泥石流平均流速预测模型及敏感因子研究.pdf

《泥石流平均流速预测模型及敏感因子研究.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、JournalofEngineeringGeology工程地质学报1004—9665／2014／22(3)-0355—06DOI：i0．13544／j．cnki．jeg．2014．03．001泥石流平均流速预测模型及敏感因子研究术于国强张茂省王根龙(国土资源部黄土地质灾害重点实验室，中国地质调查局西安地质调查中心西安710054)摘要为了探求泥石流平均流速敏感因子及影响因素耦合关系，本文采用BP神经网络和支持I~W：ql,模型对蒋家沟泥石流数据进行预测，对两种泥石流平均流速预测模型的学习与泛化能力进行比较，并对平均

2、流速各影响因素的敏感程度进行分析，建立了泥石流平均流速敏感因子预测模型。结果表明：支持向量机的泛化能力优于BP网络，更适合样本数量较少的泥石流动态预测。沟道比降和不稳定层厚度是泥石流平均流速的主要影响因子，各因子之间存在复杂的耦合关系。基于不稳定层厚度和泥面比降的泥石流平均流速预测模型精度较高，能够定量描述泥石流动态与影响因子间的响应关系。研究成果可为泥石流防治提供依据。关键词泥石流蒋家沟平均流速预测模型敏感因子中图分类号：P642文献标识码：APREDICTIoNMoDELANDSENSITIVEFACToRSF

3、oRAVERAGESPEEDoFDEBRISFLOWSATJIANGJIAGULLYYUGuoqiangZHANGMaoshengWANGGenlong(KeyLaboratoryforGeo-hazardsinLoessArea，MLR，anCenterofGeologicalSurvey，ChinaGeologicalSurvey，Xian710054)AbstractThispaperinvestigatesthesensitivefactorsoftheaveragespeedofdebrisflowsan

4、dcouplingrelationshipoftheinfluencefactorswiththedataofdebrisflowsinJiangjiagully．TheBPANNandSVMareproposedforbuildingpredictivemodelofaveragevelocityofdebrisflows．Predictionaccuracyandscopeofapplicationofthetwomodelsarecompared．AnalysisisdoneOfvariousfactorsf

5、ortheirsensitivitiestoaveragespeedofdebrisflowswiththedefaultfactormethod．Apredictionmodelfordebrisflowsisestablishedandverified．TheresultsshowthatthepredictionaccuracyoftheSVMmodelisbetterthanthatoftheBPANNmodelinthevalidationphase．Theextrapolatingabilityandp

6、redictingcapabilityhavebeenvalidatedbyusingthesuppo~vectormachinepredictionmodel，andtheSVMmodelismoresuitableforitspredictionofdebrisflow．Channelslopeandthethicknessoftheunstablelayerarethemainsensitivefactorsaffectingaveragevelocityofdebrisflows．Theinteractio

7、namongstvariousfactorsformedcouplingrelationshipunderthecomplicatedcondition．Therelationshipmodelofaveragespeedofdebrisflowsestablishedbaseuponchannelslope，andthethicknessoftheunstablelayercanwellexpressquantitativelytheresponserelationshipbetweendebrisflowsdy

8、namicandvariousfactorswithsufficienthighaccuracy．Itisanaccuratescientificbasisforpreventionofdebrisflow．KeywordsDebrisflow，Jiangjiagully，Averagevelocity，Predictionmodel，Sensitivefa