Time-variant error-adaptive interval inversion for resilience improvement based on nonlinear redundancy of steel-concrete composite girder bridges

The system nonlinear redundancy of the multi-girder Steel-Concrete Composite Girder (S-CCG) bridge provides the necessary internal buffering mechanism against asymmetric extreme loads. In the process of developing a bridge resilience improvement strategy, error-adaptive capacity from system nonlinear redundancy against inevitable deviations holds the key to ensuring optimal comprehensive performance including reliability, redundancy, cost, and constructability. However, as a measure of such capacity, the “error-adaptive interval” in the identification of ductility degradation and transverse stiffness is not accounted for. This paper proposes a time-variant Error-Adaptive Interval Inversion (E-AII) framework to exploit the error-adaptive capacity, in which the inversion is formulated as a reliability-constrained, multi-objective interval optimization problem with the aid of the machine-learning method and the Fiber Macro-Element Numerical Simulation Method (FM-ENSM). The proposed method is validated using a real S-CCG bridge. Numerical validation shows that the FE-ENSM accelerates one single nonlinear analysis from 1568 s to 2 s while maintaining simulation accuracy. Furthermore, the deployed surrogate learning strategy for worst-case reliability index predictions achieves a coefficient of determination of R 2 of 0.99103 and a maximum relative error of 2.52%. These indicate that the proposed method is able to significantly mitigate the computational burden while sustaining high predictive precision. The multiple comprehensive performances of the resilience improvement process can be greatly optimized with the guidance of the maximized error-adaptive interval of the ductility factor and transverse stiffness. This method can be extended to the E-AII of other scenarios of bridge resilience improvement.