Page 130 - 2023年第54卷第4期
P. 130
selectionwithPCA - RF[J].JournalofCivilStructuralHealthMonitoring,2022,12(3):557 - 578.
[24] 韩腾飞,李亚平.基于 Stacking集成学习的剩余使用寿命预测[J?OL].计算机集成制造系统:1 - 18[2022 -
12 - 04].http:??kns.cnki.net?kcms?detail?11.5946.TP.20220314.1224.016.html.
[25] 王惠文,孟洁.多元线性回归的预测建模方法[J].北京航空航天大学学报,2007(4):500 - 504.
[26] 方匡南,吴见彬,朱建平,等.随机森林方法研究综述[J].统计与信息论坛,2011,26(3):32 - 38.
[27] 丁世飞,齐丙娟,谭红艳.支持向量机理论与算法研究综述[J].电子科技大学学报,2011,40(1):2 - 10.
[28] 刘波,张斌,喻佳,等.基于多元线性回归模型的大坝变形预报研究[J].人民长江,2010,41(20):53 - 55.
[29] 罗浩,郭盛勇,包为民.拱坝变形监测预报的随机森林模型及应用[J].南水北调与水利科技,2016,14
(6):116 - 121.
[30] 张志威,戴妙林,刘晓青,等.基于 GA - SVM 的 碾 压 混 凝 土 重 力 坝 参 数 反 演 [J].人 民 黄 河,2020,42
(6):112 - 116.
[31] JONESB,JOHNSONRT.DesignandanalysisfortheGaussianprocessmodel[J].QualityandReliabilityEngi
neeringInternational ,2009,25(5):515 - 524.
[32] 刘紫亮,居翔,张 永 芳,等.基 于 改 进 随 机 搜 索 算 法 的 随 机 森 林 调 参 优 化 [J].网 络 安 全 技 术 与 应 用,
2022(4):49 - 51.
[33] 李峰,舒斐,李明轩,等.基于深度学习的 Linux远控木马检测[J].计算机工程,2020,46(7):159 - 164.
[34] 贾俊平,何晓群,金勇进.统计学[M].7版.北京:中国人民大学出版社,2018.
Combinedpredictionmodelofdam deformationbasedonmulti - factorfusion
andStackingensemblelearning
1
1,2,3
1
1
1
WANGRuijie,BAOTengfei ,LIYangtao,SONGBaogang,XIANGZhenyang
(1.CollegeofWaterConservancyandHydropowerEngineering,HohaiUniversity,Nanjing 210098,China;
2.StateKeyLaboratoryofHydrology - WaterResourcesandHydraulicEngineering,HohaiUniversity,Nanjing 210098,China;
3.CollegeofHydraulic&EnvironmentalEngineering,ChinaThreeGorgesUniversity,Yichang 443002,China)
Abstract:Deformationistheintuitivereflectionofthechangeofdams’operatingbehavior.It’scrucialtobuilda
deformationpredictionmodelwithhigherefficiencyandaccuracyfordam structuralsafetymonitoring.Traditional
single - factorandsingle - algorithmpredictionmodelsinevitablyhaveseriesofproblemssuchasinsufficientgenerali
zationabilityandpoorrobustness ,whichwillinducedeviationsandevenmisjudgments.Tosolvethisproblem,
thispaperselectsdifferentdeformationinterpretationfactorsandregressionalgorithmstobuildmultiplesingle - factor
single - algorithmpredictionmodels.Next ,thesemodelsareintegratedtoproposeacombineddamdeformationpre
dictionmodelthroughStackingensemblelearning.Detailedly ,thiscombinedmodeladoptsGaussianProcessRe
gressionasthemeta - learnerandintegratesthesingle - factorsingle - algorithm modelsfrom algorithm andfactor
thesetwoaspects.Toreducetheriskofoverfitting ,k - foldcross - validationisalsointroducedingeneratingthenew
dataset.Referringtothedeformationdataofaconcretearchdam ,themodel’saccuracyandeffectivenesshave
beenevaluatedbymulti - modelconstructionandperformancecomparison.Theresultsshowthatthesingle - factor
single - algorithmmodelsarecharacterizedbyaccuracyanddiversity.Throughtheintegrationofalgorithmsandfac
tors ,thepredictionaccuracyandrobustnesshavebeensignificantlyimproved,andthepredictioncapabilityofthe
modelshasbeeneffectivelyenhancedduringthewater - fluctuatingperiod.Aboveall,thecombineddamdeforma
tionpredictionmodelhasexcellentnonlinearinformationminingabilityandpredictiveperformance ,andcouldpro
videareliablebasisfordamsafetymonitoring.
Keywords:multi - factorfusion;damsafetymonitoring;predictionmodel;Stackingensemblelearning;Support
VectorMachine ;RandomForest
(责任编辑:韩 昆)
— 5 0 —
6