Técnicas avançadas de normalização de dados aplicadas ao método de monitoramento de integridade estrutural baseado em impedância eletromecânica

• Ano de defesa: 2017
• Autor(a) principal: Rabelo, Diogo de Souza
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Engenharia Mecânica
• Palavras-chave em Português: Monitoramento de integridade estrutural; impedância eletromecânica; normalização de dados; limiar discriminante; trinca de fadiga; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA::MECANICA DOS SOLIDOS
• Link de acesso: https://repositorio.ufu.br/handle/123456789/19907; http://dx.doi.org/10.14393/ufu.te.2017.1

This work is dedicated to the Structural Health Monitoring (SHM), an area that proposes to interrogate a structure with the purpose of detecting, locating and identifying structural damage. Currently, SHM represents one of the areas of great interest in engineering, with the goals of increasing safety while reducing maintenance costs, as well as decreasing downtime from machines and equipment. For this aim, the impedance-based SHM utilizes piezoelectric transducers that are used in high frequency ranges (typically above 30 kHz), providing high sensitivity to incipient type of damage. However, changes in environmental or operational conditions can cause changes in the frequency spectrum of the acquired signals, thus leading the system to a false diagnosis regarding the structural health. In this thesis, it was aimed to develop data normalization techniques that, when applied to the acquired impedance signals, the changes are compensated for without loss of important information. Furthermore, a method for threshold determination was proposed through the use of a statistical model, as well as advanced data normalization techniques were proposed, one of which has used optimization methods and the other is based on the minimization of the Euclidian distance between a point and a curve. It is important to point out that, despite the absence of precise numeric models, the studies presented here come from an experimental proposal, using innovative techniques, optimization methods, data treatment and analysis, statistics and dedicated programming. Finally, the results conveyed demonstrate the great potential of the use of the impedance-based SHM in conjunction with properly applied statistics and data normalization techniques.