ABSTRACT
Visual inspection of bridges is important to maintain the sustainability of civil infrastructure. A hammering test is one of the non-destructive and verified methods used for this purpose, by which an expert inspector identifies spalling inside a concrete structure. The reliability of this method depends on the technique of the inspector. However, Japan society faces the problem of a scarcity of labor across all fields. Because of this, there are few training opportunities for new inspectors to learn the technique. This study aims to identify spalling concrete by using an autoencoder, instead of professional inspectors. First, a spalled concrete specimen was used to record the hammering sounds. The experimental method and a comparison of frequency spectra at each level of damage were described. Next, the autoencoder was applied to identify spalling concrete, and a comparison of the identification precision at each damage level was conducted. The results of the analysis show that spalling concrete can be identified by autoencoder to a concrete thickness of less than 30 mm.
