Multi-Objective Optimization of Studs in Steel–Concrete Composite Beams

Abstract

The arrangement of shear studs in steel–concrete composite beams (SCCBs) significantly influences both mechanical performance and economic efficiency. However, existing optimization studies have predominantly focused on cross-sectional dimensions and material grades, with systematic optimization of the mechanical parameters and spatial layout of shear connectors remaining largely unexplored. This study proposes a framework that couples machine learning surrogate models with the NSGA-II algorithm to address this gap. For a continuous SCCB, stud spacing, shear stiffness, and pull-out stiffness are adopted as optimization variables. The surrogate-assisted NSGA-II is employed to generate the Pareto front, and the analytic hierarchy process (AHP) combined with the technique for order preference by similarity to ideal solution (TOPSIS) is subsequently applied to identify the global optimal solution. The results demonstrate that the multilayer perceptron (MLP) delivers the best overall predictive accuracy. Feature importance analysis reveals that the shear stiffness of the studs over the pier, denoted KS-P, dominates the axial force response of the deck slab, contributing 48% of the total importance. After optimization, the deflection increases by 3.9%, whereas the axial force in the concrete slab decreases by 11.7% and the stud consumption is reduced by 88%, confirming the effectiveness and engineering practicality of the proposed method. The proposed methodology provides a theoretical basis and technical support for the refined design of shear connectors in SCCBs.