Detecção de falha aplicada na atualização de probabilidade de falha

Detalhes bibliográficos
Ano de defesa: 2013
Autor(a) principal: Renan Nominato Oliveira Souza
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Minas Gerais
UFMG
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://hdl.handle.net/1843/BUOS-B33HJS
Resumo: Maintenance data collection, short and long term, is a task of great importance in a context that involves issues related to budget planning in industry. A reliable dataset of failures enables the denition of strategies that generate cost savings, reduction of incidents, increased productivity, customer satisfaction and make system upgrade scan be performed considering important information that can be viewed in the history of failures. The present work deals with the application of techniques for fault detection in systems subject to failures with the objective is update the life of component in real time. For this, the life of the components was adjusted in systems with a single type of failures, in others words, with only one failure mode the life of component was adjusted. Moreover in systems whose two or more types failure could happen, thats why was necessary another stage which occur the failure classication as well as then the life of component was updated. For fault detection was used Fuzzy / Bayesian algorithm. In order to failure mode ling was used statistical distribution of Weibull which generally is applied in problems related to life time of the component. The dataset of fails after upgrade, was tested from a Kolmogorov-Smirnov test to validate if the Weibull distribution t to the failure data. In systems that occurred more than one type of failure was used a classier NFC (Neuro Fuzzy Classier) for the purpose of to determine the type of failure. After determining the type of failure the problem was treated in the same way when there was only one type of failure. The classier achieve an accuracy rate of 85.04 % when applied to the classication of faults in electrical transformers as well as 88.88 % when applied in a interactive tank system. From the results obtained it was observed that the methodology can direct the maintenance policy to be used. Another important point was that the data now have much greater reliability to be collected as well as be classied. Thus, it is clear that the results of using the system are highly benecial for data storage in a maintenance environment