|Table of Contents|

Structural acceleration data anomaly detection based on 2D convolutional neural network(PDF)

《建筑科学与工程学报》[ISSN:1673-2049/CN:61-1442/TU]

Issue:
2025年01期
Page:
112-120
Research Field:
建筑结构
Publishing date:

Info

Title:
Structural acceleration data anomaly detection based on 2D convolutional neural network
Author(s):
MA Shenglan1;  ZHONG Jiankun1;  LIU Yuhao1;  ZHENG Xiang2
(1. Fujian Provincial Key Laboratory of Advanced Technology and Informatization in Civil Engineering, Fujian University of Technology, Fuzhou 350118, Fujian, China; 2. Xiamen Yijian Group, Xiamen 361004, Fujian, China)
Keywords:
structural health monitoring;  2D convolutional neural network;  truss structure;  deep learning;  acceleration;  data anomaly detection
PACS:
TU317
DOI:
10.19815/j.jace.2023.06082
Abstract:
In order to improve the efficiency and accuracy of anomaly detection in structural acceleration data, a structural acceleration data anomaly detection method based on 2D convolutional neural network(2D-CNN)was proposed. The effectiveness of the proposed method was verified through a two-dimensional truss numerical model, and the effects of 2D-CNN convolution layers and acceleration noise levels on data anomaly detection were studied. The results show that the proposed structural acceleration data anomaly detection method can quickly and accurately distinguish the types of acceleration data anomalies, and the accuracy of anomaly detection can reach over 97%. For samples with complex information and large data scales, using 2D-CNN with more than 4 layers can help improve the accuracy of acceleration data anomaly detection. Using 2D-CNN with 5 convolutional layers can achieve a data anomaly identification accuracy up to 98%. When the acceleration signal-to-noise ratio is greater than 1, the accuracy of data anomaly detection is above 90%. When the acceleration signal-to-noise ratio is 10, the accuracy is above 97%. The proposed method has good noise tolerance and robustness. The use of 2D-CNN data anomaly detection method can provide technical support for the effective operation of sensor networks.

References:

[1] 鞠翰文,邓 扬,李爱群.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 Hanwen, DENG Yang, LI Aiqun. Restoring method of structural abnormal monitoring data based on GRU neural network[J]. Journal of Vibration and Shock, 2023, 42(9): 328-338.
[2]YI T H, LI H N, SONG G B, et al. Detection of shifts in GPS measurements for a long-span bridge using CUSUM chart[J]. International Journal of Structural Stability and Dynamics, 2016, 16(4): 1640024.
[3]RAO A R M, KASIREDDY V, GOPALAKRISHNAN N, et al. Sensor fault detection in structural health monitoring using null subspace-based approach[J]. Journal of Intelligent Material Systems and Structures, 2015, 26(2): 172-185.
[4]MA S L, JIANG S F, LI J. Structural damage detection considering sensor performance degradation and measurement noise effect[J]. Measurement, 2019, 131: 431-442.
[5]LI Z L, KOH B H, NAGARAJAIAH S. Detecting sensor failure via decoupled error function and inverse input-output model[J]. Journal of Engineering Mechanics, 2007, 133(11): 1222-1228.
[6]范时枭,张金辉,张其林.结构健康监测系统的数据异常识别[J].计算机辅助工程,2016,25(5):60-65.
FAN Shixiao, ZHANG Jinhui, ZHANG Qilin. Abnormal data recognition in structural health monitoring system[J]. Computer Aided Engineering, 2016, 25(5): 60-65.
[7]游 颖,王 建,刘学刚,等.改进BP神经网络的钢结构应力缺失数据重构[J].建筑科学与工程学报,2022,39(4):166-173.
YOU Ying, WANG Jian, LIU Xuegang, et al. Reconstruction of missing stress data for steel structure based on improved BP neural network[J]. Journal of Architecture and Civil Engineering, 2022, 39(4): 166-173.
[8]LECUN Y, BENGIO Y,HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436-444.
[9]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.
[10]潘成龙,应雨龙.基于二维卷积神经网络的滚动轴承变工况故障诊断方法[J].上海电力大学学报,2022,38(1):29-34.
PAN Chenglong, YING Yulong. A fault diagnosis method for rolling bearings under variable condition based on two-dimensional convolutional neural network[J]. Journal of Shanghai University of Electric Power, 2022, 38(1): 29-34.
[11]顾梅花,苏彬彬,王苗苗,等.彩色图像灰度化算法综述[J].计算机应用研究,2019,36(5):1286-1292.
GU Meihua, SU Binbin, WANG Miaomiao, et al. Survey on decolorization methods[J]. Application Research of Computers, 2019, 36(5): 1286-1292.
[12]YI T H, HUANG H B, LI H N. Development of sensor validation methodologies for structural health monitoring: a comprehensive review[J]. Measurement, 2017, 109: 200-214.
[13]梁振隆.基于递归图和卷积神经网络的结构损伤识别研究[D].合肥:合肥工业大学,2020.
LIANG Zhenlong. Structural damage identification based on recurrence plot and convolutional neural network[D]. Hefei: Hefei University of Technology, 2020.
[14]CHOPRA S, HADSELL R, LECUN Y. Learning a similarity metric discriminatively,with application to face verification[C]//IEEE. 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition(CVPR'05). San Diego: IEEE, 2005: 539-546.
[15]李书进,赵 源,孔 凡,等.卷积神经网络在结构损伤诊断中的应用[J].建筑科学与工程学报,2020,37(6):29-37.
LI Shujin, ZHAO Yuan, KONG Fan, et al. Application of convolutional neural network in structural damage identification[J]. Journal of Architecture and Civil Engineering, 2020, 37(6): 29-37.
[16]麻胜兰,翁柳青,姜绍飞.传感器性能退化与结构损伤响应异常区分研究[J].西南交通大学学报,2013,48(6):1024-1030.
MA Shenglan, WENG Liuqing, JIANG Shaofei. Distinguishing between sensor deterioration and structural damage[J]. Journal of Southwest Jiaotong University, 2013, 48(6): 1024-1030.
[17]KULLAA J. Detection,identification,and quantification of sensor fault in a sensor network[J]. Mechanical Systems and Signal Processing, 2013, 40(1): 208-221.
[18]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.

Memo

Memo:
-
Last Update: 2025-01-20