Prediction of construction condition of stud dowel by digital hammering inspection using machine learning

Abstract

Digital hammering inspection using acoustic emission (AE) sensor is expected to be a method to investigate for improper construction of stud dowel in concrete structures. AE sensor measure vibration waveforms using mechanical resonance of piezoelectric elements. Since there is no established method for the predicting the state of the stud dowel using the waveforms obtained from AE sensor, machine learning, such as random forests and XGBoost, is used to make predictions. The features used for prediction are the decay time and the frequency and amplitude of the waveform when it was converted to the frequency domain by the fast Fourier transform. Binary classification is used to predict whether an abnormality of stud dowel is present or not. It is possible to improve the accuracy by dividing the data according to the state of the stud dowel.