Prediction of seismic moment-rotation response in cold-formed steel bolted connection frames: A data-driven approach

초록

This study presents a data-driven framework for predicting the nonlinear moment-rotation response of cold-formed steel (CFS) bolted connection frames under seismic loading. A mechanics-based analytical model capable of simulating slip, bearing-induced hardening, and local-buckling softening was validated against five experimental tests and used to generate a large-scale parametric dataset of 56,700 frame configurations, covering a wide range of crosssectional geometry, bolt group arrangement, frame length, material yield strength, bolt-hole clearance, and slip strength. Particle swarm optimization (PSO) regression was employed to develop explicit predictive equations for eleven key response coordinates defining the full moment-rotation curve, achieving coefficients of determination predominantly above 0.9. For key seismic design quantities, the peak moment was predicted with near-zero bias and an MAE of 2.34%, and residual moments at the IMF and SMF classification thresholds (0.02 and 0.04 radians) were estimated with MAE values of 6.71% and 9.10%, respectively. The proposed framework enables rapid estimation of seismic performance parameters, including slip resistance, peak moment, and post-buckling strength, without detailed numerical analysis, thereby providing a practical system-level design approach that complements current component-based standards for CFS bolted moment connections.

키워드