In-process Optical Coordinate Measuring Machine (OCMM) offers the potential of diagnosing in a timely manner variation sources that are responsible for product quality defects. Such a sensor system can help manufacturers improve product quality and reduce process downtime. Effective use of sensory data in diagnosing variation sources depends on the optimal design of a sensor system, which is often known as the problem of sensor placements. This thesis addresses coordinate sensor placement in diagnosing dimensional variation sources in assembly processes. Sensitivity indices of detecting process mean and variance components are defined as the design criteria and are derived in terms of process layout and sensor deployment information. Exchange algorithms, originally developed in the research of optimal experiment deign, are employed and revised to maximize the detection sensitivity. A sort-and-cut procedure is used, which remarkably improve the algorithm efficiency of the current exchange routine. The resulting optimal sensor layouts and its implications are illustrated in the specific context of a panel assembly process.

In-process Optical Coordinate Measuring Machine (OCMM) offers the potential of diagnosing in a timely manner variation sources that are responsible for product quality defects. Such a sensor system can help manufacturers improve product quality and reduce process downtime. Effective use of sensory data in diagnosing variation sources depends on the optimal design of a sensor system, which is often known as the problem of sensor placements. This thesis addresses coordinate sensor placement in diagnosing dimensional variation sources in assembly processes. Sensitivity indices of detecting process mean and variance components are defined as the design criteria and are derived in terms of process layout and sensor deployment information. Exchange algorithms, originally developed in the research of optimal experiment deign, are employed and revised to maximize the detection sensitivity. A sort-and-cut procedure is used, which remarkably improve the algorithm efficiency of the current exchange routine. The resulting optimal sensor layouts and its implications are illustrated in the specific context of a panel assembly process.

Significant research has been initiated recently to devise control strategies that could predict and compensate manufacturing errors using so called explicit Stream-of-Variation(SoV) models that relate process parameters in a Multistage Manufacturing Process (MMP) with product quality. This doctoral dissertation addresses several important scientific and engineering problems that will significantly advance the model-based, active control of quality in MMPs. First, we will formally introduce and study the new concept of compensability in MMPs, analogous to the concept of controllability in the traditional control theory. The compensability in an MMP is introduced as the property denoting one’s ability to compensate the errors in quality characteristics of the workpiece, given the allocation and character of measurements and controllable tooling. The notions of “within-station” and “between-station” compensability are also introduced to describe the ability to compensate upstream product errors within a given operation or between arbitrarily selected operations, respectively. The previous research also failed to concurrently utilize the historical and on-line measurements of product key characteristics for active model-based quality control. This dissertation will explore the possibilities of merging the well-known Run-to-Run (RtR) quality control methods with the model-based feed-forward process control methods. The novel method is applied to the problem of control of multi-layer overlay errors in lithography processes in semiconductor manufacturing. In this work, we first devised a multi-layer overlay model to describe the introduction and flow of overlay errors from one layer to the next, which was then used to pursue a unified approach to RtR and feedforward compensation of overlay errors in the wafer. At last, we extended the existing methodologies by considering inaccurately indentified noise characteristics in the underlying error flow model. This is also a very common situation, since noise characteristics are rarely known with absolute accuracy. We formulated the uncertainty in process noise characteristics using Linear Fractional Transformation (LFT) representation and solved the problem by deriving a robust control law that guaranties the product quality even under the worst case scenario of parametric uncertainties. Theoretical results have been evaluated and demonstrated using a linear state-space model of an actual industrial process for automotive cylinder head machining. / text

Failure diagnosis in large and complex systems is a critical task. In the realm of discrete event systems, Sampath et al. proposed a language based failure diagnosis approach. They introduced the diagnosability for discrete event systems and gave a method for testing the diagnosability by first constructing a diagnoser for the system. The complexity of this method of testing diagnosability is exponential in the number of states of the system and doubly exponential in the number of failure types. In this thesis, we give an algorithm for testing diagnosability that does not construct a diagnoser for the system, and its complexity is of 4th order in the number of states of the system and linear in the number of the failure types. In this dissertation we also study diagnosis of discrete event systems (DESs) modeled in the rule-based modeling formalism introduced in [12] to model failure-prone systems. The results have been represented in [43]. An attractive feature of rule-based model is it's compactness (size is polynomial in number of signals). A motivation for the work presented is to develop failure diagnosis techniques that are able to exploit this compactness. In this regard, we develop symbolic techniques for testing diagnosability and computing a diagnoser. Diagnosability test is shown to be an instance of 1st order temporal logic model-checking. An on-line algorithm for diagnosersynthesis is obtained by using predicates and predicate transformers. We demonstrate our approach by applying it to modeling and diagnosis of a part of the assembly-line. When the system is found to be not diagnosable, we use sensor refinement and sensor augmentation to make the system diagnosable. In this dissertation, a controller is also extracted from the maximally permissive supervisor for the purpose of implementing the control by selecting, when possible, only one controllable event from among the ones allowed by the supervisor for the assembly line in automaton models.

Depuis plusieurs années, de nombreuses recherches ont été menées autour du diagnostic. Cependant, il est impératif de se préoccuper dès la phase de conception d’un système des objectifs de diagnostic à atteindre. Aussi, de nombreux travaux se sont intéressés à analyser et à caractériser les propriétés de la diagnosticabilité d’un système. La diagnosticabilité est la propriété d’un système garantissant qu’il génère des observations permettant de détecter et discriminer les fautes en temps fini après leur occurrence.Le sujet de cette thèse porte sur les méthodes permettant d’établir les propriétés de la diagnosticabilité des systèmes à événements discrets dans le cadre distribué, sans construction du modèle global du système. Ce cadre est de première importance pour les applications réelles : systèmes naturellement distribués, systèmes trop complexes pour traiter leur modèle global, confidentialité des modèles locaux les uns par rapport aux autres. L’analyse de la diagnosticabilité de tels systèmes distribués se fonde sur des opérations de synchronisation des modèles locaux, par les observations et les communications. D’abord, nous étudions comment optimiser cette analyse de la diagnosticabilité en faisant abstraction de l’information nécessaire et suffisante à partir des objets locaux pour décider la diagnosticabilité globale. L'efficacité de l’algorithme peut être grandement améliorée par la synchronisation des objets locaux et abstraits en comparaison avec celle des objets locaux et non abstraits.Ensuite, nous proposons, dans le cadre distribué, l'algorithme de la diagnosticabilité de motifs d'événements particuliers a priori inobservables dans les systèmes. Ces motifs peuvent être simplement l’occurrence, brutale ou graduelle, d’une faute permanente ou transitoire, plusieurs occurrences d’une faute, plusieurs fautes en cascade, etc. Dans le cadre distribué, la reconnaissance du motif d’événements s’effectue d’abord progressivement dans un sous-système et ensuite la diagnosticabilité de ce motif peut être déterminée par la méthode abstraite et distribuée. Nous prouvons la correction et l'efficacité de notre algorithme à la fois en théorie et en pratique par la mise en œuvre de l’implémentation sur des exemples.Finalement, nous étudions le problème de la diagnosticabilité dans les systèmes distribués avec composants autonomes, où l’information observable est distribuée au lieu d’être centralisée comme jusqu’alors. En d'autres termes, chaque composant ne peut appréhender que ses propres événements observables. Nous donnons la définition de la diagnosticabilité conjointe. Et puis nous discutons de l'indécidabilité de diagnosticabilité conjointe dans le cas général, c'est à dire, les événements de communication ne sont pas observables, avant de proposer un algorithme pour tester sa condition suffisante. De plus, nous obtenons également un résultat de décidabilité et de l'algorithme lorsque les communications sont observables. / Over the latest decades, much research work has been done on automatic fault diagnosis. However, it is imperative to analyze at system design stage how correctness and efficiency and diagnosis algorithm can achieve. Thus many studies were interested in analyzing and characterizing the properties of diagnosability of a system. Diagnosability is the property of a system ensuring that it generates observations for detecting and discriminating faults in finite time after their occurrence.In this thesis, we investigate how to optimize distributed diagnosability analysis by abstracting necessary and sufficient information from local objects to decide global diagnosability decision. The algorithm efficiency can be greatly improved by synchronization of abstracted local objects compared to that of non abstracted local ones.Then we extend the distributed diagnosability algorithm from fault event first to simple pattern and then to general pattern, where pattern can describe more general objects in the diagnosis problem, e.g., multiple faults, multiple occurrences of the same fault, ordered occurrences of significant events, etc. In the distributed framework, the pattern recognition is first incrementally performed normally in a subsystem and then pattern diagnosability can be determined by adjusting abstracted method used in fault event case. We prove the correctness and efficiency of our proposed algorithm both in theory through proof and in practice through implementation.Finally we study joint diagnosability problem in systems with autonomous components, i.e., observable information is distributed instead of centralized. In other words, each component can only observe its own observable events. We give joint diagnosability definition. And then we discuss the undecidability of joint diagnosability in the general case, i.e., communication events are not observable, before proposing an algorithm to test its sufficient condition. In addition, we also get a decidability result and algorithm when communications are observable.

Un système technologique piloté est constitué d’une variété de composants interagissant ensemble et combinant de multiples phénomènes physiques. Ces composants, subissant un stress en fonctionnement, finissent par être affectés par des défauts pouvant avoir de graves conséquences pour l’intégrité du système lui-même ou son environnement. Ces défauts sont généralement répertoriés par des techniques de sureté de fonctionnement et classés par ordre de criticité.Une solution, pour réduire les sinistres consécutifs à la survenue d’un défaut critique, est de mettre en place un diagnostiqueur embarqué capable, dans un délai bref, de détecter cette survenue puis d’identifier le défaut, en vue de placer le système dans un mode de fonctionnement plus sûr. Il apparaît alors impératif, durant la conception du diagnostiqueur, d’étudier la diagnosticabilité des défauts, étude consistant à s’assurer que le diagnostiqueur sera toujours capable de détecter et d’identifier sans ambiguïté les défauts préalablement répertoriés.Cette thèse établit une chaîne complète de développement d’un diagnostiqueur embarqué pour les systèmes technologiques pilotés. Y sont décrites, dans un cadre théorique unifié et cohérent basé sur l’utilisation de modèles homogènes, toutes les étapes depuis la conception jusqu’à la réalisation du diagnostiqueur : représentation comportementale du système et modélisation des défauts, étude de la diagnosticabilité de ces défauts, puis génération du diagnostiqueur lui-même. Le lien rigoureux établi entre l’étude de la diagnosticabilité et la génération du diagnostiqueur, qui fonde la correction et la cohérence de l’approche, constitue un aspect saillant et original de ces travaux.Cette thèse résulte d’un projet collaboratif réunissant l’entreprise Sherpa Engineering, le Laboratoire de Recherche en Informatique (LRI) unité mixte de recherche de l'Université Paris-Sud et du CNRS et enfin le Laboratoire d’Ingénierie des Systèmes Embarqués (LISE) du CEA LIST. / A technological system is constituted with many components interacting with each other and combining multiple physical phenomena. Those components may be affected by faults resulting in serious damage to the system integrity or its environment. Those faults are generally listed by using safety analysis methodology and classified according to their severity level.A solution for reducing damages resulting from the occurrence of a critical fault is to embed a diagnosis system, which can quickly detect its occurrence and identifying the fault, this in order to put the system in the appropriate safety mode. Therefore the design process of this diagnosis system must include a diagnosability study of faults. It consists in checking that the diagnosis system will always be able to detect and identify any of the listed faults without ambiguity.This thesis introduces a complete tool-chain to develop a diagnosis system for technological systems. By using a unified theoretic and coherent framework relying on homogeneous models, all steps from the conception to the construction of the diagnosis system are described: behavioral representation of the system and faults modeling, diagnosability study of faults, generation of the diagnosis system. The rigorous link established between the diagnosability study and the associated diagnosis system generation, which ensures the coherence and correction of this approach, constitutes a salient and original aspect of this work.This thesis results from a collaborative project between the company Sherpa Engineering, the Laboratory for Computer Science (LRI) at Université Paris-Sud, joint with CNRS, and the Laboratory of Model driven engineering for embedded systems (LISE) from the CEA/LIST.

Cette thèse porte sur le diagnostic des systèmes à événements discrets modélisés par des Réseaux de Petri labellisés (RdP-L). Les problèmes de diagnostic monolithique et de diagnostic modulaire sont abordés. Des contributions sont proposées pour résoudre les problèmes d'explosion combinatoire et de complexité de calcul. Dans le cadre de l'analyse de la diagnosticabilité monolithique, certaines règles de réduction sont proposées comme un complément pour la plupart des techniques existantes de l'analyse de la diagnosticabilité, qui simplifient le modèle RdP-L tout en préservant sa propriété de diagnosticabilité. Pour un RdP-L sauf et vivant, une nouvelle condition suffisante pour la diagnosticabilité est proposée. Pour un RdR-L borné et non bloquant après l'occurrence d'une faute, l'analyse à-la-volée est améliorée en utilisant la notion d'explications minimales qui permettent de compacter l'espace d'état ; et en utilisant des T-semiflots pour trouver rapidement un cycle indéterminé. Une analyse à-la-volée utilisant Verifier Nets (VN) est proposée pour analyser à la fois les RdP-L bornés et non-bornés, ce qui permet d'obtenir un compromis entre efficacité du calcul et limitation des explosions combinatoires. Dans le cadre de l'analyse de la diagnosticabilité modulaire, une nouvelle approche est proposée pour les RdP-Ls décomposés. Les règles de réduction, qui préservent la propriété de la diagnosticabilité modulaire, sont appliquées pour simplifier le modèle initial. La diagnosticabilité locale est analysée en construisant le VN et le Graphe d'Accessibilité Modifié (MAG) du modèle local. La diagnosticabilité modulaire est vérifiée en construisant la composition parallèle du MAG et des graphes d'accessibilités d'autres modules du système. La complexité de calcul est inférieure à celles des autre approches dans la littérature. D'autre part, l'explosion combinatoire est également réduite en utilisant la technique de ε-réduction / This thesis deals with fault diagnosis of discrete event systems modeled by labeled Petri nets (LPN). The monolithic diagnosability and modular diagnosability issues are addressed. The contributions are proposed to reduce the combinatorial explosion and the computational complexity problems. Regarding monolithic diagnosability analysis, some reduction rules are proposed as a complement for most diagnosability techniques, which simplify the LPN model and preserve the diagnosability property. For a safe and live LPN, a new sufficient condition for diagnosability is proposed. For a bounded LPN that does not deadlock after a fault, the on-the-fly diagnosability analysis is improved by using minimal explanations to compact the state space; and by using T-invariants, to find quickly an indeterminate cycle. An on-the-fly diagnosability analysis using Verifier Nets (VN) is proposed to analyze both bounded and unbounded LPN, which achieves a compromise between computation efficiency and combinatorial explosion limitation. Regarding modular diagnosability analysis, a new approach is proposed for decomposed LPNs model. Reduction rules, that preserve the modular diagnosability property, are applied to simplify the model. The local diagnosability is analyzed by building the VN and the Modified Reachability Graph (MRG) of the local model. The modular diagnosability is verified by building the parallel composition of the MRG and the reachability graphs of other modules of the system. We prove in this study that the computational complexity of our approach is lower than existing approaches of literature. The combinatorial explosion is also reduced by using the ε -reduction technique.

Les systèmes hybrides sont au cœur des systèmes cyber-physiques. De tels systèmes représentent l’interaction de processus physiques continus modélisant généralement l'environnement avec des décisions discrètes issues d'un système de contrôle commande électronique. La vérification de ces systèmes est cruciale pour assurer leur sûreté dès la phase de modélisation. Les recherches sur les systèmes hybrides ont de nombreux domaines d’application, notamment le transport, l’aéronautique et la biologie. La thèse étudie des principes du raisonnement qualitatif et les applique à la vérification des systèmes hybrides. Le travail consiste à élaborer une méthode pour abstraire le système hybride en utilisant des principes qualitatifs. On recourt à une discrétisation finie de l'espace d'état tout en conservant des caractéristiques qualitatives du système. L'abstraction calculée permet de prouver des propriétés au niveau du système hybride concret et fournit une représentation du comportement global du système. Un outil développé en C++ permet de calculer l'abstraction d'un système hybride donné. Une évaluation de ses performances est établie. On s'intéresse particulièrement à une propriété de sûreté des systèmes appelée diagnosticabilité. Un modèle de système est dit diagnosticable s'il permet d'identifier sans ambiguïté la survenue de toute faute modélisée à partir des seules observations disponibles du système jusqu’à un certain délai après l’occurrence de la faute. Une méthode qui consiste à utiliser l'abstraction établie précédemment pour vérifier la diagnosticabilité d'un système hybride est proposée. / Hybrid systems are at the core of cyber-physical systems. Such systems represent the interaction between continuous physical processes generally modelling the environment with discrete decisions from control electronic signaling. The verification of these systems is crucial to ensure safety at the modeling stage. The application of hybrid systems is present in many fields such as transportation, biology and avionics. The thesis studies principals from the qualitative reasoning domain and applies them to the verification of hybrid systems. The accomplished work elaborates methods to abstract a hybrid system using qualitative principles. These methods consist in discretizing the state space to a finite number of states while conserving qualitative characteristics. The computed abstraction allows to prove properties at the level of the concrete hybrid system and presents a representation of the global behavior of the system. A tool developed in C++ computes the abstraction of a given hybrid system. An evaluation of its performance is performed. We are also interested in a particular property called diagnosability. The system is said to be diagnosable when it is capable to identify modeled faults using limited specified observations. A method that uses the computed abstraction to verify diagnosability of a given hybrid system is proposed.

While conventional test point insertions commonly used in design for testability can improve fault coverage, the test points selected may not necessarily be the best candidates to aid <em>silicon diagnosis</em>. In this thesis, test point insertions are conducted with the aim to detect more faults and also synergistically distinguish currently indistinguishable fault-pairs. We achieve this by identifying those points in the circuit, which are not only hard-to-test but also lie on distinguishable frontiers, as Testability-Diagnosability (TD) points. To this end, we propose a novel low-cost metric to identify such TD points. Further, we propose a new DFT + DFD architecture, which adds just one pin (to identify test/functional mode) and small additional combinational logic to the circuit under test. Our experiments indicate that the proposed architecture can distinguish 4x more previously indistinguishable fault-pairs than existing DFT architectures while maintaining similar fault coverages. Further, the experiments illustrate that quality results can be achieved with an area overhead of around 5%. Additional experiments conducted on hard-to-test circuits show an increase in <em>fault coverage</em> by 48% while maintaining similar diagnostic resolution. Built-in Self Test (BIST) is a technique of adding additional blocks of hardware to the circuits to allow them to perform self-testing. This enables the circuits to test themselves thereby reducing the dependency on the expensive external automated test equipment (ATE). At the end of a test session, BIST generates a signature which is a compaction of the obtained output responses of the circuit for that session. Comparison of this signature with the reference signature categorizes the circuit as error free or buggy. While BIST provides a quick and low cost alternative to check circuit's correctness, diagnosis in BIST environment remains poor because of the limited information present in the lossily compacted final signature. The signature does not give any information about the possible defect location in the circuit. To facilitate diagnosis, researchers have proposed the use of two additional on-chip embedded memories,response memory to store reference responses and fail memory to store failing responses. We propose a novel architecture in which only one additional memory is required. Experimental results conducted on benchmark circuits substantiate that the same fault coverage can be maintained using just 5% of the available test vectors. This reduces the size of memory required to store responses which in turn reduces area overhead. Further, by adding test points to the circuit using our proposed architecture, we can improve the diagnostic resolution by 60% with respect to external testing. / Master of Science

Les systèmes avioniques s'étoffent et se complexifient de plus en plus. Avec l'augmentation des capacités de calcul, de nouvelles architectures basées sur le partage de ressources émergent. Effectuer le diagnostic d'un système n'est désormais plus une opération anodine. L'enjeu actuel est donc de mettre en place des techniques de diagnostic performantes tout en optimisant les capacités de monitoring nécessaires.Ce mémoire donne une caractérisation basée modèle d'un système sous diagnostic, puis propose des techniques pour en évaluer les performances de diagnostic, ainsi que celles de son monitoring (relativement à ces performances). Le contexte industriel dans lequel s'inscrit cette thèse amène d'autres contraintes, notamment la prise en compte de la taille des systèmes avioniques à analyser. Cette thèse étudie alors l'applicabilité des techniques introduites dans ce contexte et en propose une adaptation. / Avionics systems become more and more complex. With the improvment of computing possibilities, new architectures based on resources sharing are growing up. Perform diagnosis of a system is no longer a trivial operation. The challenge is to develop efficient techniques of diagnosis while optimizing capabilities of monitoring required.This thesis give a model-based characterization of a system under diagnosis, and proposes techniques to assess diagnostic performances, as well as its monitoring ones (with respect to these diagnostic performances). The industrial context of this thesis brings other constraints, and in particular the need to handle the size of avionics systems to analyze. That thesis then examines the applicability of the introduced techniques to this particular context, and proposes an adaptation.