The use of data-driven algorithms for Structural Health Monitoring (SHM) allows continuous monitoring and online damage identification in structures subjected to changes in its operational and environmental conditions along its lifetime. The principal goal for SHM is oriented to the development of efficient methodologies to process the data obtained directly from the structures under inspection and provide results associated with the different levels of the damage identification process. Some advantages in the use of data from the structure are for instance that data are obtained from a sensor network which is permanently installed to the structure allowing to know in any time the state of the structure, it is possible to identify different kind of damages and the possibility of enhancing the information about the damage size, position, among others. However, one of the challenges in SHM is the development of algorithms or methodologies with an accuracy that allows to avoid false damage identification. As a contribution, this work presents a damage detection and classification methodology which use multivariate analysis and machine learning algorithms as a pattern recognition point of view. The validation of the methodology is supported by some performance measures that provide relevant information about the classification behaviour to select the different elements in the methodology. The methodology is tested with data from a structure with a piezoelectric sensor network and the results show that it is possible to demonstrate its usefulness in the damage detection and classification process.
展开▼
