Calculation and Intelligent Prediction of Long-Term Subgrade Settlement on Soft Soil Interlayer Foundations Under Secondary Consolidation in the Yellow River Floodplain

Highways constructed on stratified foundations with thick soft soil interlayers in the Yellow River floodplain of Shandong Province have experienced long-term settlement. However, accurately predicting subgrade settlement caused by the secondary consolidation of soft soils remains a major engineering challenge. In this study, PLAXIS 3D numerical simulation was combined with a neural network model to predict the long-term temporal and spatial settlement behavior of highway subgrades. The results show that the soft soil creep (SSC) constitutive model better represents the consolidation process of the soft soil interlayer than the soft soil (SS) model. A decrease in permeability will prolong the dissipation time of excess pore water pressure and the settlement stabilization time, leading to an increase in the proportion of post-construction settlement in the total settlement. The final settlement increases linearly with the thickness of the soft soil interlayer and embankment height, while it decreases following a power-law function with increasing interlayer burial depth. By comprehensively considering the combined effects of multiple factors, a genetic algorithm–optimized backpropagation neural network (GA-BP) model was developed. The testing dataset achieved a root mean square error (RMSE) of 0.01488 m, a mean absolute percentage error (MAPE) of 7.0562%, and a coefficient of determination (R2) of 0.9706, demonstrating the model’s ability to achieve intelligent full-period and full-section settlement prediction for subgrades with soft soil interlayers. Overall, this study developed an intelligent framework for predicting long-term settlement in subgrades with soft soil interlayers, offering practical guidance for evaluation and timely settlement control.