Abstract During the operation of metro tunnels, structural performance could inevitably degrade due to the combined effects of the stochastic and disadvantageous environment. In order to reduce the randomness and uncertainty underlying the structural safety risk analysis in operational tunnels, this paper develops a novel hybrid approach to perform global sensitivity analysis. The deterministic and stochastic finite element (FE) model is used to develop the approximate relationship between input and output parameters with a high level of accuracy. Based on the simulated data from an FE model, a meta-model is constructed by a built Particle Swarm Optimization-Least Square Support Vector Machine (PSO-LSSVM) model. In this research, 10,000 groups of data are generated by the built PSO-LSSVM model, which provides data support for the global sensitivity analysis through Extended Fourier Amplitude Sensitivity Test (EFAST). The input variables with a high global sensitivity are identified as crucial variables which should be well controlled and managed during tunnel operation. A Hankou-Fanhu (H-F) tunnel section in the Wuhan metro system is utilized as a case study to verify the applicability of the proposed approach. Global sensitivity analysis enables the reduction of the epistemic uncertainty in tunnel structural safety management, providing insight into a better understanding of (1) the input-output causal relationships of the structural safety risk in operational tunnels, (2) the reduction of the epistemic uncertainty in project safety management of operational tunnels.
Read full abstract