Acoustic Emission-based Damage Detection and Classification in Steel Frame Structure Using Wavelet Transform and Random Forest

Abstract

This research proposes a unique approach for detecting damage locations and identifying damage kinds. This method is beneficial for discovering and categorizing internal structural faults that vision-based approaches cannot locate. Construction-related vibrations in a steel frame structure can be used as a source for acoustic emission. Sensor devices detect the stress waves produced by structure collapse, and spectrum analysis using wavelet transform of such data is valuable in pinpointing the location of the damage. The collected characteristics from these signals are input into the most effective RF (Random Forest) classifier, which are used to categories damage types like cracks and bolt loosening. When compared to previous damage localization approaches, the findings show that the proposed strategy is more efficient and has a higher classification accuracy.