Damage detection of aircraft skin in complex regions remains a persistent challenge in the aerospace industry due to the compact structures and limited accessibility of areas such as wing-fuselage junctions, engine nacelles, and flaps. This article presents a miniaturized damage detection device tailored for such environments and investigates the physical characteristics of the acquired signals, including impact-response behavior, energy-induced temporal causality, and spatial spike distributions. These characteristics are leveraged to design a spatio-temporal neural network architecture. The detection process is modeled as an impact-response dynamic system to guide the construction of wavelet-based convolutional kernels with learnable scale, translation, damping, and phase parameters. A causal convolutional network captures temporal dependencies of spike features, while a graph attention network extracts spatial features based on transformed spike intervals and an adjacency matrix. A point-wise attention mechanism is then applied to fuse multi-dimensional features for final classification. Experimental results demonstrate that the proposed method outperforms existing approaches and provides strong interpretability.
GuoDWuEQWuY, et al. Flightbert: binary encoding representation for flight trajectory prediction. IEEE Trans Intell Transp Syst2022; 24(2): 1828–1842.
2.
SandamaliGGNSuRSudheeraKLK, et al. A safety-aware real-time air traffic flow management model under demand and capacity uncertainties. IEEE Trans Intell Transp Syst2021; 23(7): 8615–8628.
3.
HuZXangZ. Damage detection for continuous beams by using the tap-scan method. Appl Math Model2024; 135: 524–540.
4.
WangFLiuJDongB, et al. Optimization of thermal-wave radar thermography by transverse heat flow suppression technique for accurate defect detection of CFRP laminates. IEEE Trans Instrument Measure2021; 70: 1–10.
5.
GhandourahEEHamidiSHAMohd SallehKA, et al. Evaluation of welding imperfections with X-ray computed laminography for NDT inspection of carbon steel plates. J Nondestruct Evaluation2023; 42(3): 77.
6.
ZhangHZhangMZhangH, et al. High quality ultrasonic imaging at low detection frequency for defects in thick composites. J Nondestruct Evaluation2025; 44(1): 21.
7.
HuZLinPGuoH, et al. Detect the stiffness transition in beam structures by using the passive tap-scan method. Mech Syst Sig Process2023; 192: 110211.
8.
ZhuZKangRQinK. Design of metamaterial sensor for non destructive testing of aircraft composite skin damage. Micromach2025; 16(3): 284.
9.
WuJZhangYShanT, et al. An additive feature fusion attention based on YOLO network for aircraft skin damage detection. J Supercomput2025; 81(4): 627.
10.
ChenHXuZQuZ, et al. Review of nondestructive testing methods for honeycomb sandwich structures. Mater Rep2024; 38(10): 296–310. (in Chinese)
11.
LiuZLiYCaoY, et al. Research on knock detection defect recognition based on EM clustering. Aeronaut Comput Tech2022; 52(2): 53–57. (in Chinese)
12.
RabcanJLevashenkoVZaitsevaE, et al. Non-destructive diagnostic of aircraft engine blades by fuzzy decision tree. Eng Struct2019; 197: 109396.
13.
ShrifanNHJawadGNIsaNAM, et al. Microwave nondestructive testing for defect detection in composites based on K-means clustering algorithm. IEEE Access2020; 9: 4820–4828.
14.
GaoMHuoLWangF, et al. CNN-based similar microwave reflection signals for improved detectability and intelligent characterization of internal defects in composite materials. J Nondestruct Evaluation2025; 44(1): 28.
15.
ZhangFWangLYeW. Aluminum plate defect detection based on multilevel LSTM. Chinese J Theoretic Appl Mech2023; 55(11): 2566–2576. (in Chinese)
16.
ZhangYLiuC. Real-time pavement damage detection with damage shape adaptation. IEEE Trans Intell Transp Syst2024; 25(11): 18954–18963.
17.
NingSWuWWuY, et al. Rail track surface defect detection algorithm integrating MobileNetv3 and transformer. ICAIT2024: 13401: 47–52.
18.
LiuJLiQLiL, et al. Structural damage detection and localization via an unsupervised anomaly detection method. Reliab Eng Syst Safe2024; 252: 110465.
19.
GrezmakJWangPSunC, et al. Explainable convolutional neural network for gearbox fault diagnosis. Procedia CIRP2019; 80: 476–481.
20.
ZhouBKhoslaALapedrizaA, et al. Learning deep features for discriminative localization. In: Proceedings of the IEEE conference on computer vision and pattern recognition, Las Vegas, NV, 12 December 2016, pp.2921–2929. New York: IEEE.
21.
BojarskiMYeresPChoromanskaA, et al. Explaining how a deep neural network trained with end-to-end learning steers a car. arxiv:1704.07911, 2017.
22.
RavanelliMBengioY. Speaker recognition from raw waveform with sincnet. In: 2018 IEEE spoken language technology workshop (SLT), 2018, Athens, Greece, 18–21 December 2018, pp.1021–1028. New York: IEEE.
23.
LiTZhaoZSunC, et al. WaveletKernelNet: an interpretable deep neural network for industrial intelligent diagnosis. IEEE Trans Syst Man Cybernet Syst2021; 52(4): 2302–2312.
24.
HanGChenJLiuL, et al. An interpretable CNN with wavelet group policy embedded for intelligent fault diagnosis. IEEE Trans Instrument Measure2024; 73: 1–15.
25.
ZhaoLMaoYand Qin YA lightweight dynamic dual-damped wavelet-based convolutional neural network for interpretable bearing fault diagnosis. Measure Sci Technol2024; 35(8): 086115.
26.
QiJMoJNiuY, et al. Recognition of fiber optic vibration signals based on laplace wavelet transform and deep learning. IEEE Trans Electri Electronic Eng2024; 19(6): 1026–1034.
27.
BaraldiPBattistiniCAntonelloF, et al. Physics-informed neural networks for the safety analysis of nuclear reactors. Prog Nucl Energy2025; 185: 105745.
28.
MadhusoodananNChiplunkarRPuliVK, et al. Physics-informed sparse causal inference for source detection of plant-wide oscillations. AIChE J2024; 70(4): e18362.
29.
SongPZhaoCHuangB. Addressing heterogeneous time–frequency causality: source consistency exploring for industrial root cause alignment and diagnosis. IEEE Trans Cybernet2024; 55: 1107–1120.
30.
ZhaoSZhangTShmaliyYS, et al. Bayesian transfer filtering using pseudo marginal measurement likelihood. IEEE Trans Cybernet2024; 55: 562–573.
31.
LiQWangFWangC. STGFP: information enhanced spatio-temporal graph neural network for traffic flow prediction. Appl Intell2025; 55(6): 485.
32.
ZhengCFanXPanS, et al. Spatio-temporal joint graph convolutional networks for traffic forecasting. IEEE Trans Knowl Data Eng2023; 36(1): 372–385.
33.
WuSHuY. Traffic flow prediction based on spatio-temporal aggregated graph neural networks. Transp Res Record2025; 2679: 573–588.
34.
JinGLiangYFangY, et al. Spatio-temporal graph neural networks for predictive learning in urban computing: a survey. IEEE Trans Knowl Data Eng2023; 36(10): 5388–5408.
35.
ZhouMLiYCaoY, et al. Physics-informed spatio-temporal hybrid neural networks for predicting remaining useful life in aircraft engine. Reliab Eng Syst Safe2025; 256: 110685.
36.
JiangGShenKLiuX, et al. Hierarchical spatio-temporal graph network for fault diagnosis of industrial processes. IEEE Int Thing J2024; 12(3): 3043–3054.
37.
HuXMaJZhangR, et al. Ts-GSAN: a two-stage graphical spatiotemporal attention network fault localization method for distributed energy systems. IEEE Trans Instrument Measure2025; 74: 3516408.
38.
LiangPWangXAiC, et al. SRSGCN: a novel multi-sensor fault diagnosis method for hydraulic axial piston pump with limited data. Reliab Eng Syst Safe2025; 253; 110563.
39.
WeiPZhuWYangY, et al. Spatiotemporal transform network-based anomaly detection and localization of distributed parameter systems. IEEE Trans Ind Inform2025; 21(4): 2770–2778.
40.
CawleyPAdamsRD. The mechanics of the coin-tap method of non-destructive testing. J Sound Vib1988; 122(2): 299–316.
41.
LiJXuLXuX, et al. Research on the distribution rule of percussion response pulse width in coin-tap test. Shock Vib2021; 2021(1): 5535687.
42.
IoffeSSzegedyC. Batch normalization: accelerating deep network training by reducing internal covariate shift. PMLR2015; 37: 448–456.
43.
VaswaniAShazeerNParmarN, et al. Attention is all you need. Adv Neur Inform Process Syst2017; 30: 1–11.
