Performance assessment of corroded stiffened steel plates considering non-Gaussian spatial correlated corrosion random field

This paper investigates the effect of corrosion randomness on the compression performance of stiffened steel plates. An anisotropic non-Gaussian random filed model is proposed to characterize the corrosion depth of the corroded steel plate surface. The specimens of corroded steel plate were obtained utilizing the artificial climate environment method. Corrosion characteristics were measured through 3D-scanning and reverse modeling. Statistical methods are utilized to analyze corrosion depth, including probability distribution and correlation coefficient. Two models are proposed for steel plate placing flatly and inclinedly, respectively. It is found that for both models corrosion depth follows truncated GEV distribution, and the correlation coefficient follows a bidirectional exponential function. To investigate the effect of non-Gaussian and correlation characteristics on bearing capacity, the proposed model is used for generating anisotropic non-Gaussian corrosion random fields for stiffened steel plates. Generated samples were used for finite element analysis. It is found the ignorance of non-Gaussian or correlation of the corrosion depth field will underestimate the deterioration of bearing capacity of stiffened steel plates. Finally, a parametric study on the correlation coefficients of the corrosion field indicates that correlation length will affect the residual bearing capacity of corroded steel plate. The proposed random field for corrosion depth of steel plate could be used for potential application on the safety assessment of corroded steel structures.