Abstract

Over the last decades, in spite of its cost, the use of stainless alloys has grown significantly owing to environmental concerns, low maintenance, and high corrosion resistance. This paper is aimed at developing a prediction model for the resistance of slender austenitic stainless steel plate girder subjected to patch loading using a machine learning based approach. Firstly, the study is conducted through geometrically and materially nonlinear with imperfection analyses (GMNIA). The numerical methodology is validated by comparison with experimental results available in the literature. Secondly, an extensive parametric analysis covering a wide range of girder geometries is performed. A numerical database resulting from the parametric analysis is fitted using symbolic regression and a resistance model is attained. Finally, the accuracy of the prediction model is evaluated by comparison of the predicted resistances with values computed numerically, and resistances predicted using available formulae from the literature.