基于支持向量机的多年冻土路基融沉变形预测  

Thaw Settlement Prediction of Permafrost Subgrade Based on Support Vector Machine

作　　者：孙全胜[1] 常继峰[1] 

机构地区：[1]东北林业大学土木工程学院,黑龙江哈尔滨150040

出　　处：《公路工程》2014年第5期136-140,148,共6页Highway Engineering

摘　　要：多年冻土路基热稳定性差,工后沉降量大,路基病害较为严重。如果能够准确预测该类路基的工后融沉值,就可为工程建设提供重要的参考依据,从而提高该类路基的路用性能。为此,在对现有预测模型应用效果分析的基础上,首次将支持向量机应用于多年冻土路基融沉变形的预测中,提出了一种有效可行的新型预测方法,并以实际工程为依托,构建了基于支持向量机原理的多年冻土路基融沉变形预测模型。通过与实测值及其它预测模型的对比分析表明:该模型在预测过程中有效的避免了"过拟合"及"维数灾难",人为干预较少,具有预测精度高,泛化能力强,预测结果稳定的特点,成功的解决了多年冻土路基影响因素多,样本数量少等带来的预测难题。The permafrost subgrade has the poor thermostability and larger post-construction settlement,the disease of subgrade is more serious. If the post-construction thaw settlement can be predicted accurately for this subgrade,the important reference may be provided for the engineering construction. So the pavement performance of this subgrade would be improved greatly. Therefore,an effective and feasible new method for prediction is put forward by analyzing the application effects of some existing prediction models. This paper applied support vector machine to the thaw settlement prediction of permafrost subgrade for the first time,and seted up the corresponding prediction model based on practical project.Compared with the measured values and other prediction models,the analysis shows that the model can avoid＂overfitting＂and＂curse of dimensionality＂in the prediction process,the human intervention is also less,it has the merits of higher prediction precision,stronger generalization ability and better predictions stability,the prediction difficulties caused by more effect factors and smaller sample size of the permafrost subgrade is solved successfully.

关 键 词：多年冻土路基 融沉变形 支持向量机 预测模型 对比分析 

分 类 号：U416.168[交通运输工程—道路与铁道工程]