Study on estimation of internal cracks using electromagnetic waves and machine learning

In recent years, the aging of reinforced concrete structures and a shortage and aging of engineers have become problems in Japan. There is a need for methods that allow inspection of reinforced concrete structures without the need for skilled engineers. The focus of this study was on electromagnetic radar which can relatively easily detect reinforcing bars and internal defects. Then, machine learning approaches, convolutional neural network (CNN) and convolutional auto encoder (CAE), were used to detect small cracks inside concrete caused by rebar corrosion from electromagnetic radar measurement data. In this study, output images from electromagnetic radar were learned and judged by machine learning. The results showed that both approaches were able to determine the presence or absence of internal cracks. The CNN determination showed the existence of cracks in the range of 0.04–1.0 mm in crack width, while the CAE determination showed the possibility of estimating crack growth.