Application of machine learning in monitoring systems of civil structures
StrStructural Health Monitoring (SHM) aims primarily to accurately identify the current state of a structure, assessing damage levels and eventually allowing to predict its future performance. In civil engineering structures, SHM is often responsible for ensuring public safety while handling with large and complex data. Such monitoring problems can be difficult to solve by conventional computing techniques alone, as they require the acquisition of large data sets that need to be thoroughly and carefully analyzed. This yields big data opportunities to use artificial intelligence methodologies.
This paper presents the integration of machine learning (ML) techniques for pattern recognition in SHM systems of civil engineering structures. The developed SHM consists of data acquisition both from time series of values observed at regular intervals and from structurally relevant measured values, called events, where specific data are collected.
ML is used in the development of statistical models for feature discrimination. Events are classified into different clusters in a semi-supervised learning procedure, which is an extension of an unsupervised learning to allow their identification. A real-world SHM implementation is presented as a case study of the ML application. It consists of an industrial steel tower structure, containing several mechanical equipment with different loads, which operate at various frequencies.
A sensor network is installed, acquiring data on strains, accelerations, and weather conditions. A visualization user interface is provided to access all data through a user friendly and accessible tool. The paper presents the main results obtained and illustrates the potentialities of the applied ML methodology.
AUTORES | João Zeferinoª, Eduardo Gonçalvesª, Paulo Carapitoª, Filipe Santosª
VESAM Engenharia S.A., Zona Industrial de Cantanhede, Lote 69, 3060-197 Cantanhede, Portugal
Artigo apresentado na 10.ª Edição da Conferência Internacional sobre Monitorização da Saúde Estrutural de Infraestruturas Inteligentes (SHMII 10), de 30 de junho a 2 de julho de 2021, no Porto, Portugal.
