Hybrid-Driven Digital Twin Framework for Time-Variant Reliability Assessment of Civil Structures

This paper proposes a novel hybrid-driven digital twin (DT) framework for time-variant reliability assessment of civil structures, which mainly consists of four modules, including physics model construction, data-driven model calibration, failure probability calculation, and time-variant reliability prediction. In the first module, a DT model of a specific structure is constructed to simulate structural dynamic responses. Then, an improved unscented Kalman filter (IUKF) algorithm is performed to continuously calibrate the parameters of DT model. Subsequently, in module 3, the subset simulation (SS) approach is employed to calculate failure probability of structures subjected to various model parameter samples, and the generated input–output samples are further applied for metamodel training. A Kriging metamodeling is used to construct the correlation between model parameters and structural failure probability. Once the metamodel is well trained, the time-variant reliability assessment of structures can be continuously achieved in module 4. Numerical simulations on a Bouc–Wen model are conducted to validate the feasibility and accuracy of the proposed approach. Furthermore, a scaled column shake table structure is further employed to verify the effectiveness of the proposed approach. Both numerical and experimental results have shown that the proposed approach is capable of conducting time-variant reliability assessment of civil structures.