Response Characterization of Concrete Filled Steel Tubular K Joints: From Experimental Tests to Finite Element Modelling and ANN Predictive Approaches

Concrete filled steel tubular members are widely used in high-rise buildings, offshore structures, and tubular arch bridges due to their high strength, enhanced stiffness, and excellent seismic performance. Despite their widespread applications, the ultimate capacity of joints in a tubular structure is significantly affected by the interaction of geometry, material properties, loading conditions, etc. Experimental campaigns provide important findings on structural performance and failure mechanisms, yet their scope is often limited by cost and practical constraints, resulting in a narrow range of tested geometric configurations. Moreover, capacity models are typically developed for a single failure mode, and ignore interaction among the steel tube, concrete core, and shear studs, leading often to non-reliable predictions and inefficient design for CFST K joints exhibiting different or combined failure modes. This research addresses these limitations by employing advanced finite element modelling (FEM) and data driven artificial neural network (ANN) predictions to investigate the influence of key geometric parameters on the ultimate capacity of CFST K-joints. Multiplanar KK joints are examined under symmetric and antisymmetric brace loading, and the influence of tensile and compressive chord preload are evaluated. Maximum load capacity and stable load displacement response are observed depending on a high number of combinations of the geometries characterizing the components of a CFST joint. The influence of shear studs and ring stiffeners, generally used to enhance the load bearing capacity, stiffness, and ductility of CFST K joints, has been also investigated. The combination of compressive chord preload and shear studs or ring stiffeners has been focused for understanding the changes in stiffness, peak load and post peak behaviour as well as the effect of tensile chord preload. Overall, a validated FE framework has been provided. Further, a quantitative guidance for geometry and strengthening selection, and ANN based predictive tools, enabling safer and more efficient CFST K-joint design with lowering reliance on costly experimental testing have been produced.