A Comparison of Pattern Classification Approaches for Structural Damage Identification

A structural damage identification approach based on wavelet packet decomposition (WPD) and random forests (RF) was proposed and compared with other pattern classification approachs. The main procedure involves extracting energy features from vibration acceleration data through wavelet packet decomposition and then using these features as input for a RF classifier. The experiment was carried on an 8-storey steel shear building model in the case verification. The results show that the proposed method is effective for structure damage detection and can achieve a higher accuracy than other commonly used methods.