基于导波和轻量化卷积神经网络的复合材料结构损伤识别方法

包文强; 马济通; 赵森; 杨正岩

doi:10.21656/1000-0887.450162

基于导波和轻量化卷积神经网络的复合材料结构损伤识别方法

doi: 10.21656/1000-0887.450162

1.
大连海事大学信息与科学技术学院，辽宁大连 116026
2.
江南大学纤维工程与装备技术学院，江苏无锡 214122

基金项目:

国家自然科学基金 12102075

国家自然科学基金 62101090

中国科协青年人才托举工程 2022QNRC001

中央高校基本科研业务费 3132024236

详细信息

作者简介:
包文强(1999—)，男，硕士(E-mail: baowenqiang2022@163.com)

通讯作者:
杨正岩(1993—)，女，教授，博士(通讯作者. E-mail: zyyang1993@jiangnan.edu.cn)

中图分类号: V11
计量
- 文章访问数: 45
- HTML全文浏览量: 9
- PDF下载量: 5
- 被引次数: 0
出版历程
- 收稿日期: 2024-05-31
- 修回日期: 2024-09-18
- 刊出日期: 2025-08-01

A Lightweight Convolutional Neural Network for Guided Wave Based Damage Detection in Composite Structures

1.
School of Information Science and Technology, Dalian Maritime University, Dalian, Liaoning 116026, P.R.China
2.
School of Fiber Engineering and Equipment Technology, Jiangnan University, Wuxi, Jiangsu 214122, P.R.China

摘要

摘要: 针对有限计算资源下复合材料的实时监测问题，该文提出了基于超声导波和轻量化卷积神经网络(one-dimension convolutional neural network-deformable convolution attention，CDCA)的损伤实时识别方法. 在该方法中，为了压缩多个路径的导波信号，首先提出了改进差分驱动的平均聚合(improved differential-driven piecewise aggregate approximation，IDPAA)算法，利用该方法可以显著减少计算量；其次，提出了轻量化可变形卷积注意力(deformable convolution attention，DCA)机制，让模型聚焦在与损伤相关的像素级特征，从而实现更高效、准确的结构损伤识别；最后，通过结合一维卷积神经网络(one-dimension convolutional neural network，1D-CNN)和DCA机制，构建了CDCA模型. 该模型不仅可以在有限资源环境下运行，还能实现含噪声工况下的损伤实时识别. 在真实数据集上验证了所提出方法的有效性. 试验结果表明，所提出的损伤识别方法有较高的损伤识别准确性，准确率可达98%，并且大幅提高了模型计算效率，相较于其他先进深度学习模型具有显著的优势.
- 结构健康监测 /
- 超声导波 /
- 复合材料 /
- 深度学习 /
- 轻量化卷积神经网络
Abstract: For real-time monitoring composite structures under limited resource conditions, a machine learning-assisted method based on ultrasonic guided waves was proposed for real-time damage detection. In the proposed method, firstly, an improved differential-driven piecewise aggregate approximation (IDPAA) algorithm was designed to compress multi-path guided wave signals, thereby greatly reducing computational requirements. Secondly, a novel lightweight deformable convolution attention (DCA) mechanism was developed, which can help model focus on pixel-level feature information related to damage, enabling more efficient and accurate structural damage detection. Finally, by integrating a one-dimension CNN (1D CNN) with the designed DCA mechanism, a lightweight 1D-CNN-DCA (CDCA) model was proposed, which not only can work under the limited resource conditions but also can effectively real-time monitor structural damages under noise environments. The experiment of the proposed damage detection method on real-world datasets demonstrates its effectiveness, where the results reveal that it can reach 98% accuracy and significantly improves computation efficiency, outperforming other advanced monitoring approaches.
- structural health monitoring /
- ultrasonic guided wave /
- composite material /
- deep learning /
- lightweight convolutional neural network

HTML全文

图 1 IDPAA算法示意图

注为了解释图中的颜色，读者可以参考本文的电子网页版本，后同.

Figure 1. The schematic representation of the IDPAA algorithm

下载: 全尺寸图片幻灯片

图 2 CDCA模型结构

Figure 2. The architecture of the CDCA model

下载: 全尺寸图片幻灯片

图 3 实验设置

Figure 3. The experimental setup

下载: 全尺寸图片幻灯片

图 4 不同路径的完整信号、损伤信号对比图

Figure 4. Comparison of intact signals and damage signals for different paths

下载: 全尺寸图片幻灯片

图 5 压缩前后的导波信号

Figure 5. Guided wave signals before and after compression

下载: 全尺寸图片幻灯片

图 6 不同信噪比噪声条件下的混淆矩阵

Figure 6. Confusion matrixes under different SNR noise conditions

下载: 全尺寸图片幻灯片

图 7 模型性能对比

Figure 7. The model performance comparison diagram

下载: 全尺寸图片幻灯片

表 1 模型参数量，计算复杂度对比表

Table 1. Comparison of the model parameters and the computational complexities

model	number of parameters	FLOPs
CDCA	0.17M	7.2M
ResNet 18(1D-CNN)	3.9M	175M
CNN-LSTM	0.43M	139M
multi-scale 1D-CNN	8.6M	51M

下载: 导出CSV

参考文献(20)

[1]	ZHANG Z, LIU M, LI Q, et al. Visualized characterization of diversified defects in thick aerospace composites using ultrasonic B-scan[J]. Composites Communications, 2020, 22: 100435. doi: 10.1016/j.coco.2020.100435
[2]	YUN H, WANG R, RAYHANA R, et al. WaveCLR: contrastive learning of guided wave representations for composite damage identification[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 2515914.
[3]	BABU J, SUNNY T, PAUL N A, et al. Assessment of delamination in composite materials: a review[J]. Proceedings of the Institution of Mechanical Engineers (Part B): Journal of Engineering Manufacture, 2016, 230(11): 1990-2003. doi: 10.1177/0954405415619343
[4]	LI B, ZHANG L, YANG B. Grain refinement and localized amorphization of additively manufactured high-entropy alloy matrix composites reinforced by nano ceramic particles via selective-laser-melting/remelting[J]. Composites Communications, 2020, 19: 56-60. doi: 10.1016/j.coco.2020.03.001
[5]	YUN H, FENG K, RAYHANA R, et al. A multidimensional data fusion neural network for damage localization using ultrasonic guided wave[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 2522912.
[6]	姚浩, 夏桂然, 刘泽佳, 等. 基于机器学习的黏钢构件黏接层缺陷识别方法研究[J]. 应用数学和力学, 2024, 45(4): 429-442. doi: 10.21656/1000-0887.440365 YAO Hao, XIA Guiran, LIU Zejia, et al. A defect identification method for bonding layers of adhesive steel members based on machine learning[J]. Applied Mathematics and Mechanics, 2024, 45(4): 429-442. (in Chinese) doi: 10.21656/1000-0887.440365
[7]	LIM H J, SOHN H. Online stress monitoring technique based on lamb-wave measurements and a convolutional neural network under static and dynamic loadings[J]. Experimental Mechanics, 2020, 60(2): 171-179. doi: 10.1007/s11340-019-00546-8
[8]	LIAO Y, QING X, WANG Y, et al. Damage localization for composite structure using guided wave signals with Gramian angular field image coding and convolutional neural networks[J]. Composite Structures, 2023, 312: 116871. doi: 10.1016/j.compstruct.2023.116871
[9]	GORGIN R, LUO Y, WU Z. Environmental and operational conditions effects on Lamb wave based structural health monitoring systems: a review[J]. Ultrasonics, 2020, 105: 106114. doi: 10.1016/j.ultras.2020.106114
[10]	LIU K, MA S, WU Z, et al. A novel probability-based diagnostic imaging with weight compensation for damage localization using guided waves[J]. Structural Health Monitoring, 2016, 15(2): 162-173. doi: 10.1177/1475921715627491
[11]	ZHAO X, GAO H, ZHANG G, et al. Active health monitoring of an aircraft wing with embedded piezoelectric sensor/actuator network, Ⅰ: defect detection, localization and growth monitoring[J]. Smart Materials and Structures, 2007, 16(4): 1208-1217. doi: 10.1088/0964-1726/16/4/032
[12]	ANGRISANI L, BACCIGALUPI A, SCHIANO L M R. Ultrasonic time-of-flight estimation through unscented Kalman filter[J]. IEEE Transactions on Instrumentation and Measurement, 2006, 55(4): 1077-1084. doi: 10.1109/TIM.2006.877748
[13]	SHAO W, SUN H, WANG Y, et al. A multi-level damage classification technique of aircraft plate structures using Lamb wave-based deep transfer learning network[J]. Smart Materials and Structures, 2022, 31(7): 075019. doi: 10.1088/1361-665X/ac726f
[14]	LIAO Y, WANG Y, ZENG X, et al. Multiscale 1-DCNN for damage localization and quantification using guided waves with novel data fusion technique and new self-attention module[J]. IEEE Transactions on Industrial Informatics, 2024, 20(1): 492-502. doi: 10.1109/TII.2023.3268442
[15]	SIKDAR S, LIU D, KUNDU A. Acoustic emission data based deep learning approach for classification and detection of damage-sources in a composite panel[J]. Composites (Part B): Engineering, 2022, 228: 109450. doi: 10.1016/j.compositesb.2021.109450
[16]	RAUTELA M, GOPALAKRISHNAN S. Ultrasonic guided wave based structural damage detection and localization using model assisted convolutional and recurrent neural networks[J]. Expert Systems With Applications, 2021, 167: 114189. doi: 10.1016/j.eswa.2020.114189
[17]	ZHANG H, HUA J, LIN J, et al. Damage localization with Lamb waves using dense convolutional sparse coding network[J]. Structural Health Monitoring, 2015, 22: 1180-1192.
[18]	张巧灵, 高淑萍, 何迪, 等. 基于时间序列的混合神经网络数据融合算法[J]. 应用数学和力学, 2021, 42(1): 82-91. doi: 10.21656/1000-0887.410056 ZHANG Qiaoling, GAO Shuping, HE Di, et al. A hybrid neural network data fusion algorithm based on time series[J]. Applied Mathematics and Mechanics, 2021, 42(1): 82-91. (in Chinese) doi: 10.21656/1000-0887.410056
[19]	DAI J, QI H, XIONG Y, et al. Deformable convolutional networks[C]// 2017 IEEE International Conference on Computer Vision (ICCV). Venice, Italy: IEEE, 2017: 764-773.
[20]	RAVENSCROFT W, GOETZE S, HAIN T. Deformable temporal convolutional networks for monaural noisy reverberant speech separation[C]// 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Rhodes Island: IEEE, 2023: 1-5.