Enhancing Reliability Assessment of Civil Structures Based on Digital Twin Modeling Subjected to Multi-Source Uncertainties

In this study, a novel digital twin (DT) modeling approach is developed to enhance the reliability assessment of civil structures subjected to multi-source uncertainties. Referring to the proposed procedure, a modular Bayesian inference (MBI) with the Transitional Markov Chain Monte Carlo (TMCMC) sampling algorithm is first used to calibrate the DT model of structures based on measured structural responses. Distinguished from the classic Bayesian inference approach, the uncertainty caused by modeling errors is added to the extended likelihood function by using a bias function. Then, the DT model calibration of structures can be achieved by three modules. Based on the calibrated DT model of structures, a finite number of discrete structural response samples can be generated by performing nonlinear dynamic analysis. Then, the generalized extreme value distribution (GEVD) is used to fit the probability density function (PDF) of these discrete response samples based on the maximum likelihood estimation (MLE) algorithm. Subsequently, the earthquake-induced failure probability of structures can be assessed by directly integrating the fitted GEVD and predefined structural failure thresholds. Numerical simulations on a three-story steel frame structure subjected to seismic excitations are developed to validate the feasibility of the proposed approach. The shake table tests on a scaled reinforced concrete (RC) column structure are further conducted to verify the effectiveness of the proposed approach. Both numerical and experimental results demonstrate that the proposed approach is reliable and highly effective for structural reliability assessment.