Prediction and Engineering Verification of Building Settlement in Loess High-Fill Areas

To identify suitable settlement prediction methods for buildings constructed in loess high-fill areas, this study analyzes settlement monitoring data from a residential building in the Yan’an New District. Considering the pronounced compressibility of loess high fills and the depth-sensitive characteristics of differential settlement, four prediction approaches—the exponential curve method, the hyperbolic method, the code-based method, and the finite element method—are employed to forecast settlement. Fitting functions are established for each method to extrapolate the ultimate settlement. The research results indicate that, within the monitoring period, the exponential curve method and the hyperbolic method exhibit higher agreement with the measured settlement data, with prediction errors of 2.56% and 0.62%, respectively. These errors are significantly lower than those of the code-based method and the finite element method, which reach 45.6% and 77.4%, respectively. In particular, the hyperbolic method, through adaptive parameter iteration, controls the discrepancies between calculated and measured settlements at all monitoring points within a range of 0.62% to 8.89%. This method is therefore capable of more accurately capturing the settlement evolution characteristics, consistent with the long-term creep behavior of loess high-fill foundations, and provides a feasible reference method and practical decision-making support for building settlement prediction and engineering design in loess high-fill areas.