AI‐assisted reliability assessment of steel structures: methodology and application

Abstract

This paper presents a novel methodology based on Artificial Neural Networks (ANNs) for the reliability assessment of steel structural components within the EN 1990 framework. The approach enables large‐scale Monte Carlo simulations by replacing computationally expensive finite element analyses with validated ANN surrogate models. The methodology is established in two stages: first, through validation against deterministic datasets using the initial stiffness of welded beam‐to‐column joints, and subsequently extended to resistance‐based reliability assessment, focusing on the Column Web in Compression (CWC) component defined in EN 1993‐1‐8. Probabilistic input variability is introduced through advanced sampling techniques, including Latin Hypercube Sampling, allowing simulation of low failure probabilities with high computational efficiency. The results demonstrate the capability of ANN‐based models to support stochastic structural assessments and provide insight into the adequacy of current design expressions, establishing a scalable AI‐driven framework for reliability evaluation of steel joints.