Page 133 - 《水资源与水工程学报》2025年第1期
P. 133
!36 "!1 # & ' ( ) & * + , - Vol.36No.1
2025 $ 2 % JournalofWaterResources&WaterEngineering Feb.,2025
DOI:10.11705/j.issn.1672-643X.2025.01.14
'/ GNSS 0123456789:;
<=>?@ABCDE
1,2 1,2 1,2 1,2 1,2
!"# , $%& , ' ( , )*+ , , -
(1. ?@VTL& WXWY&Z, >? ?@ 710048;2. ?@VTL& [\WT]
E^_`N!-.abc, >? ?@ 710048;)
F G: defghijWXT]de@Nklm-I45QR, NnFopqrs(GNSS) UtdefgOim
-Iuvwx, yz{m|}~hidem@N
-I。 GNSS z{Vfg
him, x'fk^k(VMD)、 (CNN)、 ¡¢£¤(BiLSTM) ~
¥¦§¨©(AM) mde GNSS fgz{ªShi«。X¬ VMD®fg¯°±²ªV, ³´µ
¶·¸, z{|}; ¹¬ CNNJº VMDV3m»¼½f¾¿, ÀÁ½Â»mÃÄ~NÃ
Å; Ƭ BiLSTM«ÇÈÉ CNN¼m½f¾¿, ʺÀÁ½Â»mËÌ~ÍÎÏÐ, ÑfgmÇ
ÈÒ。tÓÔÕ&Öm-I×z, ØÙÇÚÛ(SA) Üݯ°ÞÓkl, Jßà«máâã~h
i;BiLSTMÑmäåʺ AMNæçè, äÈ¥¦§¨©¯xéhiäåmêë, ÞìíîhiÒ。
ïðñÈ«ò¬óWXT]dem GNSS Oiz{hi, äåôõ: SöÕ&Ö÷È«øù, mñ
È«ú 3 ûü(h、x、y) ýmhikþÿ!à 63.92%、62.06% ~ 89.10%, tWXT]de GNSS O
iz{kl"#m$üÝ。
HIJ: defg; hi«;GNSS; ½hi; Õ&Ö
KLMNO:TU43 PQRST:A PUVO:1672643X(2025)01012909
Deformationpredictionmethodsforhydraulicengineeringslopes
basedontheintegrationofGNSSanddeeplearning
1,2
1,2
1,2
1,2
1,2
SONGJintao ,ANChence ,YANGJie ,MAChunhui ,TONGFei
(1.InstituteofWaterResourcesandHydroElectricEngineering,Xi’anUniversityofTechnology,Xi’an710048,China;2.StateKey
LaboratoryofWaterEngineeringEcologyandEnvironmentinAridArea ,Xi’anUniversityofTechnology,Xi’an710048,China)
Abstract:Slopedeformationpredictionisanimportantresearchfieldinslopesafetyanalysisofhydraulic
projects.TheGlobalNavigationSatelliteSystem(GNSS)technologyisoneofthemainapproachesfor
slopedeformationmonitoring,itsdataqualityandpredictionaccuracyarecrucialforslopesafetyevalua
tion.InresponsetotheissuesofnoiseprocessinginGNSSdataandhighprecisionpredictionofdeforma
tionsequences,adenoisingandpredictionmodelwasproposedbasedonvariationalmodedecomposition
(VMD),convolutionalneuralnetworks(CNN),bidirectionallongshorttermmemory(BiLSTM)net
worksandattentionmechanism(AM).Toimprovedataquality,VMDisemployedtofilteranddenoise
theoriginaldeformationsequence ;CNNisutilizedtoextracttimevaryingfeaturesfromtheprocessedda
ta,whichisthenfittedbyBiLSTMtoobtainthedeformationvaluesbasedonthehistoricalandfuturein
formationinofthetimeseries.Tooptimizethemodelparametersandtherebyimprovepredictionaccura
cy ,thesimulatedannealing(SA)optimizationalgorithmisintroducedfortheoptimalanalysisofimpor
tantparametersinthedeeplearningnetwork.Subsequently ,theoutputofBiLSTMmodelisprocessedin
!"#$:20240913; %$:20241203
'()*: !"#$%&'()*'(+,(52109166、52409173);!"#$%&'(-.+,(52039008);!"/0
123456789(GZC20232140)
+,-.: :;<(1990—), =, >??@6, 12, ABC, D2EFG, HIJKLM@NOPQRB&S%4TU。