Fundamental Study on Damage Detection of Civil Structures Modeled as MDOF System Based on Machine Learning

Abstract

In Japan, the deterioration of civil structures constructed during the period of strong economic growth combined with the shortage of engineers has become a serious social problem. One solution to address this issue is to improve the efficiency of structural maintenance and management using sensors. The application of machine learning method is expected to advance such efforts. This research attempts the damage detection of civil structures using an autoencoder, which is a type of unsupervised machine learning. The study revealed that there is a relation between the magnitude of reconstruction error and the degree of damage to the structure.