Life-cycle seismic reliability evaluation framework for steel frames considering degradation effects based on adaptive sampling and BPNN

To the quantitative assessment of structural safety throughout the entire lifecycle and the achievement of performance-based seismic design, the multi-level life-cycle seismic reliability evaluation framework has been proposed, leveraging the time-discretization approach. It comprehensively accounts for performance degradation and quantifies life-cycle failure probabilities at different stages of its service life. The framework adopts a performance-based seismic design concept, conducting reliability assessments for structures at various seismic hazard levels and different performance levels. Moreover, given the complexity and high dimensionality of earthquake-resilient structural systems, an Adaptive Sampling Backpropagation Neural Network (AS-BPNN) with Euclidean distance constraints has been introduced to enhance the accuracy and computational efficiency of failure probability. The global seismic reliability analyses of steel frame structures with replaceable connections were performed for the load-bearing capacity and deformation capacity. The results demonstrated that the impact of life-cycle effects on deformation capacity limit states is more pronounced than on load-bearing capacity. As the seismic hazard levels increase, the impact of life-cycle degradation on the deformation capacity limit state of earthquake-resilient steel frames becomes more evident. For earthquake-resilient steel frame structures that are more sensitive to deformations, the influence of life-cycle degradation on the seismic capacity of structures should be considered.