Study on settlement prediction methods for high embankments in mountainous areas

Due to the influence of special geological conditions and economic factors, the construction of highways in the mountainous areas of Guangxi has employed a large number of high embankment structures. Characterized by significant settlement amounts and long settlement cycles, these high embankments are prone to issues such as cracking and uneven settlement. This study utilizes three prediction methods—exponential curve model fitting, grey system theory, and backpropagation neural network (BPNN)—to forecast the settlement of a high embankment section of a highway in Guangxi. The prediction results from these three methods are compared with actual measured values. The findings indicate that among the three methods, the BPNN achieves the best overall fitting performance; the grey system theory meets accuracy requirements while being the least affected by different spatial locations; and the exponential curve fitting method involves lower computational costs but shows greater dependence on parameter selection, yet its accuracy improves significantly as the time interval increases. The results provide a methodological basis for the settlement prediction and control of high embankments in mountainous expressways, offering practical significance for ensuring their long-term operational safety.