Many real-world systems experience deterioration with usage and age, which often leads to low product quality, high production cost, and low system availability. Most previous maintenance and reliability models in the literature do not incorporate condition monitoring information for decision making, which often results in poor failure prediction for partially observable deteriorating systems. For that reason, the development of fault prediction and control scheme using condition-based maintenance techniques has received considerable attention in recent years. This research presents a new framework for predicting failures of a partially observable deteriorating system using Bayesian control techniques. A time series model is fitted to a vector observation process representing partial information about the system state. Residuals are then calculated using the fitted model, which are indicative of system deterioration. The deterioration process is modeled as a 3-state continuous-time homogeneous Markov process. States 0 and 1 are not observable, representing healthy (good) and unhealthy (warning) system operational conditions, respectively. Only the failure state 2 is assumed to be observable. Preventive maintenance can be carried out at any sampling epoch, and corrective maintenance is carried out upon system failure. The form of the optimal control policy that maximizes the long-run expected average availability per unit time has been investigated. It has been proved that a control limit policy is optimal for decision making. The model parameters have been estimated using the Expectation Maximization (EM) algorithm. The optimal Bayesian fault prediction and control scheme, considering long-run average availability maximization along with a practical statistical constraint, has been proposed and compared with the age-based replacement policy. The optimal control limit and sampling interval are calculated in the semi-Markov decision process (SMDP) framework. Another Bayesian fault prediction and control scheme has been developed based on the average run length (ARL) criterion. Comparisons with traditional control charts are provided. Formulae for the mean residual life and the distribution function of system residual life have been derived in explicit forms as functions of a posterior probability statistic. The advantage of the Bayesian model over the well-known 2-parameter Weibull model in system residual life prediction is shown. The methodologies are illustrated using simulated data, real data obtained from the spectrometric analysis of oil samples collected from transmission units of heavy hauler trucks in the mining industry, and vibration data from a planetary gearbox machinery application.

Many real-world systems experience deterioration with usage and age, which often leads to low product quality, high production cost, and low system availability. Most previous maintenance and reliability models in the literature do not incorporate condition monitoring information for decision making, which often results in poor failure prediction for partially observable deteriorating systems. For that reason, the development of fault prediction and control scheme using condition-based maintenance techniques has received considerable attention in recent years. This research presents a new framework for predicting failures of a partially observable deteriorating system using Bayesian control techniques. A time series model is fitted to a vector observation process representing partial information about the system state. Residuals are then calculated using the fitted model, which are indicative of system deterioration. The deterioration process is modeled as a 3-state continuous-time homogeneous Markov process. States 0 and 1 are not observable, representing healthy (good) and unhealthy (warning) system operational conditions, respectively. Only the failure state 2 is assumed to be observable. Preventive maintenance can be carried out at any sampling epoch, and corrective maintenance is carried out upon system failure. The form of the optimal control policy that maximizes the long-run expected average availability per unit time has been investigated. It has been proved that a control limit policy is optimal for decision making. The model parameters have been estimated using the Expectation Maximization (EM) algorithm. The optimal Bayesian fault prediction and control scheme, considering long-run average availability maximization along with a practical statistical constraint, has been proposed and compared with the age-based replacement policy. The optimal control limit and sampling interval are calculated in the semi-Markov decision process (SMDP) framework. Another Bayesian fault prediction and control scheme has been developed based on the average run length (ARL) criterion. Comparisons with traditional control charts are provided. Formulae for the mean residual life and the distribution function of system residual life have been derived in explicit forms as functions of a posterior probability statistic. The advantage of the Bayesian model over the well-known 2-parameter Weibull model in system residual life prediction is shown. The methodologies are illustrated using simulated data, real data obtained from the spectrometric analysis of oil samples collected from transmission units of heavy hauler trucks in the mining industry, and vibration data from a planetary gearbox machinery application.

A paradigm shift is emerging in system reliability and maintainability. The military and industrial sectors are moving away from the traditional breakdown and scheduled maintenance to adopt concepts referred to as Condition Based Maintenance (CBM) and Prognostic Health Management (PHM). In addition to signal processing and subsequent diagnostic and prognostic algorithms these new technologies involve storage of large volumes of both quantitative and qualitative information to carry out maintenance tasks effectively. This not only requires research and development in advanced technologies but also the means to store, organize and access this knowledge in a timely and efficient fashion. Knowledge-based expert systems have been shown to possess capabilities to manage vast amounts of knowledge, but an intelligent systems approach calls for attributes like learning and adaptation in building autonomous decision support systems. This research presents an integrated knowledge-based approach to diagnostic reasoning for CBM of engineering systems. A two level diagnosis scheme has been conceptualized in which first a fault is hypothesized using the observational symptoms from the system and then a more specific diagnostic test is carried out using only the relevant sensor measurements to confirm the hypothesis. Utilizing the qualitative (textual) information obtained from these systems in combination with quantitative (sensory) information reduces the computational burden by carrying out a more informed testing. An Industrial Language Processing (ILP) technique has been developed for processing textual information from industrial systems. Compared to other automated methods that are computationally expensive, this technique manipulates standardized language messages by taking advantage of their semi-structured nature and domain limited vocabulary in a tractable manner. A Dynamic Case-based reasoning (DCBR) framework provides a hybrid platform for diagnostic reasoning and an integration mechanism for the operational infrastructure of an autonomous Decision Support System (DSS) for CBM. This integration involves data gathering, information extraction procedures, and real-time reasoning frameworks to facilitate the strategies and maintenance of critical systems. As a step further towards autonomy, DCBR builds on a self-evolving knowledgebase that learns from its performance feedback and reorganizes itself to deal with non-stationary environments. A unique Human-in-the-Loop Learning (HITLL) approach has been adopted to incorporate human feedback in the traditional Reinforcement Learning (RL) algorithm.

Technische Systeme nutzen sich ab. Dadurch bedingt kommt es zu Ausfällen. Um die Funktionstüchtigkeit von abgenutzten technischen Systemen wiederherzustellen, werden Instandsetzungsmaßnahmen durchgeführt. Da die Folgen eines unerwartet eintretenden Ausfalls drastisch sein können, ist es sinnvoll, das Abnutzungsverhalten eines technischen Systems vorherzusagen und so den Zeitpunkt von Instandsetzungsmaßnahmen zielgerichtet zu planen. Die Erstellung von Abnutzungsprognosen wird dadurch erschwert, dass sich technische Systeme oft variabel, in Abhängigkeit von auf sie einwirkenden Beanspruchungen, abnutzen. Außerdem wird diese Abnutzungsvariabilität von betriebsbedingten Einflüssen überlagert, was deren Modellierung erschwert. Im Rahmen dieser Arbeit wurden deshalb Lösungsansätze entwickelt, die es ermöglichen, die Abnutzungsvariabilität eines technischen Systems in Abnutzungsprognosen zu integrieren und dabei betriebsbedingte Einflüsse zu berücksichtigen. Somit können Instandsetzungsmaßnahmen präziser geplant, Ressourcen geschont sowie Kosten reduziert werden. / Technical systems are prone to deterioration. This leads to negative consequences like break-downs. Maintenance actions are executed in order to transfer technical systems back into healthy states. If break downs occur suddenly, the consequences can be dramatic. Therefore, it is reasonable to schedule maintenance actions based on health-state predictions. Thereby, health state predictions are impeded by the fact that technical systems often deteriorate variably, depending on certain stress factors. Furthermore, the effects of variable deterioration behavior can be hidden by system specific behavior. Thus, approaches are shown which integrate variable deterioration behavior into healthstate predictions while influences caused by the system specific behavior are considered. Consequently, maintenance actions can be scheduled more efficiently which spares resources and reduces costs.

Ces travaux de thèse traitent des problèmes de maintenance associés aux véhicules industriels. Ils se concentrent sur la planification des opérations de maintenance et sur le développement d'une méthodologie de conception pour la maintenance. Le but est de proposer une offre de maintenance personnalisée en fonction de chaque véhicule et capable de s'adapter aux contraintes des utilisateurs. Dans l'industrie du transport, ces contraintes se caractérisent par un nombre d'opportunités de maintenance limité et des immobilisations à fortes conséquences financières. Cette offre a vocation à garantir un niveau de disponibilité élevé tout en réduisant l'impact de la maintenance sur les coûts globaux d'exploitation. Dans ce cadre, la politique de maintenance développée vise à assurer, moyennant un certain risque, l'autonomie d'un système multi-composant sur des périodes d'opérations données. Pendant ces périodes, aucune opération de maintenance et aucune défaillance du système ne doivent venir perturber la réalisation des missions. A la fin de chaque période, la politique considérée évalue la nécessité d'une intervention de maintenance pour assurer la prochaine période avec un niveau de confiance spécifié. Lorsque la maintenance est jugée indispensable, des critères intégrant les coûts et l'efficacité de la maintenance sont introduits pour sélectionner les opérations à réaliser. Cette forme originale de regroupement dynamique s'appuie à la fois sur les modèles de fiabilité des composants, sur la structure fiabiliste du système et sur les informations de surveillance disponibles en ligne. Celles-ci se composent d'informations liées à l'état de santé des composants mais également à leurs conditions d'utilisation. La flexibilité du processus permet d'intégrer, dans la décision, des niveaux d'informations différents suivant les composants. Les paramètres de cette politique, à savoir la longueur de la période et le niveau de confiance, sont optimisés en fonction du coût total de maintenance. Ce coût, évalué sur un horizon fini, intègre les coûts directs associés aux opérations de maintenance et les coûts indirects engendrés par les immobilisations. Pour envisager une réduction significative des coûts d'exploitation du système, l'optimisation de la politique de maintenance seule ne suffit pas. Il est primordial de mener une réflexion plus large associant le système et sa maintenance dès la conception. Pour diriger cette réflexion, la méthodologie de conception proposée hiérarchise, à l'aide d'un facteur d'importance original, l'impact des composants sur les coûts d'exploitation. Différentes options de conceptions sont ensuite évaluées, par simulation, sur les composants jugés prioritaires. Les options retenues conduisent à réduire les coûts globaux d'exploitation. Des résultats de simulation permettent d'illustrer les méthodes développées. Une application sur un sous-système du véhicule industriel est également réalisée. / This thesis research work focuses on the maintenance operations scheduling and the development of a design methodology for maintenance. The aim is to suggest a customized maintenance service offer for each vehicle and able to adapt to user constraints. In the transport industry, these constraints are defined by a limited number of maintenance opportunities and vehicle unplanned stops with significant financial consequences. This service offer should enable both to improve the vehicle uptime and to reduce the maintenance impact on operating costs. In this framework, the developed maintenance policy ensures, with a given risk probability, maintenance free operating periods for a multi-component system. During these periods, the system should be able to carry out all its assigned missions without maintenance actions and system fault. And the end of each period, the considered policy evaluates if a maintenance action is required to ensure maintenance-free and fault-free operation on the next period with a specified confidence level. When a maintenance action is mandatory, decision criteria considering the maintenance costs and the maintenance efficiency are used to select the operations to be performed. This form of dynamic clustering, called time-driven clustering, integrates both the component reliability models, the system structure and the available monitoring information. In our case, the monitoring information refers to the component state information and information on the component operating conditions. The process flexibility makes possible to make a maintenance decision in using different information levels for system components. The policy parameters, namely the period length and the confidence level value, are optimized based on the total maintenance cost. This cost, evaluated on a finite horizon, is composed of directs costs related to maintenance operations and indirect costs generated by system immobilizations. In order to reach a significant operating costs reduction, the maintenance policy optimization alone is not sufficient. It is essential to have a broader approach to involve the system and its maintenance since the conception. In this context, the developed design methodology suggests to prioritize the components impact on the operating costs. This prioritization is performed thanks to a defined importance factor. Then, multiple design options are evaluated by simulation in priority component. The selected options lead to reduce the operating costs. This work contains simulation results that illustrate the methods mentioned above. Moreover, a heavy vehicle sub-system is used as a test-case.

Purpose: Explore how historical data from a CMMS can be used in order to improve maintenance effectiveness and efficiency of activities, and investigate the possibilities for CBM in the manufacturing industry in the context of digitalization.  Research questions: RQ1: To what extent could condition-based maintenance or other maintenance types being used in order to predict, prevent or in other way eliminate historical breakdowns/faults?  RQ2: Which significance has an organization's degree of maturity to reduce the number of breakdowns?  Method: A case study was performed at Volvo Construction Equipment Operations in Eskilstuna, who manufactures machinery for the construction industry. The case study was compiled in two phases. Phase one was a quantitative study where raw data were collected from a CMMS and tabulated in order to later perform in-depth analysis. Phase two was designed to collect information that generated a wider understanding of the research area, by performing interviews and observations. A literature study was performed to compare the empirical findings with peer-reviewed information to ensure the quality of the study. The data is compiled and analyzed with an abductive approach. The analysis was followed by a discussion of how the research findings could support identifying possibilities of different maintenance types in the future.  Conclusion: The result showed that using historical breakdown data from a CMMS can be useful in order to identify organization’s current state and what possibilities different maintenance types have to decrease the number of breakdowns. To what extent the breakdowns can be decreased relies not only on the maintenance type but also an organizations maturity level. The case study´s result showed that by combining different maintenance types and increasing degree of maturity, Volvo could decrease the historical breakdowns with 86,5%. By only using CBM with current maturity level, 56% of the historical breakdowns could be predicted. However, to decide how many breakdowns that is cost-effective to prevent and precisely what maintenance type that should be used requires a cost analysis which this study is not covering.

Made available in DSpace on 2016-08-17T14:52:58Z (GMT). No. of bitstreams: 1 Raimundo Nonato Rosa Filho.pdf: 1125835 bytes, checksum: 91689e7b58443f6d0eb73d752860ce37 (MD5) Previous issue date: 2005-03-31 / In this work a methodology for reduction of maintenance cost in the on-load tap changers (OLTC) of extra high voltage is proposed. The methodology is based on the use of Artificial Neural Networks (ANN) for the intelligent processing of input signals of the commutator. The neural nets adequately trained allow to create an information system and dedicated diagnosis of the OLTC. This system can interpret and diagnosis the components through the real time input signals in order to delay the power transformer maintenance intervals, foreseeing when the OLTC is going to maintenance have intervention based on its condition. It has been adopted a multiperceptron ANN architecture in which the input vector has 22 components and the output considers only one component with the status of the OLTC condition in function of its operation time. This output information is used to determine the periods of maintenance of the commutators. It is reported an application of the proposed system considering the on load tap changer of an autotransformer bank of 500/230/13.8 kV, 600MVA of Centrais Elétricas do Norte do Brasil S/A (ELETRONORTE). The results indicate the advantages of the maintenance based on the condition using ANN. / Neste trabalho é proposta uma metodologia para redução de custo de manutenção nos comutadores de tap sob carga (OLTC) dos transformadores de potência de extra alta tensão. A metodologia está baseada na utilização de redes neurais artificiais (RNA) para o processamento inteligente dos sinais de entrada dos comutadores. As redes neurais adequadamente treinadas permitem criar um sistema de informação e diagnóstico dedicado a OLTC que podem interpretar e diagnosticar os componentes através das entradas em tempo real de forma a, postergar os intervalos de manutenção, prevendo quando o OLTC deverá sofrer intervenção de manutenção baseada na condição do OLTC. Foi adotada uma arquitetura de RNA de multiperceptron na qual a entrada considera um vetor com 22 entrada e apenas uma saída com o status da condição do OLTC em função do tempo de operação. Essa informação de saída é utilizada para determinar os períodos de manutenção dos comutadores de tap. É realizada uma aplicação do sistema proposto considerando o comutador de tap sob carga de um banco de autotransformador de 500/230/13.8kV, 600MVA da Centrais Elétricas do Norte do Brasil S/A( ELETRONORTE) e os resultados indicam as vantagens da manutenção baseada na condição usando RNA.

Cette thèse est consacrée à la description, la prédiction et la prévention des défaillances de systèmes. Elle se compose de quatre parties relatives à la modélisation stochastique de dégradation, au pronostic de défaillance du système, à l'estimation du niveau de défaillance et à l'optimisation de maintenance.Le processus d'Ornstein-Uhlenbeck (OU) dépendant du temps est introduit dans un objectif de modélisation des dégradations. Sur la base de ce processus, le premier instant de passage d’un niveau de défaillance prédéfini est considéré comme l’instant de défaillance du système considéré. Différentes méthodes sont ensuite proposées pour réaliser le pronostic de défaillance. Dans la suite, le niveau de défaillance associé au processus de dégradation est estimé à partir de la distribution de durée de vie en résolvant un problème inverse de premier passage. Cette approche permet d’associer les enregistrements de défaillance et le suivi de dégradation pour améliorer la qualité du pronostic posé comme un problème de premier passage. Le pronostic de défaillances du système permet d'optimiser sa maintenance. Le cas d'un système contrôlé en permanence est considéré. La caractérisation de l’instant de premier passage permet une rationalisation de la prise de décision de maintenance préventive. L’aide à la décision se fait par la recherche d'un niveau virtuel de défaillance dont le calcul est optimisé en fonction de critères proposés / This thesis is dedicated to describe, predict and prevent system failures. It consists of four issues: i) stochastic degradation modeling, ii) prognosis of system failures, iii) failure level estimation and iv) maintenance optimization. The time-dependent Ornstein-Uhlenbeck (OU) process is introduced for degradation modeling. The time-dependent OU process is interesting from its statistical properties on controllable mean, variance and correlation. Based on such a process, the first passage time is considered as the system failure time to a pre-set failure level. Different methods are then proposed for the prognosis of system failures, which can be classified into three categories: analytical approximations, numerical algorithms and Monte-Carlo simulation methods. Moreover, the failure level is estimated from the lifetime distribution by solving inverse first passage problems. This is to make up the potential gap between failure and degradation records to reinforce the prognosis process via first passage problems. From the prognosis of system failures, the maintenance optimization for a continuously monitored system is performed. By introducing first passage problems, the arrangement of preventive maintenance is simplified. The maintenance decision rule is based on a virtual failure level, which is solution of an optimization problem for proposed objective functions

Aujourd’hui la prédiction des défaillances de certains systèmes industriels est devenue indispensable pour l’amélioration de la fiabilité et de la rentabilité de ces derniers. Cette prédiction s’appuie principalement sur l’analyse d’évolution du niveau de dégradation du système. Pour les systèmes dont l’état de détérioration n’est pas directement observable, la définition d’indicateurs de santé mesurables est nécessaire. Une modélisation du processus de dégradation à partir de ces données peut être ensuite effectuée. Dans cette thèse, nous considérons un ensemble d’indicateurs non-monotones pour un système opérant dans un environnement dynamique. Compte tenu des principales caractéristiques des données ainsi que de l’impact des conditions environnementales et de leur instabilité, une modélisation stochastique de l’évolution de ces indicateurs est proposée. Les modèles proposés se basent principalement sur une combinaison d’un processus de Wiener et de processus de sauts. Les motivations, les méthodes de calibration, l’utilité et les limites de chaque modèle sont discutées. Nous proposons ensuite une approche pour l’aide à la décision concernant les actions de maintenance préventive. Cette approche consiste à évaluer la valeur d’une option réelle qui présente la possibilité d’«Attendre avant d’Agir» suite à un signal d’avertissement sur une défaillance probable. Une application de cette approche pour le cas d'une éolienne équipée d’un système de surveillance et de gestion est traitée / A major concern for engineers and managers nowadays is to make high quality products and highly reliable systems. In this context, reliability analysis and failure prediction, besides of efficient maintenance decision-making are strongly required. Deterioration modeling and analysis is a fundamental step for the understanding and the anticipation of system behavior. Consider a functional system operating in unstable conditions or environment where the deterioration level is not observable and could not be determined by direct measures. For this system a set of measurable health indicator that indirectly reflects the system working conditions and deterioration level can be defined and examined. Considering these indicators, the development of a mathematical model describing the system behavior is required.In this thesis, we consider a set of non-monotone indicators evolving in a dynamic environment. Taking into account the major features of the data evolution as well as the impact of dynamic environment consequences and potential shocks, stochastic models based on Wiener and jump processes are proposed for these indicators. Each model is calibrated and tested, and their limits are discussed. A decision-making approach for preventive maintenance strategies is then proposed. In this approach, knowing the RUL of the system, a simulation-based real options analysis is used in order to determine the best date to maintain. Considering a case study of a wind turbine with PHM structure, the decision optimization approach is described

De nos jours, les systèmes industriels sont de plus en plus complexes tant du point de vue de leur structure logique que des diverses dépendances (dépendances économique, stochastiques et structurelles) entre leurs composants qui peuvent influencer l'optimisation de la maintenance. La Maintenance conditionnelle qui permet de gérer les activités de maintenance en fonction de l’information de surveillance a fait l’objet de beaucoup d'attention au cours des dernières années, mais les dépendances stochastiques sont rarement utilisées dans le processus de prise de décision. Par conséquent, cette thèse a pour objectif de proposer des politiques de maintenance conditionnelle tenant compte des dépendances économiques et stochastiques pour les systèmes multi-composant. En termes de dépendance économique, les politiques proposées sont conçues pour permettre de favoriser les opportunités de grouper des actions de maintenance. Une règle de décision est établie qui permet le groupement de maintenances avec des périodes d'inspection différentes. La dépendance stochastique causée par une part de dégradation commune est modélisée par copules de Lévy. Des politiques de maintenance conditionnelle sont proposées pour profiter de la dépendance stochastique.Nos travaux montrent la nécessité de tenir compte des dépendances économiques et stochastiques pour la prise de décision de maintenance. Les résultats numériques confirment l’avantage de nos politiques par rapport à d’autres politiques existant dans la littérature / Nowadays, industrial systems contain numerous components so that they become more and more complex regarding the logical structures as well as the various dependences (economic, stochastic and structural dependences) between components. The dependences between components have an impact on the maintenance optimization as well as the reliability analysis. Condition-based maintenance which enables to manage maintenance activities based on information collected through monitoring has gained a lot of attention over recent years but stochastic dependences are rarely used in the decision making process. Therefore, this thesis is devoted to propose condition-based maintenance policies which take advantage of both economic and stochastic dependences for multi-component systems. In terms of economic dependence, the proposed maintenance policies are designed to be maximally effective in providing opportunities for maintenance grouping. A decision rule is established to permit the maintenance grouping with different inspection periods. Stochastic dependence due to a common degradation part is modelled by Lévy and Nested Lévy copulas. Condition-based maintenance policies with non-periodic inspection scheme are proposed to make use of stochastic dependence. Our studies show the necessity of taking account of both economic and stochastic dependences in the maintenance decisions. Numerical experiments confirm the advantages of our maintenance policies when compared with other existing policies in the literature