准确识别与定位内部缺陷对于确保混凝土结构的安全性和耐久性至关重要，冲击回波法针对只具备单一检查面的混凝土结构具有独特优势，利用冲击回波法进行混凝土结构内部缺陷测试时，其难点在于冲击回波信号特征的分析及缺陷分类识别。提出基于冲击回波信号时频图深度学习的混凝土内部缺陷识别方法：首先，对含不同缺陷类型（孔洞、蜂窝、裂缝）混凝土面板开展冲击回波试验；然后，将原始冲击回波信号通过短时傅里叶变换和连续小波变换转换为时频图；最后，将两类时频图组合起来进行数据扩充，构建包含缺陷类型、孔洞大小、裂缝深度时频图的数据集，采用Swin Transformer、Vision Transformer、ResNet18三类深度学习模型进行训练，根据深度学习模型相关评估指标，分析不同数据集下不同深度学习模型的表现情况，对比模型对缺陷的分类识别效果。结果表明，三种模型中Swin Transformer表现最佳，在缺陷类型分类中准确率超过95.00%，识别性能良好。
Abstract
Accurately identifying and locating internal defects is crucial for ensuring safety and durability of concrete structures.The impact echo (IE) method has unique advantages for concrete structures with only a single inspection surface.When using IE method for testing internal defects in concrete structures, the difficulty lies in analyzing features of IE signals and classifying and identifying defects.Here, a method for identifying internal defects in concrete was proposed based on deep learning of time-frequency maps of IE signals.Firstly, IE tests were conducted for concrete slabs containing different types of defects including holes, honeycombs and cracks.Then, the original IE signals were converted into time-frequency maps with short-time Fourier transform and continuous wavelet transform, respectively.Finally, 2 types of time-frequency maps were combined for data expansion to construct datasets containing defect types, hole sizes and crack depth time-frequency maps.3 types of deep learning models of Swin Transformer, Vision Transformer and ResNet18 were used for training.Based on the relevant evaluation indexes of deep learning models, performances of different deep learning models under different datasets were analyzed, and the classification and recognition effects of models on defects were compared.The results showed that Swin Transformer performs the best among 3 models, its correctness rate is over 95.00% in defect type classification, and it has good recognition performance.
关键词
混凝土 /
冲击回波(IE) /
深度学习 /
缺陷 /
分类识别
{{custom_keyword}} /
Key words
concrete /
impact echo (IE) /
deep learning /
defect /
classification and recognition
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
［1］KASHIF UR REHMAN S, IBRAHIM Z, MEMON S A, et al.Nondestructive test methods for concrete bridges: a review［J］.Construction and Building Materials, 2016, 107: 58-86.
［2］ZHANG Y, XIE Z.Ensemble empirical mode decomposition of impact-echo data for testing concrete structures［J］.NDT & E International, 2012, 51: 74-84.
［3］郭贵强, 乔瑞社, 解伟, 等.基于冲击回波法的水工混凝土板内部缺陷检测试验研究［J］.水力发电, 2015, 41(4): 95-99.
GUO Guiqiang, QIAO Ruishe, XIE Wei, et al.Experimental study on internal defect detection in hydraulic concrete slab based on impact-echo method［J］.Water Power, 2015, 41(4): 95-99.
［4］邹春江, 陈征宙, 董平, 等.箱梁中冲击回波主频影响因素研究及应用［J］.振动与冲击, 2010, 29(7): 126-131.
ZOU Chunjiang, CHEN Zhengzhou, DONG Ping, et al.Influencing factors on impact-echo characteristic frequency in a box beam［J］.Journal of Vibration and Shock, 2010, 29(7): 126-131.
［5］YEH P L, LIU P L.Application of the wavelet transform and the enhanced Fourier spectrum in the impact echo test［J］.NDT & E International, 2008, 41(5): 382-394.
［6］姚菲, 陆幸奇, 陈光宇.基于冲击回波法的混凝土-围岩缺陷检测与信号处理研究［J］.铁道科学与工程学报, 2021, 18(9): 2316-2323.
YAO Fei, LU Xingqi, CHEN Guangyu.Experimental and signal processing research on concrete-rock structural defects by impact-echo method［J］.Journal of Railway Science and Engineering, 2021, 18(9): 2316-2323.
［7］张景奎, 崔德密, 严卫中.混凝土缺陷智能化快速检测与定量识别技术研究［J］.科学技术与工程, 2016, 16(19): 288-293.
ZHANG Jingkui, CUI Demi, YAN Weizhong.On nondestructive testing and intelligent quick identification of concrete defects［J］.Science Technology and Engineering, 2016, 16(19): 288-293.
［8］康玉祥, 陈果, 尉询楷, 等.基于残差网络的航空发动机滚动轴承故障多任务诊断方法［J］.振动与冲击, 2022, 41(16): 285-293.
KANG Yuxiang, CHEN Guo, WEI Xunkai, et al.A multi-task fault diagnosis method of rolling bearings based on the residual network［J］.Journal of Vibration and Shock, 2022, 41(16): 285-293.
［9］XU J, YU X.Detection of concrete structural defects using impact echo based on deep networks［J］.Journal of Testing and Evaluation, 2021, 49(1): 109-120.
［10］DORAFSHAN S, AZARI H.Evaluation of bridge decks with overlays using impact echo, a deep learning approach［J］.Automation in Construction, 2020, 113: 103133.
［11］DORAFSHAN S, AZARI H.Deep learning models for bridge deck evaluation using impact echo［J］.Construction and Building Materials, 2020, 263: 120109.
［12］罗炜, 薛亚东, 贾非, 等.基于深度学习的无砟轨道砂浆层脱空病害识别［J］.现代隧道技术, 2021, 58(增刊1): 129-136.
LUO Wei, XUE Yadong, JIA Fei, et al.Identification of voids in mortar layer of ballastless track based on deep learning［J］.Modern Tunnelling Technology, 2021, 58(Suppl.1): 129-136.
［13］TORLAPATI R, AZARI H, SHOKOUHI P.Classification of impact echo signals using explainable deep learning and transfer learning approaches［J］.Transportation Research Record: Journal of the Transportation Research Board, 2023, 2677(9): 464-477.
［14］SHANG G, CHEN J.Visualization of concrete internal defects based on unsupervised domain adaptation algorithm for transfer learning of experiment-simulation hybrid dataset of impact echo signals［J］.Structural Health Monitoring, 2024, 23(3): 1875-1889.
［15］LI Z.A study on ultrasonic echo tomography for non-destructive evaluation of hardened cementitious concrete［J］.Journal of Building Pathology and Rehabilitation, 2020, 5(1): 1-18.
［16］WU Y, ZHANG J, GAO C, et al.Internal defect detection quantification and three-dimensional localization based on impact echo and classification learning model［J］.Measurement, 2023, 218: 113153.
［17］ZHU H, SUN M, FU H, et al.Training a seismogram discriminator based on ResNet［J］.IEEE Transactions on Geoscience and Remote Sensing, 2021, 59(8): 7076-7085.
［18］DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al.An image is worth 16×16 words: transformers for image recognition at scale［C］//International Conference on Learning Representations.San Diego, CA: ICLR, 2021.
［19］LIU Z, LIN Y, CAO Y, et al.Swin Transformer: hierarchical Vision Transformer using shifted windows［C］//2021 IEEE/CVF International Conference on Computer Vision.Montreal: ICCV, 2021.
［20］ZHOU Y, WANG X, ZHANG M, et al.MPCE: a maximum probability based cross entropy loss function for neural network classification［J］.IEEE Access, 2019, 7: 146331.
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(3893 KB)
