Composite materials are widely used due to their excellent properties. Accurately recognising damage to composite structures is of great significance for ensuring structural safety and extending service life. This study proposes a damage identification method that fuses deep learning with multiscale and multidirectional image phases. The Gabor filter bank, which takes into account the anisotropic properties of composite materials, is used to generate phase maps of composite structure images at different scales and directions. Anisotropic diffusion models are constructed for composite structure images and used as input to the network. A Gabor-based anisotropic filter enhanced attention convolutional long short-term memory (ConvLSTM) network is designed to detect and categorise damage states in composite structures. Cracking and delamination damage were considered in the finite element analysis used to produce the dataset for training the network model. A damage inversion experiment shows that the length and area errors do not exceed 4.5%. The detection results were compared and validated against those of the mainstream convolutional neural network, LSTM and ConvLSTM networks. Among these networks, the proposed algorithm achieves the best recognition results.
KapidzicZNylanderAM.Fatigue and failure testing of a hybrid CFRP-aluminum wing box at elevated temperature. Compos Struct2023; 305: 116469.
2.
LiSYTianXXLiQB, et al. Advancing structural health monitoring: deep learning-enhanced quantitative analysis of damage in composite laminates using surface strain field. Compos Sci Technol2024; 258: 110880.
3.
AzadMMKimHS.Noise robust damage detection of laminated composites using multichannel wavelet-enhanced deep learning model. Eng Struct2025; 322: 119192.
4.
WangZTZhuPLiuL, et al. An accurate and robust strain field smoothing method based on polynomial fitting and anisotropic diffusion in digital image correlation. Measurement2025; 225: 113947.
5.
WangMHHeMYLiangZH, et al. Fatigue damage monitoring of composite laminates based on acoustic emission and digital image correlation techniques. Compos Struct2023; 321: 117239.
6.
LiMZLiuGMaoZ.Phased-based motion estimation through short-distance Hilbert transform. Mech Syst Signal Process2024; 211: 111219.
7.
ChenNLiYXiangH.A new simulation algorithm of multivariate short-term stochastic wind velocity field based on inverse fast Fourier transform. Eng Struct2014; 80: 251–259.
8.
BarileCCasavolaCPappaletteraG, et al. Acousto-ultrasonic evaluation of interlaminar strength on CFRP laminates. Compos Struct2019; 208: 796–805.
9.
KharratMRamassoEPlacetV, et al. A signal processing approach for enhanced acoustic emission data analysis in high activity systems: application to organic matrix composites. Mech Syst Signal Process2016; 70–71: 103855.
10.
XuDLiuPFChenZP.Damage mode identification and singular signal detection of composite wind turbine blade using acoustic emission. Compos Struct2021; 255: 112954.
11.
KatariyaPVMeharKPandaSK.Nonlinear dynamic responses of layered skew sandwich composite structure and experimental validation. Int J Non-Lin Mech2020; 125: 103527.
12.
TaoCZhangCJiH, et al. Fatigue damage characterization for composite laminates using deep learning and laser ultrasonic. Compos Part B Eng2021; 216: 108816.
ChaYJAliRLewisJ, et al. Deep learning-based structural health monitoring. Autom Constr2024; 161: 105328.
15.
Alvarez-MontoyaJCarvajal-CastrillonASierra-PerezJ.In-flight and wireless damage detection in a UAV composite wing using fiber optic sensors and strain field pattern recognition. Mech Syst Signal Process2020; 136: 106526.
16.
XuGYueQLiuX.Real-time monitoring of concrete crack based on deep learning algorithms and image processing techniques. Adv Eng Inform2023; 58: 102214.
17.
RautelaMSenthilnathJMonacoE, et al. Delamination prediction in composite panels using unsupervised-feature learning methods with wavelet-enhanced guided wave representations. Compos Struct2022; 291: 115579.
18.
WangZChaYJ.Unsupervised deep learning approach using a deep auto-encoder with a one-class support vector machine to detect damage. Struct Health Monit2021; 20: 406–425.
19.
LecunYBottouLBengioY, et al. Gradient-based learning applied to document recognition. Proc IEEE1998; 86(11): 2278–2324.
20.
HintonGESalakhutdinovRR.Reducing the dimensionality of data with neural networks. Science2006; 313: 504–507.
21.
HopfieldJJ.Neural networks and physical systems with emergent collective computational abilities. Proc Natl Acad Sci USA1982; 79(8): 2554–2558.
22.
ElmanJL.Finding structure in time. Cogn Sci1990; 14(2): 179–211.
23.
ReddyAIndragandhiVRaviL, et al. Detection of Cracks and damage in wind turbine blades using artificial intelligence-based image analytics. Measurement2019; 147: 106823.
24.
BengioYSimardPFrasconiP.Learning long-term dependencies with gradient descent is difficult. IEEE Trans Neural Netw1994; 5(2): 157–166.
ChoeDEKimHCKimMH.Sequence-based modeling of deep learning with LSTM and GRU networks for structural damage detection of floating offshore wind turbine blades. Renew Energy2021; 174: 218–235.
27.
HuoLZTangPZhangZ, et al. Semisupervised classification of remote sensing images with hierarchical spatial similarity. IEEE Geosci Remote Sens Lett2015; 12(1): 150–154.
28.
ShiXJChenZRWangH, et al. Convolutional LSTM network: a machine learning approach for precipitation nowcasting. Adv Neural Inf Process Syst, 2015; 28: 802–810.
29.
GhimireSYaseenZMFarooqueAA, et al. Streamflow prediction using an integrated methodology based on convolutional neural network and long short-term memory networks. Sci Rep2021; 11(1): 7834616.
30.
MoishinMDeoRCPrasadR, et al. Designing deep-based learning flood forecast model with ConvLSTM hybrid algorithm. IEEE Access2021; 9: 50982–50993.
31.
HuXDWangXQMengFJ, et al. Gabor-CNN for object detection based on small samples. Defence Technol2020; 16(6): 1116–1129.
32.
DaugmanJG.Uncertainty relation for resolution in space. Spatial frequency and orientation optimized by two-dimensional visual cortical filters. J Opt Soc Am A1985; 2(7): 1160–1169.
33.
WiskottLFellousJMKrugerN, et al. Face recognition by elastic bunch graph matching. IEEE Trans Pattern Anal Mach Intell1997; 19(7): 775–779.
34.
PeronaPMalikJ.Scale space and edge detection using anisotropic diffusion. IEEE Trans Pattern Anal Mach Intell1990; 12(7): 629–639.
35.
YuMTaoBALiXW, et al. Local and long-range convolutional LSTM network: a novel multi-step wind speed prediction approach for modeling local and long-range spatial correlations based on ConvLSTM. Eng Appl Artif Intell2024; 130: 107613.
36.
WeiAGuanSWangN, et al. Damage detection of jacket platforms through improved stacked autoencoder and softmax classifier. Ocean Eng2024; 306: 118036.
37.
KrizhevskyASutskeverIHintonGE.Imagenet classification with deep convolutional neural networks. Commun ACM2017; 60(6): 84–90.
38.
SuHLiYXuYF, et al. A review of deep-learning-based super-resolution: from methods to applications. Pattern Recognit2025; 157: 110935.
39.
AbediniaOAmjadyNZareipourH.A new feature selection technique for load and price forecast of electrical power systems. IEEE Trans Power Syst2017; 32(1): 62–74.
40.
XuXSXiaoGangSunLi, et al. Associative reasoning for engineering drawings using an interactive attention mechanism. Autom Constr2025; 170: 105942.
41.
JieZHZengMRZhouXH, et al. Two stream CNN with attention-ConvLSTM on human behavior recognition. J Chin Comput Syst2021; 42(2): 405–408.
42.
WangQHuangYJiaWJ, et al. FACLSTM: ConvLSTM with focused attention for scene text recognition. Sci China Inf Sci2020; 2: 35–48.
43.
VillegasRYangJHongS, et al. Decomposing motion and content for natural video sequence prediction. In: International Conference on Learning Representations (ICLR). 24–26 April 2017. Toulon, France.
44.
WangXGirshickRGuptaA, et al. Non-local neural networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), Salt Lake City, UT, USA, 18–23 June 2018, pp. 7794–7803.
45.
CiangCCLeeJRBangHJ.Structural health monitoring for a wind turbine system: a review of damage detection methods. Meas Sci Technol2008; 19(12): 310–314.
46.
YangJD.Operation, maintenance, and repair of wind farms. Beijing: China Water & Power Press, 2014.
47.
GambleKPillingMWilsonA.An automated finite element analysis of the initiation and growth of damage in carbon fibre composite materials. Compos Struct1995; 32(1): 265–274.
48.
AtayaSAhmedM.Damages of wind turbine blade trailing edge: forms, location, and root causes. Eng Fail Anal2013; 35: 480–488.
49.
ASTM. D3039 standard test method for tensile properties of polymer matrix composite materials. West Conshohocken, PA: ASTM International, 2017.
50.
ZhangTBiswalSWangY.SHMnet: condition assessment of bolted connection with beyond human-level performance. Struct Health Monit2020; 19(4): 1188–1201.
51.
YuYWangCGuX, et al. A novel deep learning-based method for damage identification of smart building structures. Struct Health Monit2019; 18(1): 143–163.
