大坝安全监测数据降噪的流形学习方法  

Manifold learning method for denoising dam safety monitoring data

作　　者：冯燕明 何杨杨 左生龙 张帅 徐朗 苏怀智[2,3] FENG Yanming;HE Yangyang;ZUO Shenglong;ZHANG Shuai;XU Lang;SU Huaizhi(POWERCHINA Kunming Engineering Corporation Limited,Kunming 650032,China;The National Key Laboratory of Water Disaster Prevention,Hohai University,Nanjing 210098,China;College of Water Conservancy and Hydropower Engineering,Hohai University,Nanjing 210098,China;POWERCHINA International Group Limited,Beijing 100089,China)

机构地区：[1]中国电建集团昆明勘测设计研究院有限公司,云南昆明650032 [2]河海大学水灾害防御全国重点实验室,江苏南京210098 [3]河海大学水利水电学院,江苏南京210098 [4]中国电建集团国际工程有限公司,北京100089

出　　处：《水利水电科技进展》2024年第4期59-64,共6页Advances in Science and Technology of Water Resources

基　　金：国家自然科学基金重点项目(52239009);云南省水利水电工程安全重点实验室开放课题基金项目(202302AN360003);中国电力建设股份有限公司重点科技项目(DJ-ZDXM-2021-11)。

摘　　要：针对大坝变形、渗流、应力应变等安全监测数据难以避免受到噪声污染,且传统线性降噪方法去噪效果不佳的问题,提出了基于相空间重构与流形学习相组合的大坝安全监测数据非线性降噪方法。该方法在重构大坝安全监测数据时间序列相空间的基础上,通过交叉应用局部切空间排列方法与极大似然估计、自适应邻域等方法,以重构的相空间为桥梁,提取大坝安全监测数据序列深层次信息,得到降噪后的大坝安全监测数据。工程实测数据验证结果表明,相比小波软阈值法和固定邻域-LTSA法,本文提出的方法降噪效果更优,具有一定的工程应用价值。The noise pollution cannot be avoided for dam deformation,seepage,stress-strain,and other safety monitoring data.It is difficult to achieve excellent denoising effect for the traditional linear noise reduction method.On the basis of reconstructing the phase space of dam safety monitoring data time sequences,with the cross application of local tangent space alignment method,maximum likelihood estimate,adaptive neighborhood and other methods,using the reconstructed phase space as a bridge,and through the extraction of deep information of the monitoring data sequences,the denoised data sequences of dam safety monitoring is obtained.Based on the application of prototype test data,it is evident that the noise reduction effect of the method proposed in this paper is superior to that of the wavelet soft-thresholding method and the fixed neighborhood local tangent space alignment(LTSA)method,demonstrating certain value for engineering applications.

关 键 词：大坝安全 监测数据 降噪处理 流形学习 相空间重构 

分 类 号：TV698.1[水利工程—水利水电工程]