Abstract
Subway operations generate loads that can cause uneven soil settlement around tunnels, impacting the safety of subway train operations. Existing research often simplifies the analysis of subway loads and modeling, which does not accurately reflect the variations in passenger flow and the changing characteristics of the surrounding soil. To address this, our study employs data mining methods to analyze patterns in subway passenger flow, categorizing it into four operational conditions: full-load, high-load, medium-load, and low-load periods. Based on these categories, we developed a differential settlement prediction model that takes into account variations in passenger flow at both the tunnel location and its symmetry center. Our findings indicate: (1) In twin tunnels, the displacement at the symmetry center of the tunnels is 41.49% of that beneath the tunnel itself, with a notable 84% difference in tunnel displacement between full-load and low-load periods. It is crucial in long-term foundation settlement studies to consider the settlement at the symmetry center and the impact of fluctuating passenger flows on subway loads. (2) The accuracy of the differential prediction model, which incorporates passenger flow variability, surpasses that of traditional models, making it suitable for long-term settlement studies of subway tunnels. (3) After twenty years of operation on Shanghai Metro Line 1, the settlements caused by metro operations under the tunnel and at the symmetry center are 20.73 millimeters and 7.20 millimeters, respectively. These settlements are primarily due to cumulative plastic strain, which progresses more rapidly in the first five years and occurs mostly within 10 m below the tunnel.