基于多維數(shù)據(jù)轉(zhuǎn)換的混凝土壩異常值智能識別方法

打開文本圖片集

關(guān)鍵詞：混凝土壩；異常值識別；數(shù)據(jù)轉(zhuǎn)換；深度學(xué)習(xí)；CWT；SVM中圖分類號：TV62 文獻標(biāo)志碼：A doi：10.3969/j.issn.1000-1379.2025.07.021[用格式：蘇曉軍，許增光，張野，等.基于多維數(shù)據(jù)轉(zhuǎn)換的混凝土壩異常值智能識別方法[J].人民黃河，2025，47（7）：131-136，143.

Intelligent Identification Method for Abnormal Values of Concrete Dams Based on Multi-Dimensional Data Transformation

SU Xiaojun’， XU Zengguang'， ZHANG Ye1， ZHONG Wen'， ZHOU Tao2，YANG Tao3，LI Kangping4 （1.State KeyLaboratoryof WaterEngineering Ecology and Environment in Arid Area，Xi'an Universityof Technology， Xi'an 710048，China；2.Huanghe Hydropower Development Co.，Ltd.，Xining 810000，China; 3.ChinaYangtzePowerCo.，Ltd.，Yibin644612，China；4.PowerChinaNorthwestEngineringCo.，Ltd.，Xi’an710065，China） Abstract：AnoalydetectioinefoatiomoitogdataofocetedamsisaeyspectofstructuralhalthdgosisHoeaditionalmethodsallsuffrosteetoaucideaeractoapalisdlofiaal nosis.Indertdestseisssiserproposdanintellgnaomalyetectionetdforoetedssdouli sionaldataafoatiodatprogtcacdfefoaleeiostltodoati monitoringdataofconcretegravitydamintooalmeasurementvaluesandfourtypicalaboalmeasurementvaluesofjump-tye，steptype，platfo-tadlto-te.Tnotiuuswavelettrasfoasudtooertodmesioaldfoatiotiesdata intotwo-dimesaliagsithfqcyjinteaus，utrafongtoalyeetiotoagecasificatioiuia high-dimensional feature space.By leveraging the Inception-ResNet- σ?v2 deep learning image pre-training model，the method extracted abnormalfeaturesfromtheimages.Fallasupprtvectormachine（SVM）wasusedtobuildananomalydetectionmodelforteconcetedam smeasurementvaluesEngineringcasevalidationshowsthattheproposedmethodexhibitsoutstandingperformanceinanomalydetection tasks，with accuracy，precision，recall and F1 score reaching 97% ， 98% ， 96% and 97 % respectively，all outperforming the comparison models.Tereseachotolyonfisteeancementfectoftifrequncyiagerepresentationondeforationanomalyfatureutalsodemonstratestedvantagsofeelaingingengeingmoiigdata.Teesultsprovideaitellgnandfntsoution for deformation anomaly detection in concrete dams.

Key words：concrete dam;abnormal value identification; data transformation；deep learning；CWT； SVR

0 引言

混凝土壩作為重要的水利樞紐工程，其長期安全運行直接關(guān)系到下游人民的生命財產(chǎn)安全以及社會經(jīng)濟的穩(wěn)定[1-2]。（剩余8791字）