基于LSTM模型的桥梁振动信号降噪研究
Research on Bridge Vibration Signal Denoising Based on LSTM Model
公路桥梁的运营条件较为复杂,实际桥梁健康监测系统的监测数据中往往包含大量噪声,严重干扰桥梁结构状态评估结果的准确性。本文提出一种基于长短期记忆(LSTM)模型的降噪方法,用于滤除实测桥梁振动加速度信号中的噪声分量。首先,对桥梁振动加速度数据的构成进行数值模拟,并在其基础上建立LSTM模型的训练集。其次,根据实际工程应用需求设计LSTM降噪模型架构,将含噪信号输入LSTM模型后得到噪声分量,通过训练使LSTM模型能够有效去除噪声分量、提高信号质量。最后,以某大跨度双塔斜拉桥和某预应力混凝土连续梁桥为例对所提方法进行验证,使用互补集合经验模态分解(CEEMD)对原始信号、小波变换(WT)降噪信号、LSTM降噪信号分别开展评估,基于相关系数及正交指数分析了降噪效果,并分别在时域和频域评估了特征分量识别的准确性。结果表明,使用LSTM模型降噪后的加速度数据在时域和频域均能更加精确地识别特征信号分量,所提方法能够实现桥梁健康监测系统加速度数据的自适应降噪,进一步提升了桥梁结构状态评估的准确性。
The operational environment of highway bridges is complex, and monitoring data from structural health monitoring systems are frequently contaminated with substantial noise. This noise significantly compromises the accuracy of bridge structural condition assessments. To address this issue, a denoising method based on the Long Short-Term Memory (LSTM) network is proposed to filter noise components from measured bridge vibration acceleration signals. First, the composition of bridge vibration acceleration data is numerically simulated, and a training dataset for the LSTM model is constructed based on these simulations. Second, an LSTM-based denoising model architecture is designed to meet practical engineering requirements. Noisy signals are input into the LSTM model to isolate noise components. Through training, the model effectively removes noise and enhances signal quality. Finally, a long-span double-tower cable-stayed bridge and a prestressed concrete continuous girder bridge are used as case studies to validate the proposed method. Complementary Ensemble Empirical Mode Decomposition (CEEMD) is employed to evaluate the original signal, Wavelet Transform (WT)-denoised signal, and LSTM-denoised signal. Denoising performance is assessed using correlation coefficients and orthogonality indices, while the accuracy of characteristic component identification is evaluated in both the time and frequency domains. Results demonstrate that acceleration data denoised by the LSTM model enable more accurate identification of characteristic signal components in both domains. The proposed method achieves adaptive denoising of acceleration data from bridge health monitoring systems, thereby improving the accuracy of bridge structural condition assessments.
| [1] |
倪艳春,蔡永谊,张启伟.大跨径斜拉桥长期健康监测数据的性能分析与应用[J].结构工程师,2024,40(5):19-27. |
| [2] |
NI Yanchun,CAI Yongyi,ZHANG Qiwei.Performance analysis and application of long-term health monitoring data for a large-span cable-stayed bridge[J].Structural Engineers,2024,40(5):19-27.(in Chinese) |
| [3] |
龙武剑,舒雨清,梅柳,土木工程智能结构健康监测应用综述[J].结构工程师,2024,40(3):203-216. |
| [4] |
LONG Wujian,SHU Yuqing,MEI Liu,et al.An overview of application of intelligent structural health monitoring in civil engineering[J].Structural Engineers,2024,40(3):203-216.(in Chinese) |
| [5] |
WANG C G,CUI Y C,LIU Y,et al.High-G MEMS accelerometer calibration denoising method based on EMD and time-frequency peak filtering[J].Micromachines,2023,14(5):970-987. |
| [6] |
GHOSH A,MITTAL S,CHAKRABORTY A,et al.Multi-channel synchrosqueezed wavelet transform based identification of modal parameters of a vibrating structure[J].Journal of Civil Structural Health Monitoring,2025,15(6):1-22. |
| [7] |
熊璐伟,李庶林,杨明辉,改进的VMD-WT微震信号联合去噪方法[J].防灾减灾工程学报,2025,45(1):188-197. |
| [8] |
XIONG Luwei,LI Shulin,YANG Minghui,et al.Improved VMD-WT microseismic signal joint denoising method[J].Journal of Disaster Prevention and Mitigation Engineering,2025,45(1):188-197.(in Chinese) |
| [9] |
REZA D,MANDALA S,M.ZAKI S,et al.Deep learning autoencoder study on ECG signals[J].Jurnal Nasional Teknik Elektro,2023,12(3):82-88. |
| [10] |
XU B C,AN Z G,HAN Y,et al.Denoising CSAMT signals in the time domain base on long short-term memory[J].Journal of Geophysics and Engineering,2024,21(2):507-516. |
| [11] |
迟玉伦,李希铭,朱文博,基于LSTM-GRU神经网络的机床主轴回转误差分离降噪研究[J].计量学报,2024,45(11):1615-1625. |
| [12] |
CHI Yulun,LI Ximing,ZHU Wenbo,et al.Research on the separation and noise reduction method of spindle radial rotation error based on LSTM-GRU neural network[J].Acta Metrologica Sinica,2024,45(11):1615-1625.(in Chinese) |
| [13] |
屈广,孙利民,辛公锋.基于LSTM-SSA-BDLM的桥梁结构变形性能动态预测与预警[J].同济大学学报(自然科学版),2025,53(1):26-34. |
| [14] |
QU Guang,SUN Limin,XIN Gongfeng.Dynamic prediction and early warning of bridge structural deformation performance based on LSTM-SSA-BDLM[J].Journal of Tongji University(Natural Science),2025,53(1):26-34.(in Chinese) |
| [15] |
张明书,姚琛,吴贤国,基于BO-Adam-Bi-LSTM的盾构下穿既有隧道变形预测及调控[J].现代隧道技术,2024,61(6):92-99,110. |
| [16] |
ZHANG Mingshu,YAO Chen,WU Xianguo,et al.Deformation prediction and control for shield tunnelling passing under existing tunnels based on BO-Adam-Bi-LSTM[J].Modern Tunnelling Technology,2024,61(6):92-99,110.(in Chinese) |
| [17] |
刘杰,丁雪,刘庆宽,基于CEEMD和统计参数的斜拉桥损伤识别方法研究[J].振动与冲击,2024,43(19):326-336. |
| [18] |
LIU Jie,DING Xue,LIU Qingkuan,et al.Damage identification method of cable-stayed bridges based on CEEMD and statistical parameters[J].Journal of Vibration and Shock,2024,43(19):326-336.(in Chinese) |
| [19] |
何玉洁,李新娥,贺俊.基于参数优化变分模态分解的信号降噪方法[J].现代电子技术,2025,48(2):70-76. |
| [20] |
HE Yujie,LI Xine,HE Jun.Method of signal denoising based on parameter-optimized VMD[J].Modern Electronics Technique,2025,48(2):70-76.(in Chinese) |
| [21] |
张震,周光伟,陈昌萍.基于希尔伯特-黄变换的桥梁非线性气动力识别[J].厦门大学学报(自然科学版),2025,64(2):336-345. |
| [22] |
ZHANG Zhen,ZHOU Guangwei,CHEN Changping.Nonlinear aerodynamic identification of bridges based on Hilbert Huang transform[J].Journal of Xiamen University (Natural Science),2025,64(2):336-345.(in Chinese) |
广西科技基地和人才专项(桂AD25069039)
/
| 〈 |
|
〉 |