The proposed doctoral project will develop a novel, reliability-based framework for the numerical design and assessment of thin-walled steel structures through the integration of advanced finite element modeling, digital twins, machine learning, and probabilistic analysis. The motivation is to facilitate current engineering practice: although geometrically and materially nonlinear analysis with imperfections (GMNIA) enables realistic simulations of the behavior of members and connections, its sensitivity to design assumptions limits its direct applicability in decision-making. The proposed approach will employ a design-consistent digital twin, in which numerical models are continuously updated using monitoring data to ensure consistency with design standards and reliability requirements defined within the European Eurocode system.