基于GRU神经网络的结构异常监测数据修复方法

摘要
图/表
参考文献(27)
相关文章 (15)

全文: PDF (2284 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要结构健康监测系统中通常存在大量的异常监测数据，为保证数据的完整性和可用性，有必要对异常监测数据进行修复。大多数基于深度学习对异常数据进行修复的研究通常使用单输入维度和单向预测的方法搭建模型。本文提出一种基于门控循环(gated recurrent unit，GRU)神经网络的结构异常监测数据修复方法，该方法充分利用深度学习神经网络适合处理复杂非线性映射问题的优势，并对GRU神经网络进行了优化与重构。利用结构温度、时序先后相关性优化神经网络的输入和输出构造，并提出了利用异常数据前后时间段的信息进行双向序列预测的方法提升数据预测和修复精度。最后，利用某古城墙的应变、裂缝与温度监测数据进行方法验证，采用重构后的GRU神经网络模型对异常数据序列进行修复，并与长短时记忆(long and short-term memory ，LSTM)神经网络和反向传播(back propagation，BP)神经网络的修复精度进行比较。结果表明：相比单输入维度、单向预测的网络模型，重构后的GRU神经网络的预测精度大幅提高，且显著优于LSTM神经网络和BP神经网络。异常数据序列修复后，应变和裂缝宽度等结构响应与结构温度的线性相关性大幅增强。该方法对具有温度相关性的结构监测数据具有良好的修复能力。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	鞠翰文1
	邓扬1
	2
	李爱群1
	2

关键词 ：结构健康监测, 数据修复, 深度学习, 神经网络, 温度

Abstract：There are usually a large number of anomaly monitoring data in structural health monitoring systems. To ensure the integrity and practicability of data, it is necessary to restore anomaly monitoring data. Most studies on restoring anomaly data based on deep learning usually used single input dimension and unidirectional prediction to build models. This paper proposed a restoration method of structural anomaly monitoring data based on a Gated Recurrent Unit (GRU) neural network. The advantage of deep learning neural network to deal with the complex nonlinear mapping problem was fully utilized in this method by optimizing and reconstructing GRU neural network. The configurations of input and output of neural network were optimized by using the correlations of temperature and time series. Meanwhile a bidirectional sequence prediction method by using the information before and after anomaly data sequences was proposed to improve the prediction and restoration accuracy. At last, the proposed method was verified based on the strain, crack, and temperature monitoring data of an ancient city wall. The reconstructed GRU neural network model was used to restore the anomaly data sequences, and the restoration accuracy was compared with Long and Short-Term Memory (LSTM) neural network and Back Propagation (BP) neural network. The results show that compared with the neural network model of single input dimension and unidirectional prediction, the reconstructed GRU neural network has better prediction accuracy. And the prediction accuracy of the reconstructed GRU neural network is also significantly better than that of LSTM neural network and BP neural network. After anomaly data sequences are restored, the linear correlation of structural temperature and responses including strain and crack width gets greatly enhanced. The proposed method has a great ability to restore structural monitoring data with temperature correlation.

Key words： structural health monitoring data restoration deep learning neural network temperature

收稿日期: 2022-04-19 出版日期: 2023-05-15

引用本文:

鞠翰文1，邓扬1,2，李爱群1,2. 基于GRU神经网络的结构异常监测数据修复方法[J]. 振动与冲击, 2023, 42(9): 328-338.
JU Hanwen1, DENG Yang1,2, LI Aiqun1,2. Restoring method of structural abnormal monitoring data based on GRU neural network. JOURNAL OF VIBRATION AND SHOCK, 2023, 42(9): 328-338.

链接本文:

http://jvs.sjtu.edu.cn/CN/ 或 http://jvs.sjtu.edu.cn/CN/Y2023/V42/I9/328

[1] 荆根强，段发阶，彭璐，等. 基于被动激励的结构应变监测系统在线校准方法[J]. 仪器仪表学报，2021,41(08):137-145.
JING Gen-qiang, DUAN Fa-jie, PENG Lu, et al. On-line calibration method for structural strain monitoring system based on passive excitation [J]. Chinese Journal of Scientific Instrument, 2021, 41(08): 137-145.
[2] 王莉，吴桐. 基于光纤传感技术的混凝土结构深层裂缝监测研究[J]. 激光杂志，2020, 41(11): 193-197.
WANG Li, WU Tong. Study on the monitoring of deep crack of concrete structure based on fiber-optic sensing technology [J]. Laser Journal, 2020, 41(11): 193-197.
[3] 马帜，罗尧治，万华平，等. 基于概率主成分分析的结构健康监测数据修复方法研究[J]. 振动与冲击，2021, 40(21): 135-141+167.
MA Zhi, LUO Yao-zhi, WAN Hua-ping, et al. Repair method of structural health monitoring data based on probabilistic principal component analysis [J]. Journal of vibration and shock, 2021, 40(21): 135-141+167.
[4] 吴迪. 基于深度学习的结构健康监测系统异常数据识别方法研究[J]. 智能城市，2020, 6(20): 10-13.
WU Di. Research on abnormal data identification method of structural health monitoring system based on deep learning [J]. Intelligent City, 2020, 6(20): 10-13.
[5] 任普. 基于大数据的桥梁监测数据清洗方法研究[D]. 南京：东南大学，2019.
REN Pu. Research on the data cleansing methods for bridge monitoring data based on big-data platform [D]. Nanjing: Southeast University, 2019.
[6] 高照明. 结构健康监测数据异常诊断与修复方法研究[D]. 大连：大连理工大学，2020.
GAO Zhao-ming. Research on anomaly diagnosis and repair methods of structural health monitoring data [D]. Dalian: Dalian University of Technology, 2020.
[7] Eskelson BNI, Temesgen H, LeMay V, et al. The roles of nearest neighbor methods in imputing missing data in forest inventory and monitoring databases [J]. Scandinavian Journal of Forest Research, 2009, 24(3): 235-246.
[8] 陆化普，屈闻聪，孙智源. 基于S-G滤波的交通流故障数据识别与修复算法[J]. 土木工程学报，2015, 48(05): 123-128.
LU Hua-pu, QU Wen-cong, SUN Zhi-yuan. Detection and repair algorithm of traffic erroneous data based on S-G filtering [J]. China Civil Engineering Journal, 2015, 48(05): 123-128.
[9] Farreras-Alcover I, Chryssanthopoulos MK, Andersen JE. Regression models for structural health monitoring of welded bridge joints based on temperature, traffic and strain measurements [J]. Structural Health Monitoring, 2015, 14(6): 648-62.
[10] Wan HP, Ni YQ. Bayesian multi-task learning methodology for reconstruction of structural health monitoring data [J]. Structural Health Monitoring, 2018, 18(8): 147592171879495.
[11] Zhang Z, Luo Y. Restoring method for missing data of spatial structural stress monitoring based on correlation [J]. Mechanical Systems and Signal Processing, 2017, 91: 266-277.
[12] Li LC, Liu HL, Zhou HJ, et al. Missing data estimation method for time series data in structure health monitoring systems by probability principal component analysis [J]. Advances in Engineering Software, 2020, 149: 102901.
[13] 谢晓凯，罗尧治，张楠. 基于神经网络的大跨度空间钢结构应力实测缺失数据修复方法研究[J]. 空间结构，2019, 25(03): 38-44.
XIE Xiao-kai, LUO Yao-zhi, ZHANG Nan, et al. Missing data reconstruction in stress monitoring of steel spatial structures using neural network techniques [J]. Spatial Structures, 2019, 25(03): 38-44.
[14] 杜曼玲，高嘉欣，张礼兵，等. 大坝监测数据的时序预测与补全[J]. 水力发电，2020, 46(11): 111-115.
DU Man-ling, GAO Jia-xin, ZHANG Li-bing, et al. Time series forecasting and imputation of dam physical quantities [J]. Water Power, 2020, 46(11): 111-115.
[15] Li S, Li S, Laima S, et al. Data driven modeling of bridge buffeting in the time domain using long short-term memory network based on structural health monitoring [J]. Structural Control and Health Monitoring, 2021, 28: e2772.
[16] Jeong S, Ferguson M, Hou R, et al. Sensor data reconstruction using bidirectional recurrent neural network with application to bridge monitoring [J]. Advanced Engineering Informatics, 2019, 42(10): 100991.1-100991.14.
[17] 盖彤彤，曾森，于德湖，等. 索力振动法测量神经网络赋泛研究[J]. 工程科学与技术，2021, 53(04): 118-127.
GAI Tong-tong, ZENG Sen, YU De-hu, et al. Research on neural network generalization of cable force vibration measurement [J]. Advanced Engineering Sciences, 2021, 53(04): 118-127.
[18] Yue ZX, Ding YL, Zhao HW. Deep learning-based minute-scale digital prediction model of temperature-induced deflection of a cable-stayed bridge: case study [J]. Journal of Bridge Engineering, 2021, 26(6): 05021004.
[19] Yang JX, Yang F, Zhou YX, et al. A data-driven structural damage detection framework based on parallel convolutional neural network and bidirectional gated recurrent unit [J]. Inform Sciences. 2021, 566: 103-117.
[20] Chen JL, Jing HJ, Chang YH, et al. Gated recurrent unit based recurrent neural network for remaining useful life prediction of nonlinear deterioration process [J]. Reliability Engineering and System Safety, 2019, 185: 372-382.
[21] 姚德臣，李博阳，刘恒畅，等. 基于注意力GRU算法的滚动轴承剩余寿命预测[J]. 振动与冲击，2021,40(17): 116-123.
YAO De-chen, LI Bo-yang, LIU Heng-chang, et al. Residual life prediction of rolling bearing based on attention GRU algorithm [J]. Journal of Vibration and Shock, 2021, 40(17): 116-123.
[22] 孟恒宇，李元祥. 基于Transformer重建的时序数据异常检测与关系提取[J]. 计算机工程，2021, 47(02): 69-76.
MENG Heng-yu, LI Yuan-xiang. Anomaly detection and relation extraction for time series data based on transformer reconstruction [J]. Computer Engineering, 2021, 47(02): 69-76.
[23] Yue ZX, Ding YL, Zhao HW, et al. Mechanics-guided optimization of an LSTM network for real-time modeling of temperature-induced deflection of a cable-stayed bridge [J]. Engineering Structures, 2022. DOI:10.1016/j.engstruct.2021. 113619.
[24] Zhao HW, Ding YL, Li AQ, et al. Digital modeling on the nonlinear mapping between multi-source monitoring data of in-service bridges [J]. Structural Control and Health Monitoring, 2020, 27: e2618.
[25] 郑直，张华钦，潘月. 基于改进鲸鱼算法优化LSTM的滚动轴承故障诊断[J]. 振动与冲击，2021, 40(07): 274-280.
ZHENG Zhi, ZHANG Hua-qin, PAN Yue. Rolling bearing fault diagnosis based on IWOA-LSTM [J]. Journal of Vibration and Shock, 2021, 40(07): 274-280.
[26] Ni YQ, Zhou HF, Ko JM. Generalization capability of neural network models for temperature-frequency correlation using monitoring data [J]. Journal of Structural Engineering, 2009, 135(10): 1290-1300.
[27] Chang YS, Chiao HT, Abimannan S, et al. An LSTM-based aggregated model for air pollution forecasting [J]. Atmospheric Pollution Research, 2020, 11: 1451-1463.