Structural Health Monitoring of Building Structures With Online Data Mining Methods

Structural health monitoring (SHM) uses building vibration responses from online sensors to diagnose and evaluate possible structural damages. This is a data classification process. Normal data mining methods, such as data clustering and support vector machine (SVM), cannot classify these online data streams directly. In this paper, the convex-concave hull and SVM are modified into online versions, such that the novel data classifiers for the SHM can be trained online. We design a laboratory-scale prototype for the experiment evaluation. The experiment results show that the proposed SHM algorithms can be trained online and detect the damage successfully.