Data-driven nonlinear shear force–slip modeling for automated optimization of grouped studs in steel–precast concrete composite bridges

Abstract In composite bridges incorporating precast concrete components, the behavior of grouped stud connectors becomes increasingly complex due to nonlinear interactions governed by additional design variables such as stud group arrangement and the geometric and material characteristics of cast-in-place concrete within pockets and surrounding precast regions. This study develops a novel and interpretable nonlinear shear force–slip model using explicit data-driven learning techniques, providing an automated optimal design tool for grouped stud connectors. An experimental database comprising 249 push-out tests with 19 input parameters was compiled from the literature to support model development and validation. Existing design codes were evaluated and shown to exhibit limited predictive accuracy, particularly for slip-related behavior. New data-driven predictive equations were derived and integrated into a continuous nonlinear shear force–slip relationship. The proposed models achieve correlation coefficients exceeding 0.86 for shear capacity and 0.7 for critical slip values prediction, substantially outperforming existing approaches. A practical case study conducted under realistic design constraints demonstrates that embedding the proposed model within an automated optimization framework enables effective optimization of grouped stud configurations, increasing total shear capacity within concrete pockets by up to 50% compared with non-optimized designs.