Remaining useful life (RUL) prediction is a problem that researchers in the prognostics and health management (PHM) community have been studying for decades. Both physics-based and data-driven methods have been investigated, and in recent years, deep learning has gained significant attention. When sufficiently large and diverse datasets are available, deep neural networks can achieve state-of-the-art performance in RUL prediction for a variety of systems. However, for end users to trust the results of these models, especially as they are integrated into safety-critical systems, RUL prediction uncertainty must be captured. This work explores an approach for estimating both epistemic and heteroscedastic aleatoric uncertainties that emerge in RUL prediction deep neural networks and demonstrates that quantifying the overall impact of these uncertainties on predictions reveal valuable insight into model performance. Additionally, a study is carried out to observe the effects of RUL truth data augmentation on perceived uncertainties in the model.

Ao longo dos últimos 20 anos, o modelo de concessão de serviço público no Brasil, tem evoluído consideravelmente, com destaque para o setor elétrico, e especialmente o segmento de transmissão de energia elétrica. O arcabouço normativo vigente tem buscado garantir a prestação adequada do serviço mediante a avaliação de desempenho do serviço prestado, por empresas públicas ou privadas, numa visão da qualidade do produto e da continuidade ou disponibilidade do serviço para o usuário. Concomitantemente, surgiu e desenvolveu-se rápida e globalmente, o conceito de sustentabilidade, no qual o desenvolvimento sustentável deve satisfazer as necessidades da geração atual, sem que se comprometam as necessidades das gerações futuras (World Comission on Environment and Development WCED), compreendendo essencialmente as três dimensões econômica, social e ambiental (conceito de Triple Bottom Line 3BL), permitindo a captura do impacto das atividades das empresas e organizações por meio de indicadores definidos para cada uma das três dimensões. Ainda, nesse contexto, desenvolveu-se as novas técnicas de planejamento estratégico e de gestão de desempenho, baseados em indicadores, com destaque para a Teoria das Partes Interessadas (Stakeholders). Adicionalmente, os diversos métodos de avaliação econômica e de demonstração contábil das empresas também realizaram um movimento de convergência internacional, principalmente, para aplicação de análise de risco e de avaliação do valor adicionado, com uso do fluxo de caixa dos benefícios esperados, incluído o período da perpetuidade ou valor residual/terminal; e de padrões contábeis internacionais (a IFRIC 12), que altera a estrutura de composição do ativo, uma vez que a norma não permite o reconhecimento da infraestrutura relacionada à concessão no ativo imobilizado das concessionárias, prática atualmente comum no Brasil. Nesse contexto, o presente trabalho propõe a formulação de um novo indicador para a prestação do serviço público de transmissão de energia elétrica, a partir das informações sistematizadas nos Procedimentos de Rede do Operador Nacional do Sistema Elétrico (ONS) referentes aos diversos indicadores de desempenho apurados para a operação e manutenção da transmissão de energia elétrica, visando facilitar o acompanhamento da gestão dos ativos físicos, quanto aos requisitos contratuais de atualidade e conservação do serviço adequado, mediante a avaliação da perda ou ganho na expectativa de vida útil dos componentes ou das funções de transmissão, a partir de uma idade determinada, ou seja, indicador de longevidade dos ativos físicos da concessão de transmissão de energia elétrica. Além disso, este mesmo indicador de longevidade pode ser usado na avaliação do valor de perpetuidade (estimativa do valor terminal) de empresas, quer pelo método tradicional de projeção dos fluxos de caixa descontado da perpetuidade ou pela simples abordagem comparativa da expectativa de vida útil ao final do período de concessão dos prestadores do serviço público. / Over the past 20 years, the model of public service concessions in Brazil, has evolved considerably, especially in the electricity sector, and the transmission segment of electricity. The existing regulatory framework has sought to ensure the proper provision of the service through the performance evaluation of the service provided by public or private, in view of product quality and continuity of service or availability to the user. Concomitantly, emerged and developed rapidly and globally, the concept of sustainability, in which \"sustainable development should meet the needs of present generation without compromising the needs of future generations\" (World Commission on Environment and Development - WCED) comprising essentially three economic, social and environmental (concept of Triple Bottom Line - 3BL), allowing you to capture the impact of corporate activities and organizations by means of indicators defined for each of the three dimensions. Still, in this context, new techniques of strategic planning and performance management were developed, based on indicators, notably the Theory of Stakeholders. Additionally, the various methods of economic evaluation and financial statement of companies have also made a movement of international convergence, especially for application of risk analysis and assessment of value added, using the cash flow of the expected benefits, including the period perpetuity or the residual value or terminal value, and international accounting standards (IFRIC 12), which alters the structure of asset composition, since the standard does not permit the recognition of infrastructure related to the granting of fixed assets in utilities, currently common practice in Brazil. In this context, this paper proposes the formulation of a new indicator for the provision of public electric power transmission, from the systematic information in the Grid Procedures of the Operador Nacional do Sistema Elétrico (ONS) for the different performance indicators established for the operation and maintenance of electric power transmission, to facilitate monitoring of the management of physical assets, the contractual requirements of current conservation and appropriate service, by assessing the loss or gain in life expectancy of the components or the functions transmission, from a certain age, ie, an indicator of longevity of the physical assets of the granting of electric power transmission. Furthermore, this same indicator of longevity can be used in assessing the value of perpetuity (estimated terminal value) of companies or by the traditional method of projection of cash flows discounted perpetuity or for the simple approach of comparative life expectancy of the end of the concession of public service providers.

Made available in DSpace on 2016-04-25T18:39:35Z (GMT). No. of bitstreams: 1 Dalgi Sequeira Santos.pdf: 459515 bytes, checksum: 067388b08d48c23b142f655159808db8 (MD5) Previous issue date: 2010-11-12 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The objective of this study is to check compliance of the depreciation rates used by the companies, in conformity with the explanatory notes of property, plant and equipment included in their annual reports, with the concept of depreciation described in the Accounting Theory. The depreciation rates suggested by the Brazilian income tax regulation were utilized to compare the depreciation rates used in the balance sheets published. Based on this objective, three questions were raised: (i) is the accounting standard effective in Brazil, as regards depreciation rate, in accordance with the Accounting Theory?; (ii) is the accounting practice followed by the companies, with regard to depreciation rate, consistent with the definitions of the standard?; and (iii) did the new accounting legislation lead to the adoption of depreciation rates different from those recommended by the Tax Authorities? In order to obtain answers to the questions above, three hypotheses were corroborated, namely: (i) I the Brazilian accounting standard is in conformity with the Accounting Theory as regards the concepts of useful life of assets; (ii) II assets submitted to different conditions and systems to explore their benefits will have various economic useful lives. However, in hypothesis III, it is expected that, according to hypotheses I and II, and considering the definition of useful life of assets given by the Accounting Theory, the depreciation rates used be determined in accordance with the useful life of the asset and, therefore, be different, depending on their use and lines of business. Accordingly, for the same asset different depreciation rates will be used in different businesses, which means that not always the depreciation suggested by the Tax Authorities can be adopted as a rule. That hypothesis was refuted, based on the empirical surveys. The research adopted three methods to obtain the necessary information and data: (i) bibliographic research for the theoretical basis; (ii) documental research for surveying the standards in effect; and (iii) empirical research for surveying the data in primary sources / O objetivo deste estudo é verificar a aderência das taxas de depreciação utilizadas pelas empresas, conforme as notas explicativas de ativo imobilizado contidas em seus relatórios anuais, com o conceito de depreciação da Teoria da Contabilidade. Para comparação das taxas de depreciação utilizadas nos balanços publicados, utilizaram-se as taxas de depreciação sugeridas pela legislação brasileira de imposto de renda. Com base neste objetivo, foram levantadas três questões: (i) a norma contábil vigente no Brasil, no tocante à taxa de depreciação, está de acordo com a Teoria Contábil?; (ii) a prática contábil observada nas empresas, no tocante à taxa de depreciação, está coerente com as definições da norma?; e (iii) a nova legislação contábil levou à adoção de taxas de depreciação diferentes das recomendadas pelo Fisco? A fim de obter as respostas às questões anteriores, três hipóteses foram corroboradas, a saber: (i) I a norma brasileira de contabilidade está alinhada com a Teoria da Contabilidade no que se refere aos conceitos de vida útil do ativo; (ii) II ativos submetidos a diferentes condições e regimes de exploração de seus benefícios terão diversas vidas útil-econômicas. No entanto, na hipótese III, espera-se que, de acordo com as hipóteses I e II e considerando a definição de vida útil do ativo dada pela Teoria da Contabilidade, as taxas de depreciação utilizadas sejam estipuladas segundo a vida útil-econômica do ativo e, dessa forma, sejam diferentes conforme seu uso e em diversos ramos de negócios. Assim, para um mesmo ativo, diferentes taxas de depreciação serão utilizadas em negócios diversos, afastando-se da depreciação sugerida pelo Fisco. A referida hipótese foi refutada com base nos levantamentos empíricos. A pesquisa foi empreendida com a utilização de três métodos para obter as informações e os dados necessários: (i) pesquisa bibliográfica para a base teórica; (ii) pesquisa documental para o levantamento das normas vigentes; e (iii) pesquisa empírica para o levantamento dos dados em fonte primária

Ao longo dos últimos 20 anos, o modelo de concessão de serviço público no Brasil, tem evoluído consideravelmente, com destaque para o setor elétrico, e especialmente o segmento de transmissão de energia elétrica. O arcabouço normativo vigente tem buscado garantir a prestação adequada do serviço mediante a avaliação de desempenho do serviço prestado, por empresas públicas ou privadas, numa visão da qualidade do produto e da continuidade ou disponibilidade do serviço para o usuário. Concomitantemente, surgiu e desenvolveu-se rápida e globalmente, o conceito de sustentabilidade, no qual o desenvolvimento sustentável deve satisfazer as necessidades da geração atual, sem que se comprometam as necessidades das gerações futuras (World Comission on Environment and Development WCED), compreendendo essencialmente as três dimensões econômica, social e ambiental (conceito de Triple Bottom Line 3BL), permitindo a captura do impacto das atividades das empresas e organizações por meio de indicadores definidos para cada uma das três dimensões. Ainda, nesse contexto, desenvolveu-se as novas técnicas de planejamento estratégico e de gestão de desempenho, baseados em indicadores, com destaque para a Teoria das Partes Interessadas (Stakeholders). Adicionalmente, os diversos métodos de avaliação econômica e de demonstração contábil das empresas também realizaram um movimento de convergência internacional, principalmente, para aplicação de análise de risco e de avaliação do valor adicionado, com uso do fluxo de caixa dos benefícios esperados, incluído o período da perpetuidade ou valor residual/terminal; e de padrões contábeis internacionais (a IFRIC 12), que altera a estrutura de composição do ativo, uma vez que a norma não permite o reconhecimento da infraestrutura relacionada à concessão no ativo imobilizado das concessionárias, prática atualmente comum no Brasil. Nesse contexto, o presente trabalho propõe a formulação de um novo indicador para a prestação do serviço público de transmissão de energia elétrica, a partir das informações sistematizadas nos Procedimentos de Rede do Operador Nacional do Sistema Elétrico (ONS) referentes aos diversos indicadores de desempenho apurados para a operação e manutenção da transmissão de energia elétrica, visando facilitar o acompanhamento da gestão dos ativos físicos, quanto aos requisitos contratuais de atualidade e conservação do serviço adequado, mediante a avaliação da perda ou ganho na expectativa de vida útil dos componentes ou das funções de transmissão, a partir de uma idade determinada, ou seja, indicador de longevidade dos ativos físicos da concessão de transmissão de energia elétrica. Além disso, este mesmo indicador de longevidade pode ser usado na avaliação do valor de perpetuidade (estimativa do valor terminal) de empresas, quer pelo método tradicional de projeção dos fluxos de caixa descontado da perpetuidade ou pela simples abordagem comparativa da expectativa de vida útil ao final do período de concessão dos prestadores do serviço público. / Over the past 20 years, the model of public service concessions in Brazil, has evolved considerably, especially in the electricity sector, and the transmission segment of electricity. The existing regulatory framework has sought to ensure the proper provision of the service through the performance evaluation of the service provided by public or private, in view of product quality and continuity of service or availability to the user. Concomitantly, emerged and developed rapidly and globally, the concept of sustainability, in which \"sustainable development should meet the needs of present generation without compromising the needs of future generations\" (World Commission on Environment and Development - WCED) comprising essentially three economic, social and environmental (concept of Triple Bottom Line - 3BL), allowing you to capture the impact of corporate activities and organizations by means of indicators defined for each of the three dimensions. Still, in this context, new techniques of strategic planning and performance management were developed, based on indicators, notably the Theory of Stakeholders. Additionally, the various methods of economic evaluation and financial statement of companies have also made a movement of international convergence, especially for application of risk analysis and assessment of value added, using the cash flow of the expected benefits, including the period perpetuity or the residual value or terminal value, and international accounting standards (IFRIC 12), which alters the structure of asset composition, since the standard does not permit the recognition of infrastructure related to the granting of fixed assets in utilities, currently common practice in Brazil. In this context, this paper proposes the formulation of a new indicator for the provision of public electric power transmission, from the systematic information in the Grid Procedures of the Operador Nacional do Sistema Elétrico (ONS) for the different performance indicators established for the operation and maintenance of electric power transmission, to facilitate monitoring of the management of physical assets, the contractual requirements of current conservation and appropriate service, by assessing the loss or gain in life expectancy of the components or the functions transmission, from a certain age, ie, an indicator of longevity of the physical assets of the granting of electric power transmission. Furthermore, this same indicator of longevity can be used in assessing the value of perpetuity (estimated terminal value) of companies or by the traditional method of projection of cash flows discounted perpetuity or for the simple approach of comparative life expectancy of the end of the concession of public service providers.

As a consequence of the recent deregulation in the electrical power production industry, there has been a shift in the traditional ownership of power plants and the way they are operated. To hedge their business risks, the many new private entrepreneurs enter into long-term service agreement (LTSA) with third parties for their operation and maintenance activities. As the major LTSA providers, original equipment manufacturers have invested huge amounts of money to develop preventive maintenance strategies to minimize the occurrence of costly unplanned outages resulting from failures of the equipments covered under LTSA contracts. As a matter of fact, a recent study by the Electric Power Research Institute estimates the cost benefit of preventing a failure of a General Electric 7FA or 9FA technology compressor at $10 to $20 million. Therefore, in this dissertation, a two-phase data analytics approach is proposed to use the existing monitoring gas path and vibration sensors data to first develop a proactive strategy that systematically detects and validates catastrophic failure precursors so as to avoid the failure; and secondly to estimate the residual time to failure of the unhealthy items. For the first part of this work, the time-frequency technique of the wavelet packet transforms is used to de-noise the noisy sensor data. Next, the time-series signal of each sensor is decomposed to perform a multi-resolution analysis to extract its features. After that, the probabilistic principal component analysis is applied as a data fusion technique to reduce the number of the potentially correlated multi-sensors measurement into a few uncorrelated principal components. The last step of the failure precursor detection methodology, the anomaly detection decision, is in itself a multi-stage process. The obtained principal components from the data fusion step are first combined into a one-dimensional reconstructed signal representing the overall health assessment of the monitored systems. Then, two damage indicators of the reconstructed signal are defined and monitored for defect using a statistical process control approach. Finally, the Bayesian evaluation method for hypothesis testing is applied to a computed threshold to test for deviations from the healthy band. To model the residual time to failure, the anomaly severity index and the anomaly duration index are defined as defects characteristics. Two modeling techniques are investigated for the prognostication of the survival time after an anomaly is detected: the deterministic regression approach, and parametric approximation of the non-parametric Kaplan-Meier plot estimator. It is established that the deterministic regression provides poor prediction estimation. The non parametric survival data analysis technique of the Kaplan-Meier estimator provides the empirical survivor function of the data set comprised of both non-censored and right censored data. Though powerful because no a-priori predefined lifetime distribution is made, the Kaplan-Meier result lacks the flexibility to be transplanted to other units of a given fleet. The parametric analysis of survival data is performed with two popular failure analysis distributions: the exponential distribution and the Weibull distribution. The conclusion from the parametric analysis of the Kaplan-Meier plot is that the larger the data set, the more accurate is the prognostication ability of the residual time to failure model.

La maintenance prédictive joue un rôle important dans le maintien des systèmes de production continue car elle peut aider à réduire les interventions inutiles ainsi qu'à éviter des pannes imprévues. En effet, par rapport à la maintenance conditionnelle, la maintenance prédictive met en œuvre une étape supplémentaire, appelée le pronostic. Les opérations de maintenance sont planifiées sur la base de la prédiction des états de détérioration futurs et sur l'estimation de la vie résiduelle du système. Dans le cadre du projet européen FP7 SUPREME (Sustainable PREdictive Maintenance for manufacturing Equipment en Anglais), cette thèse se concentre sur le développement des modèles de détérioration stochastiques et sur des méthodes d'estimation de la vie résiduelle (Remaining Useful Life – RUL en anglais) associées pour les adapter aux cas d'application du projet. Plus précisément, les travaux présentés dans ce manuscrit sont divisés en deux parties principales. La première donne une étude détaillée des modèles de détérioration et des méthodes d'estimation de la RUL existant dans la littérature. En analysant leurs avantages et leurs inconvénients, une adaptation d’une approche de l'état de l'art est mise en œuvre sur des cas d'études issus du projet SUPREME et avec les données acquises à partir d’un banc d'essai développé pour le projet. Certains aspects pratiques de l’implémentation, à savoir la question de l'échange d'informations entre les partenaires du projet, sont également détaillées dans cette première partie. La deuxième partie est consacrée au développement de nouveaux modèles de détérioration et les méthodes d'estimation de la RUL qui permettent d'apporter des éléments de solutions aux problèmes de modélisation de détérioration et de prédiction de RUL soulevés dans le projet SUPREME. Plus précisément, pour surmonter le problème de la coexistence de plusieurs modes de détérioration, le concept des modèles « multi-branche » est proposé. Dans le cadre de cette thèse, deux catégories des modèles de type multi-branche sont présentées correspondant aux deux grands types de modélisation de l'état de santé des système, discret ou continu. Dans le cas discret, en se basant sur des modèles markoviens, deux modèles nommés Mb-HMM and Mb-HsMM (Multi-branch Hidden (semi-)Markov Model en anglais) sont présentés. Alors que dans le cas des états continus, les systèmes linéaires à sauts markoviens (JMLS) sont mis en œuvre. Pour chaque modèle, un cadre à deux phases est implémenté pour accomplir à la fois les tâches de diagnostic et de pronostic. A travers des simulations numériques, nous montrons que les modèles de type multi-branche peuvent donner des meilleures performances pour l'estimation de la RUL par rapport à celles obtenues par des modèles standards mais « mono-branche ». / Predictive maintenance plays a crucial role in maintaining continuous production systems since it can help to reduce unnecessary intervention actions and avoid unplanned breakdowns. Indeed, compared to the widely used condition-based maintenance (CBM), the predictive maintenance implements an additional prognostics stage. The maintenance actions are then planned based on the prediction of future deterioration states and residual life of the system. In the framework of the European FP7 project SUPREME (Sustainable PREdictive Maintenance for manufacturing Equipment), this thesis concentrates on the development of stochastic deterioration models and the associated remaining useful life (RUL) estimation methods in order to be adapted in the project application cases. Specifically, the thesis research work is divided in two main parts. The first one gives a comprehensive review of the deterioration models and RUL estimation methods existing in the literature. By analyzing their advantages and disadvantages, an adaption of the state of the art approaches is then implemented for the problem considered in the SUPREME project and for the data acquired from a project's test bench. Some practical implementation aspects, such as the issue of delivering the proper RUL information to the maintenance decision module are also detailed in this part. The second part is dedicated to the development of innovative contributions beyond the state-of-the-are in order to develop enhanced deterioration models and RUL estimation methods to solve original prognostics issues raised in the SUPREME project. Specifically, to overcome the co-existence problem of several deterioration modes, the concept of the "multi-branch" models is introduced. It refers to the deterioration models consisting of different branches in which each one represent a deterioration mode. In the framework of this thesis, two multi-branch model types are presented corresponding to the discrete and continuous cases of the systems' health state. In the discrete case, the so-called Multi-branch Hidden Markov Model (Mb-HMM) and the Multi-branch Hidden semi-Markov model (Mb-HsMM) are constructed based on the Markov and semi-Markov models. Concerning the continuous health state case, the Jump Markov Linear System (JMLS) is implemented. For each model, a two-phase framework is carried out for both the diagnostics and prognostics purposes. Through numerical simulations and a case study, we show that the multi-branch models can help to take into account the co-existence problem of multiple deterioration modes, and hence give better performances in RUL estimation compared to the ones obtained by standard "single branch" models.

Cette thèse porte sur le diagnostic et le pronostic pour l’aide à la maintenance de systèmes complexes. Elle présente deux approches de diagnostic/pronostic qui permettent de générer les indicateurs utiles pour l’optimisation de la stratégie de maintenance. Plus précisément, ces approches permettent d’évaluer l’état de santé et de prédire la durée de vie résiduelle du système. Les approches présentées visent en particulier à pallier le problème d’absence d’indicateurs de dégradation. Les développements sont fondés sur l’utilisation d’observateurs, de formalisme de Modèle de Markov Caché, des méthodes d’inférences statistiques et des méthodes de prédiction de séries temporelles à base d’apprentissage afin de caractériser et prédire les modes de fonctionnement du système. Les deux approches sont illustrées sur des exemples de dégradation d’un système de régulation de niveau d’eau, d’une machine asynchrone et d’une batterie Li-Ion. / The research presented in this thesis deals of diagnosis and prognosis of complex systems. It presents two approaches that generate useful indicators for optimizing maintenance strategies. Specifically, these approaches are used to assess the level of degradation and estimate the Remaining Useful Life of the system. The aim of these approaches is to overcome for the lack of degradation indicators. The developments are based on observers, Hidden Markov Model formalism, statistical inference methods and learning-based methods in order to characterize and predict the system operating modes. To illustrate the proposed failure diagnosis/prognosis approaches, a simulated tank level control system, an induction motor and a Li-Ion battery were used.

La complexification des systèmes et la réduction du nombre de capteurs nécessitent l’élaboration de méthodes de surveillance de plus en plus efficaces. Le travail de cette thèse s’inscrit dans ce contexte et porte sur le diagnostic et le pronostic des Systèmes à Événements Discrets (SED) temporisés. Les réseaux de Petri stochastiques partiellement mesurés sont utilisés pour modéliser le système. Le modèle représente à la fois le comportement nominal et le comportement dysfonctionnel du système. Il permet aussi de représenter ses capteurs à travers une mesure partielle des transitions et des places. Notre contribution porte sur l’exploitation de l’information temporelle pour le diagnostic et le pronostic des SED. À partir d’une suite de mesures datées, les comportements du système qui expliqueraient ces mesures sont d’abord déterminés. La probabilité de ces comportements est ensuite évaluée pour fournir un diagnostic du système en termes de probabilité d’occurrence d’un défaut. Dans le cas où une faute est diagnostiquée, une approche permettant d’estimer la distribution de sa date d’occurrence est proposée. L’objectif est de donner plus de détails sur cette faute afin de mieux la caractériser. Par ailleurs, la probabilité des comportements compatibles est exploitée pour estimer l’état actuel du système. Il s’agit de déterminer les marquages compatibles avec les mesures ainsi que leurs probabilités associées. À partir de cette estimation d’état, la prise en considération des évolutions possibles du système permet d’envisager la prédiction de la faute avant son occurrence. Une estimation de la probabilité d’occurrence de la faute sur un horizon de temps futur est ainsi obtenue. Celle-ci est ensuite étendue à l’évaluation de la durée de vie résiduelle du système. Enfin, une application des différentes approches développées sur un cas d’un système de tri est proposée. / Due to the increasing complexity of systems and to the limitation of sensors number, developing monitoring methods is a main issue. This PhD thesis deals with the fault diagnosis and prognosis of timed Discrete Event Systems (DES). For that purpose, partially observed stochastic Petri nets are used to model the system. The model represents both the nominal and faulty behaviors of the system and characterizes the uncertainty on the occurrence of events as random variables with exponential distributions. It also considers partial measurements of both markings and events to represent the sensors of the system. Our main contribution is to exploit the timed information, namely the dates of the measurements for the fault diagnosis and prognosis of DES. From the proposed model and collected measurements, the behaviors of the system that are consistent with those measurements are obtained. Based on the event dates, our approach consists in evaluating the probabilities of the consistent behaviors. The probability of faults occurrences is obtained as a consequence. When a fault is detected, a method to estimate its occurrence date is proposed. From the probability of the consistent trajectories, a state estimation is deduced. The future possible behaviors of the system, from the current state, are considered in order to achieve fault prediction. This prognosis result is extended to estimate the remaining useful life as a time interval. Finally, a case study representing a sorting system is proposed to show the applicability of the developed methods.

La construction de modèles de pronostic nécessite la compréhension du processus de dégradation des composants critiques surveillés afin d’estimer correctement leurs durées de fonctionnement avant défaillance. Un processus de d´dégradation peut être modélisé en utilisant des modèles de Connaissance issus des lois de la physique. Cependant, cette approche n´nécessite des compétences Pluridisciplinaires et des moyens expérimentaux importants pour la validation des modèles générés, ce qui n’est pas toujours facile à mettre en place en pratique. Une des alternatives consiste à apprendre le modèle de dégradation à partir de données issues de capteurs installés sur le système. On parle alors d’approche guidée par des données. Dans cette thèse, nous proposons une approche de pronostic guidée par des données. Elle vise à estimer à tout instant l’état de santé du composant physique et prédire sa durée de fonctionnement avant défaillance. Cette approche repose sur deux phases, une phase hors ligne et une phase en ligne. Dans la phase hors ligne, on cherche à sélectionner, parmi l’ensemble des signaux fournis par les capteurs, ceux qui contiennent le plus d’information sur la dégradation. Cela est réalisé en utilisant un algorithme de sélection non supervisé développé dans la thèse. Ensuite, les signaux sélectionnés sont utilisés pour construire différents indicateurs de santé représentant les différents historiques de données (un historique par composant). Dans la phase en ligne, l’approche développée permet d’estimer l’état de santé du composant test en faisant appel au filtre Bayésien discret. Elle permet également de calculer la durée de fonctionnement avant défaillance du composant en utilisant le classifieur k-plus proches voisins (k-NN) et le processus de Gauss pour la régression. La durée de fonctionnement avant défaillance est alors obtenue en comparant l’indicateur de santé courant aux indicateurs de santé appris hors ligne. L’approche développée à été vérifiée sur des données expérimentales issues de la plateforme PRO-NOSTIA sur les roulements ainsi que sur des données fournies par le Prognostic Center of Excellence de la NASA sur les batteries et les turboréacteurs. / Constructing prognostics models rely upon understanding the degradation process of the monitoredcritical components to correctly estimate the remaining useful life (RUL). Traditionally, a degradationprocess is represented in the form of physical or experts models. Such models require extensiveexperimentation and verification that are not always feasible in practice. Another approach that buildsup knowledge about the system degradation over time from component sensor data is known as datadriven. Data driven models require that sufficient historical data have been collected.In this work, a two phases data driven method for RUL prediction is presented. In the offline phase, theproposed method builds on finding variables that contain information about the degradation behaviorusing unsupervised variable selection method. Different health indicators (HI) are constructed fromthe selected variables, which represent the degradation as a function of time, and saved in the offlinedatabase as reference models. In the online phase, the method estimates the degradation state usingdiscrete Bayesian filter. The method finally finds the most similar offline health indicator, to the onlineone, using k-nearest neighbors (k-NN) classifier and Gaussian process regression (GPR) to use it asa RUL estimator. The method is verified using PRONOSTIA bearing as well as battery and turbofanengine degradation data acquired from NASA data repository. The results show the effectiveness ofthe method in predicting the RUL.

Cette thèse développe des approches pour le diagnostic et le pronostic de systèmes dynamiques incertains en utilisant la technique de modélisation Bond Graph (BG). Tout d'abord, une représentation par intervalles des incertitudes paramétriques et de mesures est intégrée à un modèle BG-LFT (Linear Fractional Transformation). Une méthode de détection robuste de défaut est développée en utilisant les règles de l'arithmétique d'intervalle pour la génération de seuils robustes et adaptatifs sur les résidus nominaux. La méthode est validée en temps réel sur un système de générateur de vapeur.Deuxièmement, une nouvelle méthodologie de pronostic hybride est développée en utilisant les Relations de Redondance Analytique déduites d'un modèle BG et les Filtres Particulaires. Une estimation de l'état courant du paramètre candidat pour le pronostic est obtenue en termes probabilistes. La prédiction de la durée de vie résiduelle est atteinte en termes probabilistes. Les incertitudes associées aux mesures bruitées, les conditions environnementales, etc. sont gérées efficacement. La méthode est validée en temps réel sur un système mécatronique incertain.Enfin, la méthodologie de pronostic développée est mise en œuvre et validée pour le suivi efficace de la santé d'un sous-système électrochimique d’une pile à combustible à membrane échangeuse de protons (PEMFC) industrielle à l’aide de données de dégradation réelles. / This thesis develops the approaches for diagnostics and prognostics of uncertain dynamic systems in Bond Graph (BG) modeling framework. Firstly, properties of Interval Arithmetic (IA) and BG in Linear Fractional Transformation, are integrated for representation of parametric and measurement uncertainties on an uncertain BG model. Robust fault detection methodology is developed by utilizing the rules of IA for the generation of adaptive interval valued thresholds over the nominal residuals. The method is validated in real time on an uncertain and highly complex steam generator system.Secondly, a novel hybrid prognostic methodology is developed using BG derived Analytical Redundancy Relationships and Particle Filtering algorithms. Estimations of the current state of health of a system parameter and the associated hidden parameters are achieved in probabilistic terms. Prediction of the Remaining Useful Life (RUL) of the system parameter is also achieved in probabilistic terms. The associated uncertainties arising out of noisy measurements, environmental conditions etc. are effectively managed to produce a reliable prediction of RUL with suitable confidence bounds. The method is validated in real time on an uncertain mechatronic system.Thirdly, the prognostic methodology is validated and implemented on the electrical electro-chemical subsystem of an industrial Proton Exchange Membrane Fuel Cell. A BG of the latter is utilized which is suited for diagnostics and prognostics. The hybrid prognostic methodology is validated, involving real degradation data sets.