Fuzzy unsupervised-learning techniques for diagnosis in a composite UAV wing by using fiber optic sensors

Joham Alvarez-Montoya, Julian Sierra-Perez

    Resultado de la investigación: Capítulo del libro/informe/acta de congresoContribución a la conferenciarevisión exhaustiva

    1 Cita (Scopus)

    Resumen

    Pattern recognition, which aims to associate data with a condition of a structure, has been applied successfully in Structural Health Monitoring (SHM) for damage diagnosis. Such association becomes more complex when applied to strain data measured from aerospace structures where the operational conditions produce changes in the strain patterns that are not related to a damage occurrence. Moreover, a damage occurrence only produces subtle changes in such patterns. That is why novelty detection strategies based on unsupervised learning have not demonstrated suitable results for strain-based SHM in operating aerospace structures. For example, imagine that it is possible to have data from all the different operational conditions for an aerostructure and, therefore, construct a model (e.g. statistical) from these data. Such model may be too general and some data from damage conditions may fit into the model and, subsequently, classified as a normal condition. One successfully-proved approach is to use unsupervised-learning, density-based classification to create clusters according to the operational condition and, then, build models for each specific cluster. In previous works, the authors implemented such methodology in an aluminum beam under simulated environmental conditions and subsequently, in the wing's main beam of an Unmanned Aerial Vehicle (UAV) made of composites. The results for the metallic structure demonstrated a good performance since the changes in the patterns due to the operational condition variations were clear and identifiable. On the other hand, the data acquired from the UAV demonstrated to be fuzzy and without clear transitions among clusters. The aim of this work is to explore fuzzy clustering techniques in order to improve the global performance of the methodology for composite aerospace structures, which exhibit high stiffness. Fuzzy C-Means (FCM) and Gustafson- Kessel (GK) algorithms were tested using the UAV flight data and their performance was evaluated through Receiver Operating Characteristic (ROC) analysis.

    Idioma originalInglés
    Título de la publicación alojadaProceedings of the 7th Asia-Pacific Workshop on Structural Health Monitoring, APWSHM 2018
    EditoresZhongqing Su, Shenfang Yuan, Hoon Sohn
    EditorialNDT.net
    Páginas682-690
    Número de páginas9
    ISBN (versión digital)9783000603594
    EstadoPublicada - 2018
    Evento7th Asia-Pacific Workshop on Structural Health Monitoring, APWSHM 2018 - Hong Kong, China
    Duración: 12 nov. 201815 nov. 2018

    Serie de la publicación

    NombreProceedings of the 7th Asia-Pacific Workshop on Structural Health Monitoring, APWSHM 2018

    Conferencia

    Conferencia7th Asia-Pacific Workshop on Structural Health Monitoring, APWSHM 2018
    País/TerritorioChina
    CiudadHong Kong
    Período12/11/1815/11/18

    Nota bibliográfica

    Publisher Copyright:
    © APWSHM 2018. All rights reserved.

    Huella

    Profundice en los temas de investigación de 'Fuzzy unsupervised-learning techniques for diagnosis in a composite UAV wing by using fiber optic sensors'. En conjunto forman una huella única.

    Citar esto