Due to the great difference between the actual post-construction settlement and initial design settlement ,the analysis of the measured settlement data is needed to predict the late settlement .In order to improve the accuracy of high-way’s soft foundation settlement prediction ,the LIB -SVM model is used to forecast the foundation settlement ,and through using the cross validation to select the optimal parameters combination ,the LIB -SVM model could overcome the blindness of SVM model in parameter selection .Aimed at instances ,the LIB - SVM model’s prediction value and exponential model’s prediction value are compared with the field measuring values .The results show that the LIB - SVM model has higher prediction accuracy than exponential model ,so the method is feasible in the practical prediction of settlement .%由于地基实际工后沉降与初始设计沉降往往存在很大差异,因此,需要通过分析现场实测沉降资料预测后期沉降。为了进一步提高公路软基沉降预测的准确性,将 LIB - SVM 模型应用于地基沉降, LIB - SVM 通过交叉验证选取的最优参数组合克服了传统 SVM 模型参数选择的盲目性。根据实例,将LIB - SVM 模型预测值和指数模型预测值与现场量测值进行了对照。结果表明,LIB - SVM 模型比指数模型有较高的预测精度,该方法在沉降的实际预测中具有可行性。
展开▼