Probabilistic Framework for Ground Movement Induced by Shield Tunnelling in Soft Soil Based on Gap Parameter

Numerical simulation and machine learning-based methods are frequently adopted when performing ground movement probabilistic analyses, considering the various uncertainties during shield tunnelling. However, numerical simulation takes time, while machine learning lacks interpretation somehow. New methods fully reflecting mechanisms and taking advantage of field data should be proposed and applied in probabilistic analysis. This study proposes a probabilistic framework from the mechanism and data aspect based on the GAP parameter. Solutions for three components of the GAP parameter are first improved through different methods. Coupling the uncertainty of the input parameters, a probabilistic framework estimating the risks from both mechanistic and data insights is then established. Furthermore, the spatial variability in soft soil is considered in the framework by calculating the equivalent parameters. Through an analysis of a practical case, the results show that the measured data can fall within the 95% confidence interval of the predicted displacement samples. The median of the predicted samples is highly consistent with the measured value, and by considering the spatial variability in soil, the results can be more accurate. As a result, the proposed probabilistic framework is verified as practically applicable when predicting ground movement while considering multiple uncertainties.