aveugle de sources, est ensuite introduite. Ces deux méthodes sont testées sur des données issues de profils non opérationnels. Ces données sont issues d’un modèle de simulation représentant finement l’actionneur dans son environnement. L’ensemble des simulations représente des essais virtuels sur une population d’EMA. A partir de ces simulations, les signatures mises au point sont calculées. Puis, afin de valider ces signatures, des métriques de performances sont calculées. Le diagnostic par reconnaissance de formes est ensuite traité. Un algorithme reposant sur une combinaison de machine à vecteur de supports et de fonctions floues d’appartenances est proposé. Celui-ci peut notamment estimer la sévérité d’un défaut. Il permet également de détecter des points ne correspondant pas à la base d’apprentissage, qui peuvent représenter des défauts inconnus ou des points appartenant à plusieurs classes à la fois, pouvant représenter des cas de défauts combinés. L’architecture d’un système de diagnostic complet, basée sur l’algorithme conçu, est détaillée. Des validations expérimentales des méthodes de calcul de signatures et de diagnostic sont ensuite menées. Ces validations reposent sur trois bases issues de trois campagnes d’essais. La première repose sur des essais d’un EMA sain sur un banc représentatif. La deuxième concerne un moteur asynchrone en défaut en régime permanent. La dernière porte sur un moteur synchrone à aimants, de type aéronautique, en défaut de courtcircuit inter-spires en régime permanent. Le respect des contraintes par l’algorithme est vérifié. Enfin, des éléments pour aller vers le pronostic sont avancés. Le processus du pronostic est détaillé. Seule une partie de ce processus est traitée, sur des données issues de vieillissement de roulements. Dans un premier temps, le partitionnement de données de vieillissement pour créer des classes de sévérité de défaut est étudié. Cette tâche a amené à proposer une métrique, dite de cohérence temporelle, permettant de vérifier qu’un résultat de partitionnement satisfait aux contraintes pour le pronostic. Puis l’algorithme de classification proposé est validé sur les données partitionnées. Ceci amène à distinguer deux méthodes de validation, une approche dite diagnostic et une dite pronostic. Une méthode de normalisation, pour l’approche pronostic, est proposée. La prédiction des signatures dans le futur est ensuite traitée. Un algorithme de régression par vecteurs de support est utilisé [etc...] / The aeronautics industry is facing three major challenges: the reduction of its environmental impact, the absorption to the air traffic increase and a high level of safety for increasing complex systems, for equivalent costs. Predictive maintenance allows answering to these issues. Systems, for which the life can be predicted, can be used for a longer time. This reduces the number of components used in the lifetime of an aircraft. To predict failures also allow increasing the availability of aircrafts by avoiding unplanned downtime. Finally, monitoring the ageing of the aircraft allows to optimize maintenance and so to reduce costs. In the 2000s, these methodologies were applied to turbojets. It starts, now, to be generalized to others avionics systems. Therefore, this work deal with predictive maintenance methodologies for electromechanical actuator (EMA) for flight controls systems. Problems and constraints (computation time, memory quantities…) related to this subject are detailed. In a first part, fault feature computation methodologies are investigated. A first method is proposed for screw/nut systems. This method is based on the identification of a friction model. A second method, based on current analysis, is presented. This method uses a combination of empirical mode decomposition and independent component analysis. The two methods are tested on data from a non-operational profile. This data are from a simulation model which represents the EMA in the aircraft environment. The simulations performed represent virtual trials on a population of EMA. From these simulations, fault features are computed. Then, performances metrics are evaluated. Diagnosis by pattern recognition is then studied. An algorithm based on support vector machine and fuzzy membership functions is proposed. This algorithm can estimate the severity of a fault. It can also detect unknown observations, which can represent unknown faults or combined faults. The architecture of a global diagnosis system, based on the proposed algorithm, is detailed. Experimental validation of fault features computation and diagnosis algorithm is performed. These validations are based on three data bases. The first one is based on trials performed on a healthy EMA on a representative bench. The second consists in an induction motor at constant speed for different types of faults. The last one is from trials on a permanent magnet synchronous machine, of aeronautics type, for different kinds of short – circuit fault severities. The respect of the aeronautics constraints is verified. At last, elements for prognosis process are given. This process is detailed. Only a part of this process is treated, on a roller bearings benchmark database. First, the clustering for prognosis is studied. A metric, which allows verifying that the obtained clusters are coherent regarding time, and thus, checks the constraints for prognosis, is given. Then the proposed diagnosis algorithm is validated on the clustered data. This brought to perform two kinds of validation, a diagnosis oriented one and a prognosis oriented one. A method to normalize data for the prognosis oriented validation, based on sigmoid function, is given. The prediction of the features in the future is studied. A regression algorithm based on support vector regression is used. Finally, the diagnosis algorithm is applied to the predicted data. This allows to estimate the end of life, and so the remaining useful life for a given time. These estimations are evaluated regarding different kinds of performance metrics and regarding the constraints of the aeronautics applicative field

在知識經濟之社會，多元思維逐漸取代傳統二元邏輯的思考與分析方法。過去使用單一數值的樣本來計算中位數的方法，已漸不符現今複雜多變的智慧科技時代之需求。尤其是在具有多變性、不確定性、與訊息不完整性的財金與經濟環境下，過分強調對於數值之運算及數學假設的前提，反而更容易造成與現實環境及條件的背離、甚至是脫節。故在進行財金與經濟方面問題的研究時，利用隸屬度函數與模糊統計的分析將會是一種較為進步的測度方法。本文在此提出模糊中位數的分析理論，並將其應用於財務金融的分析測度上，期望能對複雜的財金經濟現狀提供一套更有效且精確合理的分析方法。 / In the society of economic knowledge, Multi-valued logic goes to replace binary logic gradually. In the traditional way, we usually ask the task-taker to response the answer according to the thinking of binary logic. But such kind of response is improper since the human thinking is fuzzy and uncertain. So it should be an improved measurement using membership functions and fuzzy statistics. 　　In this paper, we will propose the definition of fuzzy median, and present some of its application. According to the above theoretical contents, we give some examples, which is used frequently in financial and economic assessment. From the explanation and discussion of fuzzy median in these examples, we can recognize that fuzzy statistics is more meaningful and proper for research of finance and economics. At last, based upon the findings of this study, certain recommendations for further research are suggested.

Integrating decision analytical techniques in geographic information systems (GIS) can help remove the two primary obstacles in spatial decision making: inaccessibility to required geographic data and difficulties in synthesizing various criteria. I developed a GIS model to assist people seeking optimal residential locations. Fuzzy set theory was used to codify criteria for each factor used in evaluating residential locations, and weighted linear combination (WLC) was employed to simulate users' preferences in decision making. Three examples were used to demonstrate the applications in the study area. The results from the examples were analyzed. The model and the ArcGIS Extension can be used in other geographic areas for residential location selection, or in other applications of spatial decision making.

This thesis deals with the design of fuzzy system that can evaluate supplier of spare parts for service. The result should be applicable to a company Iveta Šťastníková - car and tire service. Primarily it should simplify operations associated with the selection of appropriate spare parts, tools and other equipment needed to operate with car service station. First, we introduce the theoretical basis for the paper, and then we go to the present state and the analysis itself. The result is a proposed solution which should correspond to the needs of the owner.

Orientador: Marcelo Carvalho Minhoto Teixeira / Resumo: Esta tese propõe projetos de controle H∞ chaveado para uma classe de sistemas não lineares incertos descritos por modelos fuzzy Takagi-Sugeno (T-S) com funções de pertinência desconhecidas. Os projetos de controle necessitam somente dos limites inferiores e superiores das não linearidades do sistema, que podem depender de parâmetros incertos da planta. Em diversas aplicações práticas, em virtude das restrições operacionais dos equipamentos, no projeto de controle é necessário considerar que a lei de controle está sujeita à saturação do atuador. Primeiramente, utilizando uma metodologia encontrada na literatura para resolver o problema da saturação do atuador, propõe-se uma lei de controle chaveada que escolhe um ganho do controlador de realimentação do vetor de estado, que pertence a um conjunto de ganhos conhecidos, que minimiza a derivada da função de Lyapunov do tipo quadrática. Este procedimento elimina a necessidade de encontrar as expressões das funções de pertinência para implementar a lei de controle, garante um desempenho H∞ ao sistema realimentado e assegura que as trajetórias do vetor de estado permanecem dentro de uma região de operação na qual o modelo fuzzy T-S é válido. Adicionalmente, adota-se uma outra metodologia para resolver o problema da saturação do atuador, que utiliza o sinal de controle para compor um vetor de estado expandido. Desta forma, os limites do sinal de controle fazem parte da região de operação na qual o sistema não linear incerto é exatame... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: This thesis proposes local H∞ switched control designs for a class of uncertain nonlinear plants described by Takagi-Sugeno (T-S) fuzzy models with unknown membership functions. The control designs require only the lower and upper bounds of the system nonlinearities and of the system linear parameters, which can depend on uncertain parameters. In practical applications, due to the operational restrictions of the equipments, in the control design, it is necessary to consider that the control law is subject to actuator saturation. First, using a methodology found in the literature to solve the actuator saturation problem, one proposes a switched control law that chooses a state-feedback controller gain, which belongs to a given set of gains, that minimizes the time derivative of a quadratic Lyapunov function. This procedure eliminates the necessity of finding the membership function expressions to implement the control law, guarantees an H∞ performance and ensures that the state trajectory remains within a region in which the T-S fuzzy model is valid. In addition, the actuator saturation problem is approached with another methodology, using the control signal to compose an extended state vector. Then, the region in which the uncertain nonlinear system is exactly represented via T-S fuzzy models is composed by the bounds of the control signal. A switched control law, which chooses an extended state-feedback controller gain, is proposed. The design conditions avoid the presence o... (Complete abstract click electronic access below) / Doutor