The geological condition of Ho Chi Minh (HCM) City is soft soil and high groundwater and includes two main structural layers such as Pleistocene and Holocene sediments. Therefore, deep excavation of all the high-rise buildings in the city is usually supported by concrete retaining walls such as the diaphragm or bored pile retaining walls. The system limits the excavation wall deflection during the construction process which could pose a potential risk to the construction and neighborhood areas. To estimate wall deformation at a highly accurate and efficient level, this study presents several machine learning models including feed-forward neural network back-propagation (FFNN-BP), long short-term memory (LSTM), bidirectional long short-term memory (Bi-LSTM), and support vector regression (SVR). The database for the experiment was obtained from a high building in HCM City, Vietnam. The database is deployed to implement the proposed algorithms in walk-forward validation technique. As a result, the Bi-LSTM model reduced prediction errors and improved robustness than the LSTM, FFNN-BP, and SVR models. Bi-LSTM, LSTM, and FFNN-PB could be used for predicting deep excavation wall deflection. In the meantime, not only could the estimated results support safety monitoring and early warning during the construction stages but also could contribute to legal guidelines for the architecture of deep excavations in the city’s soft ground.

The paper builds a physical model of testing in the laboratory with the parametric tempered glass box 0.5 0.5 1.2 m (length width depth) containing saturated clay to study the settlement and consolidation when loading increased gradually over time. The research covers herein to present the monitoring of settlement and pore water pressure, settlement calculation, numerical simulation using PLAXIS software V8.2 based on the results of soil physical and mechanical tests before and after loading in case of having or not prefabricated vertical drain (PVD). In case of no PVD, the calculation and numerical simulation using the soil parameters before loading have the differential settlement from the monitoring data, approximately 3.86 mm (10.45%), 0.41 mm (1.11%) respectively. Meanwhile, the deviation in the case using data after loading is about 2.29 mm (6.20%), 0.21 mm (0.56%) respectively. In case of PVD, the calculation and numerical simulation with the testing result of before loading deviation from the settlement monitoring by subsidence meter is 2.91 mm (7.88%), 44.42 mm (120.28%), calculation and simulation with the testing result of after loading deviation is 0.80 mm (2.17%), 1.26 mm (3.41%). In the case of having PVD, the difference in calculation, subsidence observation, and numerical simulation between the mechanical properties before and after loading is significant, when using the mechanical data after loading then the results are quite close to the subsidence of observation and simulation rather than before loading.

This study aims to identify the impacts of one-strut failure (OSF) during deep excavation in loose to medium-dense sand with respect to strut loads and wall deflections. The finite element software PLAXIS was used to carry out both 2D and 3D analyses. In addition, as two-dimensional (2D) analysis is commonly applied in engineering practice due to limited budget and time, the reasonable transferring of results obtained using 3D analyses to 2D analyses for OSF influences was explored. It was found that a failed strut located at the lowest level at the center of the excavation is the most critical case since it causes the most significant impacts on strut load transferring and additional wall deflection. In order to adequately demonstrate such impacts using 2D analyses, it is suggested that a reduction factor of axial stiffness of 2.5 should be adopted instead of removing an entire level of struts, as currently being used in 2D analyses of OSF.