基于多层信号分解的混凝土拱坝变形监测模型  

Deformation Monitoring Model of Concrete Arch Dam Based on Multi-Layer Signal Decomposition

作　　者：王子轩 欧斌 陈德辉 杨石勇 赵定柱 傅蜀燕 WANG Zixuan;OU Bin;CHEN Dehui;YANG Shiyong;ZHAO Dingzhu;FU Shuyan(College of Water Conservancy,Yunnan Agricultural University,Kunming,650201,China;Yunnan Small and Medium-lSized Water Conservancy Project Intelligent Management and Maintenance Engineering Research Center,Kunming 650201,China;National Key Laboratory of Water Disaster Prevention,Nanjing 210098,China)

机构地区：[1]云南农业大学水利学院,昆明650201 [2]云南省中小型水利工程智慧管养工程研究中心,昆明650201 [3]水灾害防御全国重点实验室,南京210098

出　　处：《三峡大学学报（自然科学版）》2024年第6期1-9,共9页Journal of China Three Gorges University：Natural Sciences

基　　金：国家自然科学基金项目(52069029,52369026);水灾害防御全国重点实验室2023年度“一带一路”水与可持续发展科技基金项目(2023490411);云南省农业基础研究联合专项面上项目(202401BD070001-071)。

摘　　要：为了充分挖掘大坝变形监测数据的非线性和非平稳性特征,本文提出了一种大坝变形监测模型.首先,该模型通过自适应噪声完全集合经验模态分解(CEEMDAN)对变形监测数据进行分解处理.在分解过程中融入样本熵(SE)和K-均值聚类,以确保得到的模态分量(IMF)个数能够准确描述大坝变形.然后,对于高频IMF分量,采用变分模态分解(VMD)进行二次分解,并利用偏最小二乘法(PLS)分析变形序列影响因子,以提取最佳的IMF分量作为后续模型的输入因子.最后,利用改进的共生生物搜索算法(ISOS)结合长短期记忆神经网络(LSTM)进行大坝变形的准确预测.研究结果表明:相较于单层信号处理,本文通过二次信号处理可以显著提升模型的预测精度;对二次分解后的IMFs分量进行PLS筛选可以有效避免模型的冗余性,提高计算效率;相较于各对比模型,本文模型在各测点上均具有较好的预测精度和稳定性.本文提出的模型能够深入挖掘大坝监测数据中的拓扑关系,有效保留数据中的高频有用信息,从而提高预测的准确性和平滑性,展示出较好的预测精度和泛化能力.In order to fully exploit the nonlinear and non-lstationary characteristics of dam deformation monitoring data,this paper proposes a dam deformation monitoring model.Firstly,the model decomposes the deformation monitoring data by adaptive noise complete ensemble empirical mode decomposition(CEEMDAN).Sample entropy(SE)and K-lmeans clustering are integrated into the decomposition process to ensure the obtained number of intrinsic mode function components(IMF)can accurately describe the dam deformation.Then,for the high-lfrequency IMF component,the variational mode decomposition(VMD)is used for secondary decomposition,and the partial least squares(PLS)is used to analyze the influence factors of the deformation sequence to extract the best IMF component as the input factor of the subsequent model.Finally,the improved symbiotic biological search algorithm(ISOS)combined with long short-lterm memory neural network(LSTM)is used to accurately predict the dam deformation.The results show that the prediction accuracy of the model can be significantly improved by secondary signal processing compared with single-llayer signal processing.PLS screening of IMF components after secondary decomposition can effectively avoid the redundancy of the model and improve the computational efficiency.Compared with each comparison model,the proposed model in this paper has better prediction accuracy and stability at each measuring point and can deeply mine the topological relationship in the dam monitoring data,effectively retain the high-lfrequency useful information in the data,thereby improving the accuracy and smoothness of the prediction,and showing better prediction accuracy and generalization ability.

关 键 词：大坝变形 自适应噪声完全集合经验模态分解 样本熵 K-均值聚类算法 改进的共生生物搜索算法 变分模态分解 

分 类 号：TV698.1[水利工程—水利水电工程]