公路工程
公路工程
공로공정
JOURNAL OF HIGHWAY ENGINEERING
2014年
5期
136-140,148
,共6页
多年冻土路基%融沉变形%支持向量机%预测模型%对比分析
多年凍土路基%融沉變形%支持嚮量機%預測模型%對比分析
다년동토로기%융침변형%지지향량궤%예측모형%대비분석
permafrost subgrade%thaw settlement%support vector machine%prediction model%comparative analysis
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。
多年凍土路基熱穩定性差,工後沉降量大,路基病害較為嚴重。如果能夠準確預測該類路基的工後融沉值,就可為工程建設提供重要的參攷依據,從而提高該類路基的路用性能。為此,在對現有預測模型應用效果分析的基礎上,首次將支持嚮量機應用于多年凍土路基融沉變形的預測中,提齣瞭一種有效可行的新型預測方法,併以實際工程為依託,構建瞭基于支持嚮量機原理的多年凍土路基融沉變形預測模型。通過與實測值及其它預測模型的對比分析錶明:該模型在預測過程中有效的避免瞭“過擬閤”及“維數災難”,人為榦預較少,具有預測精度高,汎化能力彊,預測結果穩定的特點,成功的解決瞭多年凍土路基影響因素多,樣本數量少等帶來的預測難題。
다년동토로기열은정성차,공후침강량대,로기병해교위엄중。여과능구준학예측해류로기적공후융침치,취가위공정건설제공중요적삼고의거,종이제고해류로기적로용성능。위차,재대현유예측모형응용효과분석적기출상,수차장지지향량궤응용우다년동토로기융침변형적예측중,제출료일충유효가행적신형예측방법,병이실제공정위의탁,구건료기우지지향량궤원리적다년동토로기융침변형예측모형。통과여실측치급기타예측모형적대비분석표명:해모형재예측과정중유효적피면료“과의합”급“유수재난”,인위간예교소,구유예측정도고,범화능력강,예측결과은정적특점,성공적해결료다년동토로기영향인소다,양본수량소등대래적예측난제。
The permafrost subgrade has the poor thermostability and larger post-construction settle-ment,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 feasi-ble new method for prediction is put forward by analyzing the application effects of some existing predic-tion 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 al-so less,it has the merits of higher prediction precision,stronger generalization ability and better predic-tions stability,the prediction difficulties caused by more effect factors and smaller sample size of the per-mafrost subgrade is solved successfully.