基于深度自编码器的混凝土坝变形异常检测模型

打开文本图片集

关键词：混凝土坝；变形监测；异常检测；自编码器；深度学习中图分类号：TV642 文献标志码：A doi：10.3969/j.issn.1000-1379.2025.07.022用格式：，，，等.基于深度自编码器的混凝土坝变形异常检测模型[J].人民黄河，2025，47（7）：137-143.

Concrete Dam Deformation Anomaly Detection Model Based on Deep Autoencoder

KANG Xinyu1，LI Yanlong'，ZHANGYe1，ZHOU Tao²，ZHONGWen’，YANG Tao³ （1.StateKeyLaboratoryofWaterEgineingEcologndEvoeninAridAra，XianUivesityofTcholog，Xianina; 2.Huanghe Hydropower DevelopmentCo.，Ltd.，Xining 810o0，China；3.ChinaYangtzePowerCo.，Ltd.，Yibin 644612，China） Abstract：Anunsupervisedanomalydetectionmodelbasedonadputoencoderwasproposedtoaddressanomalousreadingsinconcrete damdeformationmonitoring，withtheobjectiveofenhancingdetectionacuracyandautomation.Theautoencoderwas trainedinanusuper visedmaneronnoraldefomationdatatolearlow-dimensionalfeaturerepresentationsandwassubsequentlyemploedtorebuildcoming measurements.Measurementsexhitingsignificantdeviationsetwenobservedandrebuildvalueswereclasifiedasanomales.Teresult showsthat theproposedmodelachievesover97%accuracyianomalydetectionandisdemonstratedtoperforeliablyundervarioutesting conditions.Conseqentlytheeutocoderbdroachisableofectielydentigfoatiaolsioeda， exhibiting robust and precise detection capabilities.

Key words： concrete dam； deformation monitoring;anomaly detection；autoencoder; deep learning

0 引言

混凝土项具有卓越的承载能力、优异的耐久性及良好的适应性，已成为水库大坝建设的首选类型[1-4]然而，在混凝土坝长期服役过程中，外部荷载变化（如水位波动、气温变化等）与内部物理化学侵蚀（如碳化、硫酸盐侵蚀等）持续对坝体造成负面影响，威胁大坝的安全运行[5]。（剩余10445字）