Soft computing approach for prediction of surface settlement induced by earth pressure balance shield tunneling

Estimating surface settlement induced by excavation construction is an indispensable task in tunneling, particularly for earth pressure balance (EPB) shield machines. In this study, predictive models for assessing surface settlement caused by EPB tunneling were established based on extreme gradient boosting (XGBoost), artificial neural network, support vector machine, and multivariate adaptive regression spline. Datasets from three tunnel construction projects in Singapore were used, with main input parameters of cover depth, advance rate, earth pressure, mean standard penetration test (SPT) value above crown level, mean tunnel SPT value, mean moisture content, mean soil elastic modulus, and grout pressure. The performances of these soft computing models were evaluated by comparing predicted deformation with measured values. Results demonstrate the acceptable accuracy of the model in predicting ground settlement, while XGBoost demonstrates a slightly higher accuracy. In addition, the ensemble method of XGBoost is more computationally efficient and can be used as a reliable alternative in solving multivariate nonlinear geo-engineering problems.

Automated summary

This study established predictive models for assessing surface settlement induced by EPB tunneling using different soft computing techniques and validated them with datasets from three tunnel construction projects in Singapore. The results showed that the XGBoost model had slightly better accuracy in predicting ground settlement and was more computationally efficient.