In recent years, Condition Monitoring (CM), which can be performed via several sensor channels, has been recognized as an effective paradigm for failure prevention of operational equipment or processes. However, the complexity caused by asynchronous data collection with different and/or time-varying sampling/transmission rates has long been a hindrance in the effective use of multichannel data in constructing empirical models. The problem becomes more challenging when sensor readings are incomplete. Traditional sensor data recovery techniques are often prohibited in asynchronous CM environments, not to mention sparse datasets. The proposed Functional Principal Component Analysis (FPCA) methodologies, e.g., nonparametric FPC model and semi-parametric functional regression model, provide new sensor data recovery techniques to improve the reliability and robustness of multichannel CM systems. Based on the FPCA results obtained from historical asynchronous data, the deviation from the smoothing trajectory of each sensor signal can be described by a set of unit-specific model parameters. Furthermore, the relationships among these sensor signals can be identified and used to construct regression models for the correlated signals. For real-time or online implementation, use of these models along with the parameters adjusted by real-time CM data become powerful tools for dealing with asynchronous CM data while recovering lost data when needed. To improve the robustness and predictability in dealing with asynchronous data, which may be skewed in probability distribution, robust methods were developed based on Functional Data Analysis (FDA) and Local Quantile Regression (LQR) models. Case studies examining turbofan aircraft engines and an experimental two-tank flow-control loop are used to demonstrate the effectiveness and adaptability of the proposed sensor data recovery techniques. The proposed methods may also find a variety of applications in systems of other industries, such as nuclear power plants, wind turbines, railway systems, economic fields, etc., which may face asynchronous sampling and/or missing data collection problems.

Induction motors are reliable and widely used in industrialised nations. However induction motors, like any other machine, will eventually fail. If the failure is not anticipated, it can result in a significant revenue loss. Therefore, there is a strong need to develop an efficient maintenance program. The most cost-effective solution is condition-based maintenance. An effective condition-based maintenance program requires an on-line condition monitoring system that can diagnose the condition of an induction motor in order to determine the types of faults and their severity while the motor is under a normal operating condition. The work in this thesis investigates the detection of stator and rotor faults (i.e. shorted turn faults, eccentricity faults, and broken rotor bar faults) using three types of sensor signals (i.e. current, leakage flux, and vibration) under different loading conditions. The work is based on an extensive series of sensor measurements taken using a number of nominally identical healthy machines (2.2 kW) and custom-modified machines (2.2 kW) with configurable stator and rotor fault settings. The thesis starts by investigating the estimation of rotor speed and rotor slot number. These two parameters are important in determining the fault frequency components that are used for detecting the stator and rotor faults. The rotor speed investigation compares four different estimation methods from the three different sensor signal types. It is found that the speed estimation techniques based on the eccentricity harmonics and the rotor frequency in the stator current, the axial leakage flux, and the motor vibration sensor signals can detect the rotor speed very accurately even when the load is as low as 2%. Similarly, this thesis proposes three different rotor slot number estimation techniques from the three different types of sensors and demonstrates that all three techniques can estimate the rotor slot number accurately. In addition, it is shown that the reliability of the estimation techniques can be increased significantly when the three techniques are combined. The shorted turn investigation in this thesis examines and compares potential shorted turn features in the three sensor signal types under five different fault severities and ten different loading conditions. The useful shorted turn features are identified in the thesis, and then examined against variations between the healthy machines in order to determine the loads and the fault severities in which the feature can reliably detect the faults. The results show that the feature based on the EPVA (extended Park’s vector approach) is the best method. This feature can detect turn to turn faults with a severity of 3.5% or greater at loads greater than 20% and phase to phase turn faults with a severity of 1.7% or greater under all loading conditions. However, estimating the fault severity is generally found to be difficult. The thesis also examines the feasibility of detecting static eccentricity faults using the different types of sensor signals under ten different loading conditions. The thesis compares potential eccentricity features under nine different fault severities. The useful features are identified and then combined through weighted linear combination (WLC) in order to produce a better eccentricity fault indicator. The indicator begins to show significant magnitude variation when the fault severity is greater than or equal to 25% and the load is greater than or equal to 25%. The experimental results show that detecting the static eccentricity faults is possible but estimating the fault severity may be difficult. Furthermore, the effects of misalignment faults on the useful eccentricity features are investigated. In this thesis, the analysis of broken rotor bar faults is performed under motor starting and rundown operation. The starting analysis introduces a new approach to detect broken rotor bar faults that utilises the wavelet transform of the envelope of the starting current waveform. The results of the wavelet transform are then processed in order to develop a normalised parameter, called the wavelet indicator. It is found that the wavelet indicator can detect a single broken bar under all loading conditions during motor starting operation. The indicator also increases its magnitude as the severity of the fault increases. On the other hand, the rundown analysis proposes several broken rotor bar fault detection techniques which utilise the induced voltage in the stator windings and the stator magnetic flux linkage after supply disconnection. The experimental results show that detecting the faults during rundown is generally difficult. However, the wavelet approach, which is based on monitoring changes in the motor torque for a given slip, seems to give the best result. / Thesis (Ph.D.) -- University of Adelaide, School of Electrical and Electronic Engineering, 2008

In this thesis two different problems in distributed sensor networks are considered. Part I involves optimal quantiser design for decentralised estimation of a two-state hidden Markov model with dual sensors. The notion of optimality for quantiser design is based on minimising the probability of error in estimating the hidden Markov state. Equations for the filter error are derived for the continuous (unquantised) sensor outputs (signals), which are used to benchmark the performance of the quantisers. Minimising the probability of filter error to obtain the quantiser breakpoints is a difficult problem therefore an alternative method is employed. The quantiser breakpoints are obtained by maximising the mutual information between the quantised signals and the hidden Markov state. This method is known to work well for the single sensor case. Cases with independent and correlated noise across the signals are considered. The method is then applied to Markov processes with Gaussian signal noise, and further investigated through simulation studies. Simulations involving both independent and correlated noise across the sensors are performed and a number of interesting new theoretical results are obtained, particularly in the case of correlated noise. In Part II, the focus shifts to the detection of faults in helicopter transmission systems. The aim of the investigation is to determine whether the acoustic signature can be used for fault detection and diagnosis. To investigate this, statistical change detection algorithms are applied to acoustic vibration data obtained from the main rotor gearbox of a Bell 206 helicopter, which is run at high load under test conditions.

Ο σκοπός αυτής της διπλωματικής είναι η ανάπτυξη ενός προσαρμοστικού ελεγκτή για έναν εύκαμπτο ρομποτικό βραχίονα. Οι μετρήσεις του συστήματος θεωρούνται πως παρεμβάλλονται από θόρυβο, του οποίου τα όρια είναι γνωστά εξ’αρχής. Ένας Set Memebership εκτιμητής υπολογίζει το δυνατό set (ορθότοπο) μέσα στο οποίο βρίσκονται οι τιμές του διανύσματος των παραμέτρων. Από τις ακμές του ορθοτόπου αυτού προκύπτουν τα όρια μέσα στα οποία βρίσκονται οι παράμετροι του συστήματος, τα οποία χρησιμοποιούνται για τον υπολογισμό της αβεβαιότητας της εκτίμησης της εξόδου του συστήματος. Ο ελεγκτής καθορίζει τα κέρδη του μέσα σε μια online βελτιστοποίηση ενός κόστους, το οποίο βάζει κάποια βάρη στην προσπάθεια του ελέγχου (control effort), στην προκλημένη αβεβαιότητα στην έξοδο του συστήματος αλλά και στο σφάλμα παρακολούθησης της εξόδου με ένα σήμα αναφοράς. Μετά την εφαρμογή του ελεγκτή, ελέγχεται η ευστάθεια των οριακών κλειστών συστημάτων που προκύπτουν από την εφαρμογή κάθε πιθανού νόμου ελέγχου. Εξετάζεται επίσης η συμπεριφορά του Set Memebership εκτιμητή σε περίπτωση σφάλματος, δηλαδή στην περίπτωση που το σύστημά μας αλλάζει καθώς δουλεύει ο έλεγχος. / The development of an adaptive controller for a flexible link manipulator is the subject of this diploma thesis. The system’s measurements are assumed to be corrupted with noise of a priori known bounds. A Set Membership Identifier computes the feasible set (orthotope) within which the parameter vector resides. The orthotope’s vertices provide the parameter-vector’s bounds, which are used to compute the predicted system-output uncertainty. The controller tunes its gains through an on-line minimization of a cost that penalizes the control effort, the induced uncertainty on the system output, and the tracking error. After the application of the controller, the stability of the ‘extreme’ closed loop systems, derived from every possible control law, is checked. The behavior of the Set Membership Identifier is checked in the case where a fault occurs, which means that there is a change in our system’s structure while the controller is functioning.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The requirement of increasing the availability of plants is present in the current industrial park. The availability represents the satisfactory operation of a plant even during process failures. A system with these characteristics is defened as a system fault-tolerant. For the implementation of a fault-tolerant system is important to use automatic monitoring techniques capable of providing accurate information on the state of the process and form the basis for the recovery of the diagnosed condition. This structure functions as the main source of information for supervisory systems responsible for corrective actions to accommodate efects caused by failures. In this work, techniques for fault detection and diagnosis based on data from the plant operation and mathematical models of processes are used to monitor the state of industrial processes. The tolerant control techniques proposed in this work is based on the union of dierent strategies for process monitoring, a process and control reconfiguration and control goals adaptation to obtain new control structures. Thus, the proposed system will add features for fault tolerance by coupling fault detection and diagnosis with the structural reconfiguration of the control system. In this context, this work presents a novel fault-tolerant control scheme illustrating their successful application through case studies on PID control, optimal control and predictive control structures. / A exigência do aumento da disponibilidade de plantas industriais é cada vez mais presente no parque industrial atual. A disponibilidade representa a manutenção satisfatória da operação da planta mesmo após o aparecimento de falhas. Um sistema com essas características é definido como um sistema tolerante a falhas. Para a concretização de um sistema tolerante a falhas é importante a implementação de técnicas de monitoramento automático capazes de fornecer informações precisas sobre o estado do processo e formar a base necessária para a recuperação da condição diagnosticada. Essa estrutura de monitoramento funciona como a fonte principal de informações para sistemas de supervisão responsáveis por ações corretivas que acomodem efeitos causados por falhas. Neste trabalho, técnicas de detecção e diagnóstico de falhas baseadas em dados de operação da planta e em modelos matemáticos dos processos s~ao utilizadas para o monitoramento do estado de processos industriais. As técnicas de controle tolerante propostas neste trabalho baseiam-se na união de diferentes estratégias de monitoramento, na reconfiguração do processo ou controladores envolvidos e adaptação dos objetivos de controle para a obtenção de novas estruturas de controle. Assim, o sistema proposto adiciona características de tolerância a falhas através do acoplamento da detecção e diagnóstico de falhas a um sistema de reconfiguração estrutural do sistema de controle. Nesse contexto, este trabalho apresenta propostas de controle tolerante a falhas ilustrando com sucesso a sua aplicação através do estudo de casos em estruturas de controle PID, controle ótimo e controle preditivo. / Mestre em Engenharia Química

Made available in DSpace on 2014-12-17T14:08:33Z (GMT). No. of bitstreams: 1 JulianaPM.pdf: 4255756 bytes, checksum: 0f65b2b3a4f0afafcf55cda7d138bb36 (MD5) Previous issue date: 2009-06-29 / This master dissertation presents the study and implementation of inteligent algorithms to monitor the measurement of sensors involved in natural gas custody transfer processes. To create these algoritmhs Artificial Neural Networks are investigated because they have some particular properties, such as: learning, adaptation, prediction. A neural predictor is developed to reproduce the sensor output dynamic behavior, in such a way that its output is compared to the real sensor output. A recurrent neural network is used for this purpose, because of its ability to deal with dynamic information. The real sensor output and the estimated predictor output work as the basis for the creation of possible sensor fault detection and diagnosis strategies. Two competitive neural network architectures are investigated and their capabilities are used to classify different kinds of faults. The prediction algorithm and the fault detection classification strategies, as well as the obtained results, are presented / Esta disserta??o apresenta o estudo e implementa??o de algoritmos inteligentes para o monitoramento da medi??o de sensores envolvidos em processos de transfer?ncia de cust?dia de g?s natural. Para a cria??o destes algoritmos s?o investigadas arquiteturas de Redes Neurais Artificiais devido a caracter?sticas particulares, tais como: aprendizado, adapta??o e predi??o. Um preditor ? implementado com a finalidade de reproduzir o comportamento din?mico da sa?da de um sensor de interesse, de tal forma que sua sa?da seja comparada ? sa?da real do sensor. Uma rede recorrente ? utilizada para este fim, em virtude de sua capacidade em lidar com informa??o din?mica. A sa?da real do sensor e a sa?da estimada do preditor formam a base para a cria??o das estrat?gias de detec??o e identifica??o de poss?veis falhas. Duas arquiteturas de redes neurais competitivas s?o investigadas e suas potencialidades s?o utilizadas para classificar tipos diferentes de falhas. O algoritmo de predi??o e as estrat?gias de detec??o e classifica??o de falhas, bem como os resultados obtidos, ser?o apresentados

Ce travail de thèse s’inscrit dans le cadre d’une Convention Industrielle de Formation par la REcherche (CIFRE) ayant pour objectif la mise en place de solutions innovantes pour le diagnostic de défaillances. Il s’agit de répondre au besoin de la société Zodiac Actuation Systems afin de diagnostiquer les défaillances pouvant survenir sur leurs systèmes d’actionnement de sièges d’avion. Premièrement, le cadre ainsi que les motivations de l’étude sont exposés. Ensuite un état de l’art sur les méthodes de diagnostic de défaillances est donné. Puis la problématique de l’hybridation de ces méthodes est abordée. Ceci a permis d’adopter la méthode de classification supervisée pour le diagnostic. Ensuite, les campagnes de mesures, le processus de construction des bases de données ainsi que les différents algorithmes nécessaires pour la classification sont présentés. Une expérimentation sur la partie du dossier d’un siège d’avion est exposée et les résultats sont donnés. Afin d’améliorer les résultats obtenus, une approche de classification renforcée par des observateurs d’état est proposée et appliquée sur le dossier du siège. Ce renforcement est réalisé à l’aide des données estimées par les observateurs tout en construisant des bases de données augmentées. Trois types d’observateurs, linéaire, Takagi-Sugeno (TS) et TS à entrées inconnues (TSEI) sont employés. L’observateur TSEI apparait comme le mieux adapté à notre application. Finalement, une extension de l'approche proposée sur l’ensemble du siège d’avion est proposée. Celle-ci consiste en la mise en œuvre d’observateurs décentralisés TSEI pour chaque sous-ensemble du siège en tenant compte de leurs interconnexions. Ces derniers ont permis d’améliorer les résultats de détection de défaillances sur l’ensemble du siège d’avion. / This study was supported by Zodiac Actuation Systems within the framework of a ``CIFRE'' project which aims to design a Fault Detection and Diagnosis (FDD) approach for actuation systems of passengers seats in commercial aircrafts. First of all, the industrial context as well as the motivations of our project have been explained. Then, a state of the art on FDD methods is presented. Among them, hybridization of FDD methods can be found and seems interesting to our application. In a first step, the supervised classification method for the FDD has been considered. To do this, the process measurements and the concept of databases construction are presented. Then, different types of classification algorithms are explained. From experimental measurements, the classification results for FDD purpose on the recline of the seat are given. In a second step, an enhanced classification approach is proposed. It consists in estimating non-measurable variables by the state observers. These variables are then added, as estimated attributes, to the measured database. The aim is to enrich the knowledge used by the classifier and thus to improve the rate of FDD. Three types of state observers are considered: linear, then Takagi-Sugeno (TS) and Unknown Input Takagi-Sugeno (UITS) observers. It appears that the UITS observer-based results are more accurate for our application. Finally, the proposed FDD approach is extended to the hole of the seat by considering a decentralized approach. In this context, decentralized UITS are proposed for each segment of the seat by taking into account their interconnexions. It is shown that these decentralized observers improve the FDD results of the considered aircraft seat.

Ce travail se compose de deux parties, dans la première, deux types de méthodologies sont proposées pour garantir l'observabilité sur des systèmes non uniformément observables. Premièrement sont présentées les méthodes basées sur le grammien d'observabilité et, à continuation, les méthodes basées directement sur l'équation de l'observateur. Dans la deuxième partie, diverses techniques sont détaillées pour la détection de défauts (fuites et obstructions) dans les canalisations sous pressions. Pour cela on construit plusieurs modèles en discrétisant les équations du coup de bélier par différences finies, implicites et explicites dans le temps. Sur ces modèles des techniques sont développés en utilisant des observateurs et des algorithmes d'optimisation. Les modèles discrets ainsi que certains observateurs ont été validés par une série d'expériences effectuées dans des canalisations d'essai. Des résultats de convergence, expérimentaux et en simulation sont exposés dans ce mémoire. / This work consists of two parts, in the first one, two types of methods are proposed to ensure the observability of non-uniformly observable systems. Firstly methods based on the observability gramian are presented, and then some methods based directly on the equation of the observer. In the second part, various techniques are detailed for the detection of defaults (leaks and obstructions) in a pipeline under pressure. For that, we built several models by discretizing the water hammer equations using finite differences explicit and implicit in time. Then some techniques are developed using observers and optimization algorithms. Discrete models and some observers were validated by a series of experiments in pipelines. Convergence, experimental and simulation results are presented in this manuscript.

La surveillance automatique consiste à vérifier le bon fonctionnement d'un système tout au long de sa durée d'utilisation et ce, sans intervention humaine. Elle permet de mettre en place une stratégie de maintenance prévisionnelle qui présente un intérêt économique majeur, en particulier dans le cas de systèmes isolés comme les éoliennes construites en pleine mer. La surveillance automatique se base sur l'acquisition plus ou moins régulière de signaux pendant le fonctionnement du système surveillé. L'analyse de ces signaux doit permettre d'établir un diagnostic et de prendre une décision sur le déclenchement des opérations de maintenance. Dans cette thèse, nous proposons une méthode d'analyse générique permettant de s'adapter à n'importe quel système surveillé. La méthode se déroule en plusieurs étapes. Premièrement, chaque signal est analysé individuellement pour en extraire son contenu spectral, c'est-à-dire identifier les pics spectraux, les séries harmoniques et les bandes de modulation présents dans sa densité spectrale. Ensuite, ce contenu spectral est suivi au cours du temps pour former des trajectoires sur l'ensemble de la séquence de signaux acquis. Ces trajectoires permettent de générer des tendances qui sont le reflet de la santé du système. Enfin, les tendances sont analysées pour identifier un changement au cœur du système qui serait synonyme d'usure ou de défaut naissant. Cette méthodologie est validée sur de nombreux signaux réels provenant de la surveillance de différents systèmes mécaniques. / The automatic monitoring consists in verifying without any human intervention that a system is operating well. The monitoring allows to use a predictive maintenance strategy, which is economically interesting, especially in the case of isolated systems like off-shore wind turbines. The automatic monitoring is based on signals acquired more or less regularly while the monitored system is operating. The analysis of these signals should be sufficient to diagnose the system and to decide whether or not the maintenance operations should be done. In this thesis, we propose a generic analysis method able to adapt itself to any monitored system. This method is composed by several steps. First, each signal is analyzed individually in order to extract its spectral content, that is to identify the spectral peaks, the harmonic series and the modulation sidebands presents in the signal spectrum. Then, the spectral content is tracked through time to construct spectral trajectories in the sequence of acquired signal. These trajectories are used to generate trends which indicate the state of the system health. Finally, the trends are analyzed to identify a change in the system response which would indicate some wear or a fault in is early stage. This analysis method is validated on real world signals acquired on different mechanical systems.

Ces travaux de thèse portent sur le développement d’une nouvelle méthodologie pour l’identification de défauts de bâtiments performants et connectés afin d'aider à la garantie de performances. Nous avons dans un premier temps resitué nos travaux dans le contexte énergétique actuel en montrant le rôle majeur des bâtiments dans la réduction des consommations énergétiques. Nous avons ensuite présenté notre méthodologie en argumentant sur les techniques à utiliser avant d’effectuer un choix final. Cette méthodologie se compose de deux blocs principaux : le premier vise à réduire les incertitudes liées à l'occupant et à l'environnement et le second étudie l'écart entre la simulation et la mesure par une analyse de sensibilité couplée à un algorithme bayésien. Nous l'avons ensuite implémentée dans un outil que nous avons nommé REFATEC. Nous avons alors soumis notre méthodologie à différents tests dans des conditions idéales afin d’éprouver sa précision et son temps d’exécution. Cette étape a montré que la méthodologie est efficace mais montre quelques faiblesses dans le cas d’une saison estivale ou d’un défaut très localisé. Enfin, nous l’avons mise en situation face à un cas réel afin de traiter les nombreuses questions que soulèvent l’utilisation de mesures in-situ dans la perspective de la garantie de performances et de la détection de défauts, avec notamment la fiabilité des mesures et les incertitudes encore nombreuses qui doivent être traitées. / This thesis deals with the development of a new methodology for fault detection within smart high-performance buildings helping the performance guarantee. We first have placed our work in the current energy context by focusing on the major role of buildings in the decrease of energy consumption. Then we introduced our methodology and we argued about various techniques that could be used before making a choice. This methodology is made up of two main parts : the former reduces the uncertainties due to the occupant and the environment and the latter studies the gap between simulation and measurements thanks to a sensitivity analysis coupled with a bayesian algorithm. Then we implemented it within a tool that we named REFATEC. We carried out various tests in controlled conditions in order to evaluate its precision and its calculation time. This step showed that our methodology is effective but it has some difficulties when the studied period is during summer or when the faults are very located. is a very located fault. Eventually we confronted our methodology to a real case where we faced numerous questions that appear when dealing with measurements, especially their reliability and the uncertainties that still need to be taken care of, in the perspective of performance guarantee and fault detection.