Made available in DSpace on 2014-12-17T14:08:41Z (GMT). No. of bitstreams: 1 AnthonyARD_DISSERT.pdf: 2982432 bytes, checksum: 71a6d988a19819a67932642761ac3dda (MD5) Previous issue date: 2010-09-03 / In the last decades, the oil, gas and petrochemical industries have registered a series of huge accidents. Influenced by this context, companies have felt the necessity of engaging themselves in processes to protect the external environment, which can be understood as an ecological concern. In the particular case of the nuclear industry, sustainable education and training, which depend too much on the quality and applicability of the knowledge base, have been considered key points on the safely application of this energy source. As a consequence, this research was motivated by the use of the ontology concept as a tool to improve the knowledge management in a refinery, through the representation of a fuel gas sweetening plant, mixing many pieces of information associated with its normal operation mode. In terms of methodology, this research can be classified as an applied and descriptive research, where many pieces of information were analysed, classified and interpreted to create the ontology of a real plant. The DEA plant modeling was performed according to its process flow diagram, piping and instrumentation diagrams, descriptive documents of its normal operation mode, and the list of all the alarms associated to the instruments, which were complemented by a non-structured interview with a specialist in that plant operation. The ontology was verified by comparing its descriptive diagrams with the original plant documents and discussing with other members of the researchers group. All the concepts applied in this research can be expanded to represent other plants in the same refinery or even in other kind of industry. An ontology can be considered a knowledge base that, because of its formal representation nature, can be applied as one of the elements to develop tools to navigate through the plant, simulate its behavior, diagnose faults, among other possibilities / Nas ?ltimas d?cadas, o segmento de ?leo & g?s e petroqu?mica tem registrado uma s?rie de grandes acidentes. Influenciadas por esse contexto, as empresas tem sentido a necessidade de se engajar em processos de prote??o do ambiente externo, que se traduz na preocupa??o ecol?gica. No caso particular da ind?stria nuclear mundial, a educa??o sustent?vel e o treinamento, que dependem muito da qualidade e utilidade da base de conhecimento, t?m sido considerados pontos chave para utiliza??o desse tipo de energia com seguran?a. Dessa forma, a motiva??o dessa pesquisa foi aplicar o conceito de ontologia como ferramenta para melhorar a gest?o do conhecimento em uma refinaria, atrav?s da representa??o de uma planta de ado?amento de g?s combust?vel, condensando os v?rios tipos informa??es associados com o seu modo de opera??o normal. Em termos de metodologia, este estudo pode ser classificado como uma pesquisa aplicada e descritiva, em que foram analisadas, classificadas e interpretadas informa??es que possibilitaram criar a ontologia descritiva de uma planta real. A modelagem da planta DEA foi realizada de acordo com os fluxogramas de processo, fluxogramas de tubula??o e instrumenta??o, documentos descritivos de seu modo de opera??o e a rela??o de alarmes associados, que foram complementadas com uma entrevista n?o estruturada de um especialista em seu modo de opera??o. A valida??o aconteceu atrav?s da compara??o de grafos montados a partir da ontologia com a documenta??o original e debatidos com o grupo de trabalho. Os conceitos utilizados nesta pesquisa podem ser expandidos para representar outras plantas da pr?pria refinaria ou mesmo de outras ind?strias. A ontologia pode ser considerada uma base de conhecimento, que devido ao seu car?ter formal, pode ser aplicada como um dos elementos no desenvolvimento de ferramentas de navega??o da planta, simula??o de comportamento, diagn?stico de falhas, dentre outras possibilidades

Le maintien en condition opérationnelle des équipements industriels est un des enjeux importants de l'entreprise, et a fait passer la maintenance d'un centre de coût à un centre de profit, ce qui a eu pour conséquence une éclosion de logiciels d'aide à la maintenance allant de la GMAO aux plateformes de e-maintenance. Ces systèmes d'aide fournissent aux différents acteurs de la maintenance, un support à la décision et un ensemble de services permettant une gestion informatisée d'activités de base appartenant au processus de maintenance (exemple l'intervention, la planification, le diagnostic, etc.). Toutefois, les besoins des utilisateurs évoluent dans le temps en fonction de nouvelles contraintes, de leur expertise, des nouvelles connaissances. Par contre les services fournis n'évoluent pas et nécessitent une réactualisation. Afin de tenir compte de l'évolution de ces connaissances, pour que ces systèmes d'aide puissent répondre aux besoins des utilisateurs et puissent proposer des services à la demande et des services évolutifs nous avons fait le point dans cette thèse sur les avantages et limites des systèmes informatiques d'aide existants notamment les plateformes de e-maintenance (systèmes les plus avancés aujourd'hui en maintenance). Pour pallier le manque des systèmes existants, nous avons proposé le concept de s-maintenance qui est caractérisé principalement par les échanges collaboratifs entre applications et utilisateurs, par des connaissances communes du domaine de maintenance. Pour mettre en œuvre ce concept, nous avons proposé une plateforme orientée connaissances assurant des fonctionnalités auto-x (auto-traçabilité, auto-apprentissage, autogestion) qui permettent de répondre aux caractéristiques de la s-maintenance. L'architecture à base de composants de cette plateforme prend appui sur une base de connaissances partagée entre les différents composants qu'elle intègre au profit de l'interopérabilité sémantique ainsi que de la capitalisation des connaissances. Nous avons par ailleurs développé une ontologie du domaine de maintenance sur laquelle s'appuie cette base de connaissances. Finalement, afin de développer les fonctionnalités auto-x assurées par la plateforme nous avons proposé un système à base de traces exploitant la base de connaissances et l'ontologie associée / Operational condition maintenance of industrial equipment is a principal challenge for the firm production. This fact transfer the maintenance from the cost center to the profit center which has lead to massif development of maintenance support system starting from the GMAO to the e-maintenance platform. These systems provide to the maintenance agent, decision-support, and set of services allowing a computerized management of core activities for maintenance process. (e.g. intervention, planning, diagnostic,...). However, the user request continues evolving in time with respect of their expertise, their renewed knowledge and new constraints. On the other hand, the existing services are not following their requirements and they need to be updated. In this thesis, an overview on the advantage and drawback of existing computerized support system, in particular the e-maintenance platform (the most advanced maintenance system) is presented in order to meet the users needs and propose scalable and on-demand services. To overcome the existing system shortage, we propose the s-maintenance concept characterized by the collaborative exchange between users and applications and the common knowledge of the maintenance field. Thus, to implement this concept, a knowledge-oriented platform is proposed providing the auto-x functionalities (auto-traceability, auto-learning and auto-management) and meeting the s-maintenance characteristics. The architecture based on components of this platform, is also based on shared knowledge between integrated components for the benefit of the semantic interoperability as well as for the knowledge capitalization. Maintenance domain ontology is also developed on which the knowledge base is rested. Finally, in order to develop the auto-x functionalities, provided by the platform, a trace-based system is proposed by exploiting the knowledge base and the associated ontology.