|本期目录/Table of Contents|

[1]马宏伟,林逸洲,聂振华. 利用少量传感器信息与人工智能的桥梁结构安全监测新方法[J].建筑科学与工程学报,2018,35(05):9-23.
 MA Hong-wei,LIN Yi-zhou,NIE Zhen-hua.New Methods of Structural Health Monitoring Based on Small Amount of Sensor Information and Artificial Intelligence[J].Journal of Architecture and Civil Engineering,2018,35(05):9-23.
点击复制

 利用少量传感器信息与人工智能的桥梁结构安全监测新方法(PDF)
分享到:

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

卷:
35卷
期数:
2018年05期
页码:
9-23
栏目:
出版日期:
2018-09-03

文章信息/Info

Title:
New Methods of Structural Health Monitoring Based on Small Amount of Sensor Information and Artificial Intelligence
作者:
马宏伟林逸洲聂振华
东莞理工学院[生态环境与建筑工程学院
Author(s):
MA Hong-wei LIN Yi-zhou NIE Zhen-hua
School of Environment and Civil Engineering, Dongguan University of Technology
关键词:
结构安全监测传感器人工智能深度学习
Keywords:
structural health monitoring sensor artificial intelligence deep learning
分类号:
-
DOI:
-
文献标志码:
A
摘要:
分析了当前结构安全监测在工程应用中存在的诸多问题和挑战,提出应发展利用少量传感器信息及基于大数据与人工智能的安全监测新方法,来克服现有系统传感器繁多、造价昂贵、海量数据难以处理的问题。介绍了单测点信息的多维相空间方法和单传感器信息的重构相空间方法;在基于双传感器信息的移动互相关函数法基础上,提出了基于双传感器信息的移动传递熵方法;阐述了基于少量传感器信息的移动主成分分析法的物理意义及其工程应用的适用性和可行性。利用单传感器方法和移动主成分分析法,以及小波包能量法、二次协方差矩阵法对虎门大桥进行长达5年的安全监测。阐述了基于深度学习的结构安全监测人工智能方法及其发展概况,分析了基于结构动力学信息的深度学习方法及其巨大的应用潜力。在此基础上进一步思考了基于少量传感器和人工智能相结合的方法在结构安全监测应用中的发展思路。结果表明:单传感器信息的重构相空间方法更适用于实际工程;基于双传感器信息的移动互相关函数法和移动传递熵方法均能精确定位损伤;移动主成分分析法最适用于实际工程的实时监测。
Abstract:
The current problems and challenges of structural health monitoring (SHM) in engineering application were analyzed. Due to the high consumption of sensors, high cost, and difficulties in handling massive data, new methods using small amount of sensor information and artificial intelligence (AI) should be developed. The method using multiple phase spaces of multi-type responses of single measurement point information, and the method using reconstructed phase space of single sensor information were introduced. Based on the method using moving cross-correlation of two sensors information, the method using moving transfer entropy of two sensors information was presented. The method using small amount of sensors based on moving principal component analysis (MPCA) was also introduced. These methods and wavelet packet energy method and covariance matrix of covariance were applied on Humen Bridges for five years. The AI method based on deep learning and its development and potential in SHM were expounded. The development ideas combining small amount of sensors method with AI in the future application of SHM were considered. The results indicate that the method using reconstructed phase space of single sensor information is more suitable to practical engineering. The method using moving crosscorrelation of two sensors information and moving transfer entropy of two sensors information methods both can locate the damage well. The MPCA is most suitable to the realtime monitoring.

参考文献/References:

相似文献/References:

备注/Memo

备注/Memo:
更新日期/Last Update: 2018-09-03