Full-Scale Building Structural Health Monitoring by Shake Table Tests and Extreme Learning Machine

Abstract

Structural health monitoring (SHM) represents a type of techniques that enables the monitoring of structural motion of a building structure during external loading such as earthquakes. SHM always provides the in-depth understanding of the severity and location of damage to a structure without requiring visual structural inspection. In this research, it proposes a data-driven approach for building structural health monitoring based on data collected from Shake Table tests. Experimental data from shake table experiments has been utilized and analyzed. Fast Fourier Transformation is employed to extract the SHM related data-features. Engineers are also invited to label the ground truth risk levels according to the observation from the shake table test. A data-driven classifier namely extreme learning machine is introduced to classify the structural risk based on the extracted features. Comparison is performed with other state-of-art machine-learning classification algorithms. Numerical experiments validates the effectiveness, efficiency, and universality of the proposed method.