Machine learning and FEM-driven analysis and optimization of deep foundation pits in coastal area: A case study in Fuzhou soft ground

Abstract

To comply with the requirements of sustainable energy development, China has proposed the strategic goal of achieving dual carbon. Systematic and scientific development and utilization of urban underground space will provide critical support for reducing carbon emissions and enhancing carbon sink capacity. This paper examines the transmission and distribution ring pit project of Fuzhou Binhai New City, China, divided into four regions, where the selection of the support system is determined by the project’s characteristics. Stability is analyzed using in-situ monitoring data from the R4 area, and the deformation of the support system is predicted using machine learning. The predicted maximum lateral deformation of the support system may reach the warning value, necessitating corrections to the existing support parameters. On this basis, the deformation during foundation pit excavation is simulated, and the effects of key factors such as pile geometric parameters, pile penetration depth, and anchor cable insertion ratio on the deformation are analyzed. The study shows that pile deformation control is optimal when the support parameters include a 1.3 insertion ratio, a 20° anchor cable angle, and a 200 kN prestressing force, enabling the construction of the remaining three areas. This study can serve as a valuable reference for the design and analysis of deep foundation pits under special stratigraphic conditions in coastal areas.