Análise de eventos de fusão em traços de OTDR utilizando redes neurais

• Ano de defesa: 2018
• Autor(a) principal: Jorge Júnior, Evandro Monteiro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Elétrica
• Palavras-chave em Português: Backpropagation; Caracterização; Fibra Óptica; Redes Neurais Artificiais; OTDR; Artificial Neural Networks; Characterization; Optical Fibers; Engenharia elétrica; Back propagation (Inteligência artificial); Fibras óticas; Redes neurais (Computação); CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA::TELECOMUNICACOES
• Link de acesso: https://repositorio.ufu.br/handle/123456789/22575; http://dx.doi.org/10.14393/ufu.di.2018.1189

When an optical link is installed or when it will perform a technology upgrade on an existing link, it is necessary to perform tests to determine the state of that link, and if it is possible to traffic the requested data. The most important equipment in these tests is the OTDR, but the software contained in the equipment uses very simple algorithms that can not always detect events that occur in this fiber. This dissertation aims to present an algorithm based on Artificial Neural Networks, through the backpropagation training method for a new way of identifying fusion events in OTDR traces. We developed software in Matlab, using the backpropagation algorithm to perform the detection of fusion events in an OTDR trace. Then the fiber detection software was used in real cases. Finally, an analysis was made on the cases in which the Neural Network was used, and it was concluded that it is possible to use the Neural Networks technique to solve problems of this type, in addition some proposals were left for future work involving the theme.