Years of repeated wet–dry cycles of seawater corrosion during the ballasting and deballasting process changes the surface morphology of ballast tanks. The resulting irregularly shaped corrosion pits and pitting severity affect the structural performance of ballast tanks. Accordingly, this thesis focuses on the changes in surface morphology and their correlation with the geometry of the corrosion pits and pitting severity. The changes in morphology are associated with the corrosion pit’s depth distribution that is acquired through a three-dimensional (3D) laser scan. Furthermore, the corrosion pit’s depth distribution is analyzed using statistical parameters at different volumetric changes (vr). Moreover, to assess the randomness of surface characteristics, i.e., Gaussian or non-Gaussian surface, the probability model analysis and the goodness-of-fit statistical tests are conducted on the corrosion pit’s depth distribution. The obtained results showed a correlation between the differences in the pitting severity at several volumetric changes (vr) and changes in the geometry of the corrosion pits. Furthermore, corroded specimens with non-Gaussian surface and relatively less pitting severity with a less sharp corrosion pit formed at large volumetric changes (vr) are studied numerically. Moreover, considering different geometries of corrosion pits, this thesis investigates the stress distribution within the area of a corrosion pit because it simplifies the non-Gaussian surface. Therefore, 45 dog bone–shaped finite element (FE) models with a single corrosion pit are investigated. The parameters of the corrosion pit’s geometries are analyzed with quantitative correlations at different sizes, shapes, and different enlargement. In addition, the scenario changes in corrosion pit’s geometry at significant volumetric changes (vr) are proposed, for example, narrow-deep, hemispherical, and shallow-wide form. Hence, the results consider stress distribution within various geometries of corrosion pits in different scenarios. Furthermore, specimens with corrosion depth conforming to Gaussian surface are analyzed numerically to determine the correlation between surface roughness and pitting severity at different volumetric changes (vr). In particular, seven rectangular FE models with Gaussian surfaces are elaborated. The surface roughness modeling considers correlations among corrosion statistical parameters at various volumetric changes (vr) of the corroded specimen. Further, the severity of pitting corrosion characteristics is estimated using the pitting factor (pf) at different volumetric changes (vr) and the correlation between the maximum (a.max) and average (a.ave) depth of corrosion pit. The pitting severity–volumetric changes (vr) scenario is proposed to investigate the rough surface model’s stress distribution, i.e., the stress distribution on the surface of the corroded specimen considering the correlation between surface roughness and pitting severity. In summary, a procedure for numerically modeling the surface of corroded specimens is developed. The depth of a corrosion pit obtained by a 3D laser scan is analyzed with a probability model to assess changes in the morphology and characteristics of the pit’s depth at a certain volumetric change (vr). Furthermore, the goodness-of-fit statistical test is performed to observe the propensity of corrosion pit’s depth to a specific distribution, i.e., Gaussian or non-Gaussian. The corroded specimens conforming to the Gaussian distribution are numerically modeled with ANSYS II Parametric Design Language to generate a Gaussian surface. Furthermore, the non-Gaussian surface model is simplified as a single corrosion pit with various geometrical shapes. These surfaces correlate to the pitting severity and the random characteristics of the shapes of corrosion pits and volumetric changes (vr), which affect the life of the plate and residual strength of the corroded ballast tank unexpectedly. Thus, the structural assessment should be able to consider different surface morphologies of corrosion pits.weiterlesen