Global Sensitivity Analysis in the Prediction of Surface Settlement Caused by Mechanized Tunneling

Global sensitivity analysis is one of the beneficial and useful tool to identify the uncertainty of input variables that has been extremely investigated in different science such as simulations. Sensitivity analysis is an essential step in the production of meta-model, which, by identifying effective parameters in tunneling, reduces the time and computations required. In this paper, sensitivity analysis was carried out on geotechnical and operational parameters of EPB mechanized tunneling in soft soil. So, the tunneling processes were modeled using a finite difference method in FLAC 3D software, and the numerical model was validated by the monitoring data obtained from the East-West route of Tehran metro 7 line. The sensitivity analysis by using elementary effect Morris method was performed on the 6 input parameters and three parameters (face pressure, specific gravity and cohesion of the soil layer which the tunnel was excavated) were selected as effective and sensitive parameters in the maximum surface settlement. Then to construction of meta- model, 100 samples were generated from effective parameters using the Latin Hypercube method. After numerical simulation for each sample, the simulation results were used for surface settlement prediction by using artificial neural network. The results showed that the prediction of the meta-model based on the artificial neural network and the numerical model for the data in the design phase corresponded about 98%.