Abstract
This study aims to statistically analyze the distribution characteristics of localized corrosion along the length of corroded steel bars with spectral analysis techniques. Steel bars were embedded in a concrete prism and subjected to accelerated corrosion to levels ranging from 5.0 wt.% to 30.0 wt.% mass loss. After the corrosion test, the corroded steel bars were taken out of the concrete and cleaned with a sand blaster, and then scanned with a 3D laser scanner. The scanned point clouds were processed with an image processing software to determine the residual cross-sectional area distribution. Empirical mode decomposition (EMD) was performed by considering the area distribution of the steel bars as a nonlinear and non-stationary time series and the length as time. The intrinsic mode functions (IMFs) extracted from the EMD reveal the characteristics of pitting corrosion of various sizes, and the magnitude of the IMF is related to the cross-sectional area of corrosion pits. To extract the characteristic length of pitting corrosion, the fast Fourier transform (FFT) was performed on each IMF. Results show that the EMD-FFT successfully extracts the spatial distribution characteristics of corroded steel bars including surface irregularity, deformation (ribs and lugs), small corrosion pits and big corrosion pits. Presence of pitting corrosion changes the characteristic length of surface irregularity and rib spacing compared to a non-corroded bar.
