2024 Volume 80 Issue 6 Article ID: 23-00191
Periodic monitoring of the rust condition is necessary for the maintenance of weathering steel bridges. In recent years, some studies have been conducted to simplify and quantify the rust condition rating method. In this study, a rating technique that combines the optical spectra in visible light and near-infrared of a weathering steel surface with a supervised learning classifier was proposed. First, the optical spectra of weathering steel surfaces in different rust conditions were measured every 1 cm square, and four wavelengths (wavelengths of 569 nm, 694 nm, 796 nm and 896 nm) were selected from the first derivative of the measured optical spectra as characteristic wavelengths for rating the rust condition. The first and second principal components were extracted by applying kernel principal component analysis to the reflection intensities at the characteristic wavelengths, and random forest classifier was constructed. As a result, accuracy was 90.7 %.
