Highly aged infrastructure structures are required to be inspected once every five years, but there are not enough inspectors for the increasing number of structures to be inspected.Therefore, we would like to replace the decision-making process with AI-based prediction so that anyone can conduct inspections quantitatively. As a first step, we will use machine learning to predict the internal conditions of reinforced concrete structures. In a previous study, classification prediction of the presence or absence of internal cavities and regression prediction of cavity diameters were achieved by limiting the frequency band within the peak frequency used as input data.In order to perform regression prediction of reinforcing bar diameter inside reinforced concrete, this study preprocesses and selects input data using feature engineering and dimensionality reduction methods, and performs learning and prediction using XGBoost, one of the decision tree methods. We would like to discuss the relationship between the features and outputs in the specimen and the influence of the state of the specimen.
SAKUMA et al. (Wed,) studied this question.