Prediction on Ground Settlement Due to Pumping by a Hybrid Method

Groundwater pumping often serves as an auxiliary measure for the construction of urban underground projects, especially in areas with high groundwater level. The drawdown of groundwater level caused by pumping will increase the effective stress of the soil, which leads to compression. Nantong, a city in China, is a typical coastal city with thick water-rich sand strata. Groundwater pumping is required before implementation of an excavation. Excessive settlement caused by pumping would impose an adverse impact on surrounding environment and even threaten the safety of adjacent buildings. Therefore, it is necessary to predict the settlement due to groundwater pumping prior to excavation. In this study, a hybrid method combining numerical calculation and statistical learning is proposed to predict ground settlement caused by pumping. Based on the field pumping test, a calibrated model is established to obtain the water level variation with time. The obtained water level change, part of the settlement dataset measured at the site were used as the training dataset of the hybrid prediction model. Through statistical learning, a semi-empirical method to calculate the settlement caused by pumping in a water-rich sandy formation was developed. Besides, another set of ground settlement measured at the site was used as a test dataset to verify the accuracy of the model. Comparing the measured and predicted values of settlement, it is found that the method proposed exhibited good performance in predicting the ground settlement caused by pumping in water-rich sandy strata.