On the usage of active learning for SHM

The key element of this work is to demonstrate a strategy for using pattern recognition algorithms to investigate correlations between feature variables for Structural Health Monitoring (SHM). The task will take advantage of data from a bridge. An informative chain of artificial intelligence tools will allow an active learning interaction between the unfolded shapes of the manifold of online data by characterising the physical shape between variables. In many data mining and machine learning applications, there is a significant supply of unlabelled data but an important undersupply of labelled data. Semi-supervised active learning, which combines both labelled and unlabelled data can offer serious access to useful information and may be the crucial element in successful decision making, regarding the health of structures.