Automatic control techniques have been widely employed in industry to increase efficiency and profitability of the processes. However, reliability on automation increases the susceptibility of the system to faults in major control equipment such as actuators and sensors. This realization has motivated design of frameworks for fault detection and isolation (FDI) and fault tolerant control (FTC). The success of these FDI and FTC mechanisms is contingent on their ability to handle complexities associated with process systems such as nonlinearity, uncertainty, high dimensionality and the resulting effects of the existence of complexity in system structure such as faults that cannot be isolated. Motivated by the above considerations, this thesis considers the problem of fault diagnosis and fault tolerant control for complex process systems. First, an FDI framework is designed that can detect and confine possible locations for faults that cannot be isolated. Next, the problem of simultaneous actuator and sensor fault diagnosis for nonlinear uncertain systems. The key idea is to design FDI filters in a way they account for the impact of uncertainty explicitly. This work then considers the problem of simultaneous fault diagnosis in nonlinear uncertain networked systems. FDI is achieved using a distributed architecture, comprised of a bank of local FDI (LFDI) schemes that communicate with each other. The efficacy of the proposed FDI methodologies is shown via application to a number of chemical process examples. Finally, an integrated framework is proposed for fault diagnosis and fault tolerant control of variable air volume (VAV) boxes, a common component of heating, ventilation and air conditioning (HVAC) systems as an industrial case study of complex systems. The advantages of the proposed framework are diagnosing multiple faults and handling faults in stuck dampers using a safe parking strategy with energy saving capability. / Thesis / Doctor of Philosophy (PhD) / Automation is the key to increase efficiency and profitability of the processes. However, as the level of automation increases, major control equipment are more prone to faults. Thus, fault detection and isolation (FDI) and fault tolerant control (FTC) frameworks are required for fault handling. Fault handling, however, can only be efficiently achieved if the designed FDI and FTC frameworks are able to deal with complexities arising in process systems such as nonlinearity, uncertainty, high dimensionality and the resulting effects of the existence of complexity in system structure such as faults that cannot be isolated. This motivates design of FDI and FTC frameworks for complex process systems. First, FDI frameworks are presented that can diagnose faults in the presence of complexities mentioned above. Then, an integrated framework is designed for diagnosing and handling faults of heating, ventilation and air conditioning (HVAC) systems as an industrial case study of complex process systems.

The volume of data acquired in production systems continues to expand. Emerging imaging technologies, such as machine vision systems (MVSs) and 3D surface scanners, diversify the types of data being collected, further pushing data collection beyond discrete dimensional data. These large and diverse datasets increase the challenge of extracting useful information. Unfortunately, industry still relies heavily on traditional quality methods that are limited to fault detection, which fails to consider important diagnostic information needed for process recovery. Modern measurement technologies should spur the transformation of statistical process control (SPC) to provide practitioners with additional diagnostic information. This dissertation focuses on how MVSs and 3D laser scanners can be further utilized to meet that goal. More specifically, this work: 1) reviews image-based control charts while highlighting their advantages and disadvantages; 2) integrates spatiotemporal methods with digital image processing to detect process faults and estimate their location, size, and time of occurrence; and 3) shows how point cloud data (3D laser scans) can be used to detect and locate unknown faults in complex geometries. Overall, the research goal is to create new quality control tools that utilize high density data available in manufacturing environments to generate knowledge that supports decision-making beyond just indicating the existence of a process issue. This allows industrial practitioners to have a rapid process recovery once a process issue has been detected, and consequently reduce the associated downtime. / Ph. D.

This dissertation is a systematic study of artificial intelligence (AI) applications for the diagnosis of power transformer incipient fault. The AI techniques include artificial neural networks (ANN, or briefly neural networks - NN), expert systems, fuzzy systems and multivariate regression. The fault diagnosis is based on dissolved gas-in-oil analysis (DGA). A literature review showed that the conventional fault diagnosis methods, i.e. the ratio methods (Rogers, Dornenburg and IEC) and the key gas method, have limitations such as the "no decision" problem. Various AI techniques may help solve the problems and present a better solution. Based on the IEC 599 standard and industrial experiences, a knowledge-based inference engine for fault detection was developed. Using historical transformer failure data from an industrial partner, a multi-layer perceptron (MLP) modular neural network was identified as the best choice among several neural network architectures. Subsequently, the concept of a hybrid diagnosis was proposed and implemented, resulting in a combined neural network and expert system tool (the ANNEPS system) for power transformer incipient diagnosis. The abnormal condition screening process, as well as the principle and algorithms of combining the outputs of knowledge based and neural network based diagnosis, were proposed and implemented in the ANNEPS. Methods of fuzzy logic based transformer oil/paper insulation condition assessment, and estimation of oil sampling interval and maintenance recommendations, were also proposed and implemented. Several methods of power transformer incipient fault location were investigated, and a 7Ã 21Ã 5 MLP network was identified as the best choice. Several methods for on-load tap changer (OLTC) coking diagnosis were also investigated, and a MLP based modular network was identified as the best choice. Logistic regression analysis was identified as a good auditor in neural network input pattern selection processes. The above results can help developing better power transformer maintenance strategies, and serve as the basis of on-line DGA transformer monitors. / Ph. D.

Thesis (MScEng (Process Engineering))--University of Stellenbosch, 2010. / Thesis presented in partial fulfilment of the requirements for the degree of Master of Science in Engineering (Extractive Metallurgy) In the Department of Process Engineering at the University of Stellenbosch / ENGLISH ABSTRACT: Multivariate statistical process control (MSPC) approaches are now widely used for performance monitoring, fault detection and diagnosis in chemical processes. Conventional MSPC approaches are based on latent variable projection methods such as principal component analysis and partial least squares. These methods are suitable for handling linearly correlated data sets, with minimal autocorrelation in the variables. Industrial plant data invariably violate these conditions, and several extensions to conventional MSPC methodologies have been proposed to account for these limitations. In practical situations process data usually contain contributions at multiple scales because of different events occurring at different localizations in time and frequency. To account for such multiscale nature, monitoring techniques that decompose observed data at different scales are necessary. Hence the use of standard MSPC methodologies may lead to unreliable results due to false alarms and significant loss of information. In this thesis a multiscale methodology based on the use of singular spectrum analysis is proposed. Singular spectrum analysis (SSA) is a linear method that extracts information from the short and noisy time series by decomposing the data into deterministic and stochastic components without prior knowledge of the dynamics affecting the time series. These components can be classified as independent additive time series of slowly varying trend, periodic series and aperiodic noise. SSA does this decomposition by projecting the original time series onto a data-adaptive vector basis obtained from the series itself based on principal component analysis (PCA). The proposed method in this study treats each process variable as time series and the autocorrelation between the variables are explicitly accounted for. The data-adaptive nature of SSA makes the proposed method more flexible than other spectral techniques using fixed basis functions. Application of the proposed technique is demonstrated using simulated, industrial data and the Tennessee Eastman Challenge process. Also, a comparative analysis is given using the simulated and Tennessee Eastman process. It is found that in most cases the proposed method is superior in detecting process changes and faults of different magnitude accurately compared to classical statistical process control (SPC) based on latent variable methods as well as the wavelet-based multiscale SPC. / AFRIKAANSE OPSOMMING: Meerveranderlike statistiese prosesbeheerbenaderings (MSPB) word tans wydverspreid benut vir werkverrigtingkontrolering, foutopsporing en .diagnose in chemiese prosesse. Gebruiklike MSPB word op latente veranderlike projeksiemetodes soos hoofkomponentontleding en parsiele kleinste-kwadrate gebaseer. Hierdie metodes is geskik om lineer gekorreleerde datastelle, met minimale outokorrelasie, te hanteer. Nywerheidsaanlegdata oortree altyd hierdie voorwaardes, en verskeie MSPB is voorgestel om verantwoording te doen vir hierdie beperkings. Prosesdata afkomstig van praktiese toestande bevat gewoonlik bydraes by veelvuldige skale, as gevolg van verskillende gebeurtenisse wat by verskillende lokaliserings in tyd en frekwensie voorkom. Kontroleringsmetodes wat waargenome data ontbind by verskillende skale is nodig om verantwoording te doen vir sodanige multiskaalgedrag. Derhalwe kan die gebruik van standaard-MSPB weens vals alarms en beduidende verlies van inligting tot onbetroubare resultate lei. In hierdie tesis word . multiskaalmetodologie gebaseer op die gebruik van singuliere spektrumontleding (SSO) voorgestel. SSO is . lineere metode wat inligting uit die kort en ruiserige tydreeks ontrek deur die data in deterministiese en stochastiese komponente te ontbind, sonder enige voorkennis van die dinamika wat die tydreeks affekteer. Hierdie komponente kan as onafhanklike, additiewe tydreekse geklassifiseer word: stadigveranderende tendense, periodiese reekse en aperiodiese geruis. SSO vermag hierdie ontbinding deur die oorspronklike tydreeks na . data-aanpassende vektorbasis te projekteer, waar hierdie vektorbasis verkry is vanaf die tydreeks self, gebaseer op hoofkomponentontleding. Die voorgestelde metode in hierdie studie hanteer elke prosesveranderlike as . tydreeks, en die outokorrelasie tussen veranderlikes word eksplisiet in berekening gebring. Aangesien die SSO metode aanpas tot data, is die voorgestelde metode meer buigsaam as ander spektraalmetodes wat gebruik maak van vaste basisfunksies. Toepassing van die voorgestelde tegniek word getoon met gesimuleerde prosesdata en die Tennessee Eastman-proses. . Vergelykende ontleding word ook gedoen met die gesimuleerde prosesdata en die Tennessee Eastman-proses. In die meeste gevalle is dit gevind dat die voorgestelde metode beter vaar om prosesveranderings en .foute met verskillende groottes op te spoor, in vergeleke met klassieke statistiese prosesbeheer (SP) gebaseer op latente veranderlikes, asook golfie-gebaseerde multiskaal SP.

<p>Rad je posvećen problemu rane i brze detekcije i izolacije neželjenih stanja dinamičkih sistema, sa posebnim naglaskom na rano otkrivanje različitih nepravilnosti u radu i kvarova industrijskih procesa.</p> / <p>The thesys is dedicated to the problem of early and swift detection and isolation of unwanted working regimes of dynamical systems, with particular emphasis on the early detection of various irregularities and failures of industrial processes.</p>

Le test d’identification d’état d’un système à événement discret (SED) a pour but d’en identifier l’état final, lorsque son état initial est inconnu. Une solution classique à ce problème, en supposant que le SED n’ait pas de sorties observables, consiste à déterminer une séquences de synchronisation, c.à-d., une séquence d’événements d’entrée qui conduit le SED sur un état connu. Ce problème a été résolu dans les années 60’ à l’aide des automates. L’objectif principal de cette thèse est d’utiliser les réseaux de Petri (RdP) pour obtenir une résolution plus optimal de ce problème et pour une plus large classe de systèmes.Initialement, nous montrons que la méthode classique peut être aisément étendue aux RdP synchronisés. Pour cette classe de réseaux non-autonomes, toute transition est associée à un événement d’entrée.L’approche proposée est générale, dans la mesure où elle s’applique à des RdP bornés arbitraires. Cependant, elle engendre le problème d’explosion combinatoire du nombre d’états. Pour obtenir des meilleures solutions, nous considérons une classe spéciale de RdP : les graphes d’état (GdE). Pour ces réseaux, nous considérons d’abord les GdE fortement connexes et proposons des approches pour la construction de SS, qui exploitent les propriétés structurelles du réseau en évitant ainsi une énumération exhaustive de l’espace d’état. Ces résultats s’étendent aux GdE non fortement connexes et à tout RdP synchronisé composé de GdE. Enfin, nous considérons la classe des RdP non bornés et proposons des séquences qui synchronisent le marquage des places non bornées. Une boîte à outils fournit toutes les approches décrites et est appliquée à des différents bancs d’essai. / State-identification experiments are designed to identify the final state of a discrete event system (DES) when its initial state is unknown. A classical solution, assuming the DES has no observable outputs, consists in determining a synchronizing sequence (SS), i.e., a sequence of input events that drives the system to a known state. This problem was essentially solved in the 60’ using automata. The main objective of this thesis is to use Petri nets (PNs) for solving the state-identification problem more efficiently and for a wider class of systems.We start showing that the classical SS construction method based on automata can be easily applied to synchronized PNs, a class of non-autonomous nets where each transition is associated with an input event. The proposed approach is fairly general and it works for arbitrary bounded nets with a complexity that is polynomial with the size of the state space. However, it incurs in the state-space explosion problem.Looking for more efficient solutions, we begin by considering a subclass of PNs called state machines (SMs). We first consider strongly connected SMs and propose a framework for SS construction that exploits structural criteria, not requiring an exhaustive enumeration of the state space of the net. Results are further extended to larger classes of nets, namely non strongly connected SMs and nets containing SM subnets. Finally we consider the class of unbounded nets that describe infinite state systems: even in this case we are able to compute sequences to synchronize the marking of bounded places. A Matlab toolbox implementing all approaches previously described has been designed and applied to a series of benchmarks.

Sistemas Instrumentados de Segurança (SIS) são projetados para prevenir e/ou mitigar acidentes, evitando indesejáveis cenários com alto potencial de risco, assegurando a proteção da saúde das pessoas, proteção do meio ambiente e economia de custos com equipamentos industriais. Desta forma, é extremamente recomendado neste projeto de SIS o uso de métodos formais para garantir as especificações de segurança em conformidade com as normas regulamentadoras vigentes, principalmente para atingir o nível de integridade de segurança (SIL) desejado. Adicionalmente, algumas das normas de segurança como ANSI / ISA S.84.01; IEC 61508, IEC 61511, entre outras, recomendam uma série de procedimentos relacionados ao ciclo de vida de segurança de um projeto de SIS. Desta forma, destacam-se as atividades que compreendem o desenvolvimento e a validação dos algoritmos de controle em que se separam semanticamente os aspectos voltados para o diagnóstico de falhas críticas e o tratamento destas falhas associado a um controle de coordenação para filtrar a ocorrência de falhas espúrias. Portanto, a contribuição deste trabalho é propor um método formal para a modelagem e análise de SIS, incluindo o diagnóstico e o tratamento de falhas críticas, baseado em rede Bayesiana (BN) e rede de Petri (PN). Este trabalho considera o diagnóstico e o tratamento para cada função instrumentada de segurança (SIF) a partir do resultado do estudo de análise de riscos, de acordo com a metodologia de HAZOP (Hazard and Operability). / Safety Instrumented Systems (SIS) are design to prevent and/or mitigate accidents, avoiding undesirable high potential risk scenarios, assuring protection of people health, protecting the environment and saving costs of industrial equipment. It is strongly recommended in this design formal method to assure the safety specifications in accordance to standards regulations, mainly for reaching desired safety integrity level (SIL). Additionally, some of the safety standards such as ANSI/ISA S.84.01; IEC 61508, IEC 61511, among others, guide different activities related to Safety Life Cycle (SLC) design of SIS. In special, there are design activities that involves the development and validation of control algorithm that separate semantically aspects oriented to diagnosis and treatment of critical faults associated with a control coordination to filter spurious failures occurrence. In this context, the contribution of this work is to propose a formal method for modeling and analysis of SIS designed including diagnostic and treatment of critical faults based on Bayesian networks (BN) and Petri nets (PN). This approach considers diagnostic and treatment for each safety instrumented function (SIF) obtained according hazard and operability (HAZOP) methodology.

Aujourd'hui, les réseaux de communications sont largement utilisés pour relier les points de ressources, qui permettent la transmission de données à distance, de réduire la complexité dans le cadre de câblage et les coûts de support et de fournir l'aide dans la maintenance. En raison de ces avantages, les réseaux ont été introduits dans les systèmes automatiques au cours de ces dernières décennies et de nouveaux protocoles de réseau industriel ont été également développés pour assurer le contrôle à distance. Les systèmes contrôlés en réseau SCR (Networked Control System NCS) sont des systèmes automatiques traditionnels où les actionneurs, les capteurs, les contrôleurs et des autres composants sont distribués autour d'un réseau de communication, qui peut être partagé ou non avec d'autres applications. Les données de commande et de diagnostic sont échangées entre les composants du système (capteur, contrôleur, actionneur) via ce réseau partagé. Cette nouvelle architecture de système de contrôle introduit des problèmes originaux, en termes de retard variable affectant la transmission, des pertes de paquets, etc. Dans l'objectif de maintenir de bonnes performances du module de diagnostic face à des éventuelles variations introduites par le réseau, il est intéressant d'introduire des nouvelles approches. Nous avons rapporté les résultats relatifs aux techniques d'estimation optimale à base de filtre de Kalman, de façon à constituer un document aussi complet que possible traitant la génération de résidus et l'isolation des défauts dans SCR. Notre contribution consiste, dans un premier temps, à développer un modèle d'état d'un système contrôlé via un réseau. En deuxième temps, nous proposons un générateur de résidus en se basant sur les hypothèses simulant le retard induit par le réseau. Finalement, nous développons un filtre isolateur pour identifier directement les défauts affectant les actionneurs dans un SCR / Today's communications networks are widely used to connect the resources, enabling the remote data transmission, reducing the cabling complexity, minimizing costs and providing easy maintenance. Because of these advantages, the networks have been introduced in automatic systems during recent decades and new industrial network protocols were also developed for the remote control. The systems controlled by networks, the term "Networked Control Systems (NCS)" are automatic traditional systems where the actuators, the sensors, the controllers and other components are distributed around a communication network that can be shared or not with other applications. The data of both control and diagnostic are exchanged between system components (sensor, controller and actuator) via the shared network. This new architecture of control system introduces new problems in terms of variable delay affecting the transmission, the packet loss, etc. With the aim to maintain good performance of diagnostic module face of possible changes introduced by the network, it is interesting to introduce new approaches. We have reported results for the optimal estimation techniques based on Kalman filter, thus creating a report as complete as possible, treating the residual generation and fault isolation in NCS. Our contribution consists, firstly, to develop a state space model of a system controlled via a network. Secondly, we propose a residual generator based on the delay models induced by the network. Finally, we develop a isolation filter in order to directly identify in the actuators faults in a SCR

Les travaux développés dans cette thèse portent sur la "co-conception diagnostic / ordonnancement des mesures dans un système contrôlé en réseau" qui est un sujet multidisciplinaire nécessitant des compétences en théorie du contrôle et en théorie des réseaux. La thèse a pour but de développer, dans le contexte des systèmes contrôlés en réseau, une approche de co-conception qui intègre de façon coordonnée les caractéristiques qui expriment la performance du diagnostic des défauts et les paramètres de l'ordonnancement temps-réel des messages. L'intérêt de cette approche coordonnée réside essentiellement dans la minimisation des ressources nécessaires pour atteindre la performance du diagnostic requise, minimisation qui prend tout son sens dans le contexte des systèmes embarqués. Nous nous sommes intéressés plus particulièrement à l'étude des problèmes liés à l'élaboration d'algorithmes de diagnostic efficaces et adaptés aux caractéristiques de l'application de façon tout en prenant en compte différents types de contraintes liées au réseau. En conjonction avec ces algorithmes, deux ensembles de techniques d'ordonnancement des mesures ont été développés : - ordonnancement hors ligne - ordonnancement évènementiel en ligne Pour l'ordonnancement hors ligne, les séquences de communication sont conçues en amont, préalablement à la mise en oeuvre de l'algorithme de diagnostic (implémentation). D'autre part, nous proposons aussi des techniques d'ordonnancement en ligne basées sur l'échantillonnage évènementiel développé au cours de la dernière décennie. Au contraire de la plupart des recherches en automatique classique, considérant que l'échantillonnage des signaux continus est réalisé d'une manière périodique, les mesures dans cette approche sont transmises si et seulement si la condition de transmission (évènement) est vérifiée / The works developed in this thesis deal with 'fault diagnosis and sensor scheduling co-design' in networked control system. This multidisciplinary subject requires theoretical knowledge in both fault diagnosis and communication networks. Our contribution consists in developing a co-design approach that integrates in the same framework the characteristics of fault diagnosis performance and real time sensor scheduling. The main benefit of this approach is minimizing the required network resources for attending acceptable fault diagnosis performances. We are interested in the development of more efficient and more adapted for real time implementation fault diagnosis algorithms while taking into account different types of communication constraints. In conjunction with these algorithms, two sets of sensor scheduling techniques are used : - Off-line scheduling - On-line scheduling (event triggered sampling) For off-line scheduling, the communication sequences are designed before the implementation of the diagnostic algorithm. In this context, we proposed several techniques for scheduling with different spatial and temporal complexity and adapted to different operating condition for the detection and the isolation of faults based on the information provided by the selected communication sequences. Moreover, we deal also with on-line scheduling techniques based on the event triggered sampling developed during the last decade. In This approach measurement packets are transmitted if and only if the transmission condition (event) is verified. This saves resources provided by the network while maintaining acceptable performance of fault diagnosis. The objective of these algorithms is to minimize the number of transmitted information which means less energy consumption and has a major importance in wireless networked control systems

Le travail présenté dans ce mémoire de thèse porte sur la tolérance aux défauts dans le cas des systèmes linéaires. Les moyens de communication numériques sont utilisés dans le cadre de la mise en oeuvre d'une architecture de commande tolérante aux défauts pour des systèmes complexes. Une coopération entre les modules de commande/diagnostic assure la tolérance à certains types de défauts qui affectent le système. La commande des systèmes est traditionnellement réalisée à partir d'un calculateur central qui collecte l'ensemble des informations relevées sur le procédé, puis les traite pour élaborer un ensemble de commande qui est appliqué au procédé. Avec le développement des systèmes commandés en réseaux (Networked Control System) et des systèmes embarqués, l'architecture des systèmes s'oriente vers une distribution des algorithmes de commande et de diagnostic. On se propose d'aborder le problème de la conception des stratégies de distribution de diagnostic/commande et de coopération des tâches de commande entre les sous-contrôleurs associés à chaque sous-système qui composent le système complexe et de prendre en compte les défauts des actionneurs et de capteurs affectant les sous-systèmes. Il s'agit alors d'élaborer une stratégie de commande coopérative visant à compenser les effets des défauts affectant le système. Les commandes locales sont des commandes prédictives à base de modèle (MPC: Model Predictive Control). Une analyse de stabilité a été faite en prenant en considération la défaillance du réseau de communication. / The work presented in this memory of thesis focuses on fault tolerance in the case of linear systems. Digital communication tools are used in the context of the implementation of an architecture for fault tolerant control of complex systems. A cooperation between the control/diagnosis blocks ensures the tolerance to certain types of faults which affect the system. Control systems is traditionally carried out starting from a central computer that collects all information gathered on the process. Then, these information are treated in order to develop a set of command which is applied to the process. Thanks to the development of the Networked System Control and embedded systems, systems architecture is oriented towards a distributed control and diagnostic algorithms. One proposes to address the problem of designing distribution strategies for diagnosis/control and control tasks cooperation between sub-controllers associated at each subsystem comprising the complex system and to take into account the faults on the actuators and sensors that affect the subsystems. Then a cooperative control strategy is proposed. It aims at compensating the effects of the faults affecting the system. Local controls are based on Model Predictive Control (MPC). An analysis of stability was made taking into account the failure of the communication network