«上一篇
 文章快速检索 高级检索

 应用科技  2020, Vol. 47 Issue (3): 63-68  DOI: 10.11991/yykj.202004001 0

### 引用本文

XU Changzhe, YU Qinglin, YANG Qingsong. Structural damage identification technology based on BP neural network[J]. Applied Science and Technology, 2020, 47(3): 63-68. DOI: 10.11991/yykj.202004001.

### 文章历史

Structural damage identification technology based on BP neural network
XU Changzhe, YU Qinglin, YANG Qingsong
Nuclear Power Institute of China, Chengdu 610213, China
Abstract: Based on the high requirements of structural safety, the structural health monitoring system based on various monitoring technologies has been widely studied and applied. The structural damage identification system is one of the core components of the structural health monitoring system. In this paper, taking a cantilever beam as the engineering background, the structural damage identification technology is studied based on the BP neural network combined with information fusion. The BP neural network is constructed by MATLAB software, the accuracy of neural network damage identification after training is higher than 90%. The reliability of the structural damage identification technology based on neural network is discussed. The advantages and disadvantages of the damage identification technology combined with information fusion and neural network are summarized. The result of network identification proves feasibility of the technology, providing a reference for the application and further study of damage identification of engineering structures.
Keywords: health monitoring    damage identification    neural network    information fusion    network construction    damage sensitive features    data enhancement    modal analysis

1 损伤识别技术

1.1 BP神经网络的构建

1)搭建神经网络结构。

2)样本数据处理。

3)网络训练。

4)网络测试。

1.2 结构损伤敏感特征的选取

1.3 信息融合技术

2 特征数据提取与处理 2.1 损伤敏感特征提取

2.2 数据增强

3 损伤识别算例 3.1 基于单特征的结构损伤识别 3.1.1 分别基于3种特征参数的结构损伤识别

3.1.2 结果分析

3类特征数据中，固有频率对应的输出结果对损伤的识别效果最佳。神经网络的实际网络输出与期望值之间差值的绝对值如图6所示。

3.2 基于信息融合的结构损伤识别 3.2.1 相同损伤尺寸特征信息的信息融合

3.2.2 不同损伤尺寸特征信息的信息融合

3种不同训练情况下，网络输出结果如表8所示。

3.2.3 结果分析

4 结论

1)针对结构损伤识别进行研究。讨论分别用频率、位移模态以及MAC值作为网络输入时，BP神经网络识别损伤的能力。对输出结果分析得到基于上述3种特征，BP神经网络能对结构损伤进行准确的识别。在损伤尺寸发生变化时，网络依然能完成准确的损伤识别。3种特征中，位移模态作为特征输入时，网络识别精度最高，20组数据均能准确区分，应用频率与MAC值时准确率均达到90%。

2)讨论信息融合技术在结构损伤识别中的应用。相比仅用单类型特征作为网络输入参数，在融合3种特征的情况下，网络识别损伤的准确性更高，不存在因为结构某几阶特征对损伤不敏感造成的误差所引起的无法区分损伤状况的数据。网络的输出结果相比单特征损伤识别精确度高，更接近期望输出。

 [1] CAWLEY P, ADAMS R D. The location of defects in structures from measurements of natural frequencies[J]. The journal of strain analysis for engineering design, 1979, 14(2): 49-57. DOI:10.1243/03093247V142049 (0) [2] ADAMS R D, CAWLEY P, PYE C J, et al. A vibration technique for non-destructively assessing the integrity of structure[J]. Journal of mechanical engineering science, 1978, 20(2): 93-100. DOI:10.1243/JMES_JOUR_1978_020_016_02 (0) [3] DOEBLING S W, FARRAR C R, PRIME M B, et al. Damage identification and health monitoring of structural and mechanical systems from changes in their vibration characteristics: a literature review[M]. Los Alamos: Los Alamos National Lab, 1996. (0) [4] MASRI S F, NAKAMURA M, CHASSIAKOS A G, et al. Neural network approach to detection of changes in structural parameters[J]. Journal of engineering mechanics, 1996, 122(4): 350-360. DOI:10.1061/(ASCE)0733-9399(1996)122:4(350) (0) [5] WU X, GHABOUSSI J, GARRETT J H JR. Use of neural networks in detection of structural damage[J]. Computers & structures, 1992, 42(4): 649-659. (0) [6] KIRKEGAARD P H, RYTTER A. Use of neural networks for damage detection and location in a steel member[C]//Proceedings of 3rd International Conference on the Application of Artificial Intelligence to Civil and Structural Engineering. Edinburgh, Britain, 1993. (0) [7] WORDEN K, MANSON G, ALLMAN D. Experimental validation of a structural health monitoring methodology Part I: novelty detection on an aircraft wing[J]. Journal of sound and vibration, 2003, 259(2): 323-343. DOI:10.1006/jsvi.2002.5168 (0) [8] MANSON G, WORDEN K, ALLMAN D. Experimental validation of a structural health monitoring methodology: Part Ⅲ. Damage location on an aircraft wing[J]. Journal of sound and vibration, 2003, 259(2): 365-385. DOI:10.1006/jsvi.2002.5169 (0) [9] MANSON G, PAPATHEOU E, WORDEN K. Genetic optimisation of a neural network damage diagnostic[J]. The aeronautical journal, 2008, 112(1131): 267-274. DOI:10.1017/S0001924000002219 (0) [10] SHABBIR F, KHAN M I, AHMAD N, et al. Structural damage detection with different objective functions in noisy conditions using an evolutionary algorithm[J]. Applied sciences, 2017, 7(12): 1245. DOI:10.3390/app7121245 (0) [11] KUOK S C, YUEN K V. Investigation of modal identification and modal identifiability of a cable-stayed bridge with Bayesian framework[J]. Smart structures and systems, 2016, 17(3): 445-470. DOI:10.12989/sss.2016.17.3.445 (0) [12] XIA Yong, HAO Hong. Statistical damage identification of structures with frequency changes[J]. Journal of sound and vibration, 2003, 263(4): 853-870. DOI:10.1016/S0022-460X(02)01077-5 (0) [13] YANG Zhibo, CHEN Xuefeng, YU Jie, et al. A damage identification approach for plate structures based on frequency measurements[J]. Nondestructive testing and evaluation, 2013, 28(4): 321-341. DOI:10.1080/10589759.2013.801472 (0) [14] JAHANGIRI M, NAJAFGHOLIPOUR M A, DEHGHAN S M, et al. The efficiency of a novel identification method for structural damage assessment using the first vibration mode data[J]. Journal of sound and vibration, 2019, 458: 1-16. DOI:10.1016/j.jsv.2019.06.011 (0) [15] 李德葆, 陆秋海. 实验模态分析及其应用[M]. 北京: 科学出版社, 2001. (0) [16] 李德葆, 诸葛鸿程, 王波. 实验动态应变分析中的模态分析方法[J]. 固体力学学报, 1990, 11(4): 360-364. (0) [17] WEST W M. Illustration of the use of modal assurance criterion to detect structural changes in an orbiter test specimen[C]//Proceedings of the 4th International Modal Analysis Conference. Schenectady, USA, 1986: 1-6. (0) [18] KO J M, WONG C W, LAM H F. Damage detection in steel framed structures by vibration measurement approach[C]//Proceedings of the 12th International Modal Analysis Conference. Honolulu, USA, 1994: 280-286. (0) [19] 郭耀华. 基于深度学习的车辆驾驶状态识别算法研究[D]. 北京: 北京邮电大学, 2019. (0)