Surrogate model-based uncertainty analysis of stiffened plate ultimate strength accounting for spatio-temporal corrosion degradation

Kakaie, A.; Garbatov, Y.; Guedes Soares, C.

Key information

Authors:

Kakaie, A. (Arman Kakaie); Garbatov, Y. (Yordan Garbatov); Guedes Soares, C. (Carlos Guedes Soares)

Published in

February 15, 2026

Abstract

This study introduces a method for propagating uncertainty and conducting sensitivity analyses of the ultimate strength of a ship's structural components, accounting for the spatiotemporal variability of the corrosion process. The proposed approach combines non-linear finite element analysis with a surrogate modelling technique to enable accurate and efficient probabilistic modelling of ageing ship structures. An extended non-stationary stochastic Gamma process, in conjunction with random field theory, is employed to model the temporal and spatial variability of corrosion. To develop a polynomial chaos Kriging surrogate model, numerous random samples of stiffened plates are generated across various time intervals and analysed using non-linear finite element methods, accounting for uncertainties in geometry, material properties, initial imperfections, and the non-stationary, non-uniform corrosion behaviour. These surrogate models are subsequently utilised to propagate uncertainty in the ultimate strength of the stiffened plates and to identify the most influential factors affecting their structural performance. The results of this study emphasise the importance of incorporating both spatial and temporal variations of corrosion into probabilistic models of the ultimate strength of stiffened plates, as non-uniform thickness reduction can lead to local failures before global collapse, thereby reducing the overall ultimate strength and increasing its uncertainty.