Post-Construction Settlement Prediction of Deep Lacustrine-Fluvial Soft Ground Improved by Vacuum Wellpoint Dewatering and Dynamic Compaction: An Engineering Case Study

Deep lacustrine–fluvial soft ground at the Songyanghu Terminal (Phase III), Yueyang Chenglingji Port, was treated using vacuum wellpoint dewatering combined with dynamic compaction. A two-dimensional coupled solid-mechanics-Darcy model in coupled COMSOL Multiphysics 6.3 model was used to examine deformation and excess pore-water-pressure response, while an archived 360-day settlement series was used to evaluate three empirical prediction methods. The field dataset contains 12 settlement observations at a constant 30-day interval from Day 30 to Day 360. The numerical outputs indicate surface settlement approaching approximately 100 mm by Day 100; at Day 10, localised excess pore-water pressure remained at roughly 100 kPa near the mid-depth drainage zone, while most of the model domain was substantially lower. To avoid treating the same observations as both fitting and validation data, the Asaoka, hyperbolic and exponential methods were recalibrated using data ending at Days 180, 240 and 300 and then evaluated against the subsequent withheld observations. Across these three windows, the mean RMSE values were 0.02, 0.59 and 0.19 mm for the Asaoka, hyperbolic and exponential methods, respectively. The Day-360 measurement of 108.84 mm is therefore used as a reference observation rather than as a proven final settlement. For this single site and dataset, the Asaoka method produced the lowest numerical withheld-data errors; however, the differences between the Asaoka and exponential predictions are smaller than the stated ±1 mm levelling accuracy and therefore do not establish statistically or physically significant superiority. Broader applicability requires validation at additional sites and with independent monitoring datasets. These results provide a case-specific basis for post-construction settlement assessment of similarly treated soft ground.