Abstract: Soft soil subgrade of highway will have large settlement after the completion of construction，and it may produce an impact on the vehicle traffic at the post stage.Therefore，the research on the prediction for soft soil subgrade settlement was carried out.According to the settlement data detected in practical engineering，a non-equidistance grey Verhulst model was established by using the scientific computing packages of python.Then the SA algorithm (simulated annealing algorithm) was applied to optimize the model，and the optimal solution of model parameters as well as final settlement was obtained.Meanwhile，the settlement prediction curve of subgrade changing with time was given.Comparative analysis and error analysis with the curve of conventional non-equidistance grey Verhulst model was carried out.The result shows that the model optimized with SA algorithm can better reflect the settlement regulation of soft soil subgrade，which can be applied to the settement prediction of soft soil subgrade.

Key words: highway engineering, subgrade settlement prediction, non-equidistance grey Verhulst model, soft soil subgrade, SA algorithm

摘要： 公路软土路基在施工完成后会产生较大的工后沉降，对后期的车辆通行有一定的影响，因此对软土路基的沉降预测进行了研究。根据实际工程中所 测得的沉降数据，利用python编程语言的科学计算库，建立了非等时距灰色Verhulst模型。采用SA算法(模拟退火算法)对模型进行优化，求得模型参数的 最优解和最终沉降量，给出了路基随时间变化的沉降预测曲线，并与传统非等间隔灰色Verhulst模型曲线进行了对比和误差分析。结果表明：经SA算法优 化后的模型能更好地反映软土路基的沉降规律，可较好地应用于软土路基的沉降预测研究中。

关键词: 道路工程, 路基沉降预测, 非等时距灰色Verhulst模型, 软土路基, SA算法