[1] WANG Yawei, ZHU Qingxin, WANG Hao, FU Wenwei, et al. Semi-supervised diagnostic method for abnormal vibration data in SHM via SCNs and Bayesian-optimized LSTM [J]. Journal of Southeast University (English Edition), 2026, 42 (3): 298-305. [doi:10.3969/j.issn.1003-7985.2026.03.003]

Copy

Semi-supervised diagnostic method for abnormal vibration data in SHM via SCNs and Bayesian-optimized LSTM()

基于SCN与贝叶斯优化LSTM的结构健康监测异常振动数据半监督诊断方法

Journal of Southeast University (English Edition)[ISSN:1003-7985/CN:32-1325/N]

Volumn:

42

Issue:

2026 3

Page:

298-305

Research Field:

Civil Engineering

Publishing date:

2026-09-06

Info

Title:

Semi-supervised diagnostic method for abnormal vibration data in SHM via SCNs and Bayesian-optimized LSTM

基于SCN与贝叶斯优化LSTM的结构健康监测异常振动数据半监督诊断方法

Author(s):

WANG Yawei¹, ZHU Qingxin^1,2, WANG Hao¹, FU Wenwei³

1.Key Laboratory of Concrete and Prestressed Concrete Structures of Ministry of Education, Southeast University, Nanjing 211189, China
2.School of Environment and Architecture, University of Shanghai for Science and Technology, Shanghai 200093, China
3.School of Civil Engineering, Suzhou University of Science and Technology, Suzhou 215011, China

王亚伟¹, 祝青鑫^1,2, 王浩¹, 傅文炜³

1.东南大学混凝土及预应力混凝土结构教育部重点实验室, 南京 211189
2.上海理工大学环境与建筑学院, 上海 200093
3.苏州科技大学土木工程学院, 苏州 215011

Keywords:

semi-supervised diagnostic method;  scattering convolution network;  long short-term memory;  density-based spatial clustering;  structural health monitoring

半监督诊断方法;  散射卷积网络;  长短期记忆网络;  基于密度的空间聚类方法;  结构健康监测

PACS:

TU356

DOI:

10.3969/j.issn.1003-7985.2026.03.003

Abstract:

Structural health monitoring (SHM) systems frequently operate under harsh environmental conditions, inevitably leading to data anomalies such as outliers, deviations, and drifts. This study develops a semi-supervised learning framework leveraging long short-term memory (LSTM) networks to systematically diagnose and categorize abnormal acceleration signals. The proposed method integrates an enhanced density-based spatial clustering of applications with noise (DBSCAN) algorithm to automatically assign labels to acceleration data, substantially reducing the need for labor-intensive manual annotation of raw datasets. A scattering convolution network (SCN) is employed to decompose raw acceleration signals. The first three orders of wavelet scattering coefficients are extracted as the training input features for the subsequent model. This approach achieves a 97.01% reduction in feature dimensionality compared with the original signal, thereby optimizing the computational efficiency and diagnostic accuracy of the LSTM model. The effectiveness and applicability of the methodology are rigorously validated through a case study involving the Suzhou Science and Technology Museum (SSTM). Experimental results confirm that when the minimum points threshold (M) and scanning radius(ε) of the enhanced DBSCAN algorithm are configured at 5 and 0.1, respectively, the unsupervised labeling module attains high data annotation accuracy. Furthermore, Bayesian optimization is utilized to fine-tune the hyperparameters of the LSTM model, yielding a missed detection rate of 0%, a false detection rate of 1.4%, and an overall diagnostic accuracy of 96%. These findings demonstrate the proposed method significantly enhances computational efficiency while preserving diagnostic accuracy, providing an effective reference for anomaly diagnosis in measurement data.

结构健康监测(SHM)系统通常在恶劣环境条件中运行,不可避免地会出现离群值、偏差和非线性漂移等数据异常。因此,本文提出了基于长短期记忆网络(LSTM)的异常数据半监督诊断方法。该方法利用改进的基于密度的空间聚类方法(DBSCAN),自动为加速度数据分配标签,解决了传统监督学习中标注成本高的问题。在此基础上,采用散射卷积网络(SCN)对原始加速度信号进行多尺度分解,利用前3阶小波散射系数作为训练数据,大幅降低了冗余信息(数据压缩率达97.01%),提升了LSTM的计算效率与分类精度。最后,通过苏州科技馆(SSTM)的振动监测数据验证了该方法的有效性。结果表明,当改进的DBSCAN算法的最小包含点数M和扫描半径ε分别设置为5和0.1时,可有效实现振动数据的高精度标注。此外,利用贝叶斯优化对LSTM模型的超参数进行调优后,漏检率降低至0%,误检率降低至1.4%,诊断准确率提升至96%。该方法在保证诊断精度的同时显著提升了计算效率,为结构健康监测数据异常诊断提供了有效参考。

References:

[1]YANG D H, SUN J Z, YI T H, et al. Early warning technology of long-span bridge bearing deterioration considering time lag effects of thermal-induced displacement[J]. Journal of Southeast University (Natural Science Edition), 2024, 54(2): 268-274. (in Chinese)
[2]ZHU Q X, WANG H, CHEN Z Y, et al. Statistic model of vehicle loads for arterial highway bridges in Jiangsu Province based on WlM system[J]. Journal of Southeast University (Natural Science Edition), 2023, 53(6): 1148-1155. (in Chinese)
[3]ZHU Q X, WANG H, ZHU X J, et al. Investigation on the pattern for train-induced strains of the long-span steel truss railway bridge[J]. Engineering Structures, 2023, 275: 115268.
[4]ZHU Q X, WANG H, SPENCER B FJr, et al. Mapping of temperature-induced response increments for monitoring long-span steel truss arch bridges based on machine learning[J]. Journal of Structural Engineering, 2022, 148(5): 04022034.
[5]GOLDSTEIN M, UCHIDA S. A comparative evaluation of unsupervised anomaly detection algorithms for multivariate data[J]. PLoS One, 2016, 11(4): e0152173.
[6]WENG S, CHEN C J, ZHANG J S, et al. Construction simulation and monitoring of Zhengzhou Zhongyuan international exhibition center assisted by flexible sensors[J/OL]. Journal of Southeast University (Natural Science Edition), 2026[2026-02-04]. https: //link. cnki. net/urlid/32. 1178. N. 20260203. 1504. 007. (in Chinese)
[7]WORDEN K, MANSON G, FIELLER N R J. Damage detection using outlier analysis[J]. Journal of Sound and Vibration, 2000, 229(3): 647-667.
[8]MALLAT S, HWANG W L. Singularity detection and processing with wavelets[J]. IEEE Transactions on Information Theory, 1992, 38(2): 617-643.
[9]KULLAA J. Detection, identification, and quantification of sensor fault in a sensor network[J]. Mechanical Systems and Signal Processing, 2013, 40(1): 208-221.
[10]HUANG H B, YI T H, LI H N. Bayesian combination of weighted principal-component analysis for diagnosing sensor faults in structural monitoring systems[J]. Journal of Engineering Mechanics, 2017, 143(9): 04017088.
[11]ZHOU G D, YANG J Y, CHEN J M, et al. Bridge temperature-thermal displacement CNN LSTM model efficient updating method with incremental monitoring data[J/OL]. Journal of Southeast University (Natural Science Edition), 2026[2026-02-09]. https: //link. cnki. net/urlid/32. 1178. N. 20260206. 1755. 002. (in Chinese)
[12]WANG L X, LIU H B, CHEN Z H, et al. Combined digital twin and hierarchical deep learning approach for intelligent damage identification in cable dome structure[J]. Engineering Structures, 2023, 274: 115172.
[13]ZHANG Y X, WANG X Y, WU W L, et al. Anomaly detection of sensor faults and extreme events by anomaly locating strategies and convolutional autoencoders[J]. Structural Health Monitoring, 2025: 14759217251333059.
[14]FU Y G, PENG C, GOMEZ F, et al. Sensor fault management techniques for wireless smart sensor networks in structural health monitoring[J]. Structural Control and Health Monitoring, 2019, 26(7): e2362.
[15]BAO Y Q, TANG Z Y, LI H, et al. Computer vision and deep learning-based data anomaly detection method for structural health monitoring[J]. Structural Health Monitoring, 2019, 18(2): 401-421.
[16]NI F T, ZHANG J, NOORI M N. Deep learning for data anomaly detection and data compression of a long-span suspension bridge[J]. Computer-Aided Civil and Infrastructure Engineering, 2020, 35(7): 685-700.
[17]ZHU Q K, LI W Q, WANG X Y, et al. Anomaly detection in bridge structural health monitoring via 1D-LBP and statistical feature fusion[J]. Structures, 2024, 70: 107734.

Memo

Memo:

Received: 2026-01-02; Revised: 2026-04-10.
Biographies: WANG Yawei (1986—), male, Ph.D.candidate; WANG Hao (corresponding author), male, doctor, professor, wanghao1980@seu.edu.cn.
Foundation items: The National Natural Science Foundation of China (No.52408342, 52378291), Shanghai Pujiang Program (No.23PJ1410000).
Citation: WANG Yawei, ZHU Qingxin, WANG Hao, et al. Semi-supervised diagnostic method for abnormal vibration data in SHM via SCNs and Bayesian-optimized LSTM[J]. Journal of Southeast University (English Edition), 2026, 42(3): 298-305. DOI: 10. 3969/j. issn. 1003-7985. 2026. 03. 003.

Common functions