Predicting the maximum ground subsidence (Smax) in the construction of soil pressure balanced shield tunnel, particularly on soft foundation soils, is essential for safe operation and to minimize the possible risk of damage in urban areas. Although some research has been done, this issue has not been solved because of its complexity and many other influencing factors. Due to the increasing accuracy of machine learning (ML) in predicting surface deformation of shield tunneling and the development of automated machine learning (AutoML) technology. In the study, different ML prediction models were constructed using an open source AutoML framework. The prediction model was trained by the dataset, which contains 14 input parameters and an output (i.e., Smax). Different AutoML frameworks were employed to compare their validities and efficiencies. The performance of the model is estimated by contrasting the prediction accuracy parameters, including root mean square error (RMSE), mean absolute error (MAE) and determinant coefficient (R²).With a coefficient of determination (R²) of 0.808, MAE of 3.7, and RMSE of 5.2 on the testing dataset, the best prediction model i.e., extra tree regressor showed better performance, proving that our model has advantages in predicting Smax. Furthermore, the SHAP analysis reveal that the soil type (ST), torque (To), cover depth (H), groundwater level (GW), and tunneling deviation have a significant effect on Smax compared to other model inputs.

Accurately forecasting tunnel-induced ground settlement is crucial for mitigating risks to urban infrastructure and ensuring the safety of tunnelling operations. This study introduces advanced predictive frameworks that incorporate enhancements to both deep learning (DL) models and statistical techniques to handle the intricate, nonlinear, and dynamic characteristics of settlement data. The proposed DL models, Convolutional Long Short-Term Memory (Conv-LSTM2D) and Convolutional Gated Recurrent Unit (Conv-GRU2D), extend traditional LSTM and GRU architectures with 2D convolutional mechanisms to capture complex spatiotemporal dependencies. Additionally, the statistical Autoregressive Integrated Moving Average (ARIMA)/Seasonal ARIMA (SARIMA) models were enhanced through seasonality removal, automated model selection using the auto_arima algorithm, and parameter fine-tuning via grid search to improve their predictive accuracy. Extensive evaluations using real-world sensor datasets from tunnelling projects demonstrated that the proposed Conv-GRU2D model achieved superior performance, with a R-squared of 0.94, an RMSE of 0.12, and MAPE of 0.32, surpassing both Conv-LSTM2D and optimized ARIMA/SARIMA models. These results highlight the effectiveness of convolutional mechanisms in deep learning and the advantage of optimized ARIMA/SARIMA models in capturing complex temporal patterns and improving forecasting precision for optimizing ground settlement. Notably, our findings demonstrate that the proposed models outperform existing methods, providing superior prediction accuracy and computational efficiency, even with limited training data. This underscores the potential of these innovative methods to significantly enhance the reliability of ground settlement forecasts in urban tunnel construction, offering valuable insights that can further advance geotechnical engineering practices and contribute to safer and more efficient tunnelling operations in urban environments.

The authors wish to make the following corrections to this paper