Predictions of excavation responses frequently differ from monitoring data due to geotechnical uncertainties. This paper proposes an efficient Bayesian updating approach for excavation responses that considers the uncertainties of soil properties and the calculation model. To evaluate the depth-dependent characteristic of model uncertainty, the model factor is quantified by a constant part and a trending component. Bidirectional long short-term memory (BiLSTM) neural networks are constructed to act as a substitute for the finite-element method to achieve higher computational efficiency. An excavation project in Taipei, Taiwan, is used in this study to illustrate the proposed approach. The results demonstrate that the BiLSTM successfully learns the mapping between soil parameters and deflection responses. The uncertainties of key soil parameters and model factors are significantly reduced when observed deflections at multiple points are incorporated on a stage-by-stage basis. The trending component of the model factor plays an essential role in the early stages, but its impact decreases as the excavation progresses. The prediction intervals using the updated parameters generally cover the monitoring data. The proposed method can rapidly update and improve the predictions of subsequent responses once the monitoring data is obtained. This means early remedial actions can be taken and construction safety ensured.
29 April 2021 in International Journal for Numerical and Analytical Methods in Geomechanics
This paper develops a hierarchical Bayesian model (HBM) that integrates the physical knowledge and the test data to predict the small-strain shear modulus Gmax for a target sand type. The limited target-specific data are combined with the abundant generic data through a hierarchical structure so that the variability of Gmax within one sand type and across different sand types can be captured. The hyperparameters that characterize the same underlying distribution of physical model parameters across all the sand types are first estimated from the abundant generic data. The model parameters for the new sand type are then updated as the limited site-specific data become available. The approach is illustrated using a generic database and two real examples not covered by the generic database. Multiple possible hierarchical models are compared in terms of model complexity and goodness-of-fit. The results show that the hierarchical modeling of small-strain shear modulus data is reasonable and necessary. The hierarchical model can provide less biased and more accurate predictions of Gmax compared to the commonly used complete pooling model, especially in cases where the site-specific data are quite different from the overall average of the generic database.