Mestrado em Modelagem e Otimização - PPGMO

URI Permanente para esta coleção

https://repositorio.ufcat.edu.br/handle/tede/40
O Mestrado em Modelagem e Otimização tem um papel essencial no desenvolvimento de novas tecnologias. Atualmente, se faz muito intensamente o uso de modelos matemáticos, simulações avançadas e sofisticados desenvolvimentos computacionais na pesquisa científica em geral.

Navegar

Navegando Mestrado em Modelagem e Otimização - PPGMO por Assunto "Aprendizado de máquina"

Agora exibindo 1 - 1 de 1
  • Imagem de Miniatura
    Item
    Modelagem com aprendizado de máquina aplicada aos sistemas de monitoramento de integridade estrutural
    (Universidade Federal de Goiás, 2021-03-25) Barella, Bruno Pereira; Cunha, Márcio José da; https://orcid.org/0000-0002-4173-8031; http://lattes.cnpq.br/5012626154282569; Moura Júnior, José dos Reis Vieira de; https://orcid.org/0000-0002-8701-8355; http://lattes.cnpq.br/2479225716217997; Moura Júnior, José dos Reis Vieira de; Cunha, Márcio José da; Costa, Vaston Gonçalves da; Gallo, Carlos Alberto
    Structural Health Monitoring (SHM) using the electromechanical impedance method focuses on the development of systems responsible for monitoring the integrity of structures, such as, aerospace fuselage and metallic structures. Therefore, the use of computational tools and resources are essential. The identification, location, quantification of damage or even the prediction of the useful life of mechanical systems plays a fundamental role in ensuring the financial and public safety. That way, the use of machine learning tools is possible to perform such functions due to their high ability to identify patterns without needing to compensate for natural effects such as loading and temperature, which often act on mechanical structures. Thus, three possible approaches were applied for the development of models to be used in structural health monitoring systems, which are: anomalies detection, multi classification of damage and mass variation regression. To develop the approaches, an experiment was carried out simulating the structural damage produced by the surface machining process in a aluminum beam exposed to thermal variations between 10 ◦ C and 40 ◦ C. Systems that use machine learning have models developed with unique specifications intrinsic to the problem and, for their development, it is necessary to use tools and methods capable of performing their functions with high reliability and precision. In this sense, the main concepts of structural integrity monitoring techniques and the characteristics of intelligent materials used as sensors were presented. Subsequently, the electromechanical impedance method was approached with the essential concepts for its understanding, as well as the main systems available for collecting impedance signatures and the commonly used statistical models. Then, the common concepts and resources in the artificial intelligence field were presented, as well as tools for the analysis, interpretation and evaluation of models. Finally, a system was deployed containing the anomalie detection approach by means of a monitoring application, using the resources of containerization and cloud computing.