Physically synthesized data for deep learning-based visual scratch inspection of aerospace alloys with complex geometries

Abstract

Aerospace alloys often operate under extreme conditions. Accurate defect segmentation in images of aerospace components is the key to quantifying the defects and evaluating their impact for part lifespan. The components usually have complex free-form surfaces, leading to uneven light distribution in images. The variable image presentations pose a great challenge for accurate segmentation, especially with limited data. Generative adversarial networks and other training-based methods are commonly used for image generation, but they still rely on sufficient high-quality training data. In this paper, a physical-based image generation method is proposed to create any possible scratches according to physical laws to improve the scratch segmentation capability with limited data. First, an efficient scratched blade surface image generation pipeline is developed. Then, a systematic strategy to maximize the effect of physical synthetic scratch images is presented. The experiments show that the segmentation intersection-over-union could be improved from 0.66 to 0.83 with only 20 real images for training, and reveal the influences of network structure, image and label quality, data fusion strategy on segmentation performance.

Keywords

Automatic visual inspection / physical image synthesis / defect segmentation / deep learning / aerospace alloys

References

[1] Aust J.Taxonomy of gas turbine blade defects.Aerospace2019;6:58

[2] Han K,Yao C.Evolution and anti-fatigue mechanism of surface characteristics of Ti60 alloy induced by ball burnishing and shot peening during tensile-compression fatigue.Eng Fail Anal2024;159:108136

[3] Zhang T,Dong R.Physics-guided neural network for grinding temperature prediction.J Eng Design2024;36:707-30

[4] Sun Y,Tan L.Surface integrity and fatigue failure behavior of nickel based alloy blades: after cutting, vibration finishing and shot peening.Eng Fail Anal2025;167:109034

[5] Ma Y,Zhang Y,Wang Y.SPDP-Net: a semantic prior guided defect perception network for automated aero-engine blades surface visual inspection.IEEE Trans Automat Sci Eng2025;22:2724-33

[6] Shang H,Liu J,Yan R.Defect-aware transformer network for intelligent visual surface defect detection.Adv Eng Inform2023;55:101882

[7] Wang WY,Li G.Artificial intelligence enabled smart design and manufacturing of advanced materials: the endless Frontier in AI⁺ era.Mater Genome Eng Adv2024;2:e56

[8] Wang WY,Liu Z.Materials genome engineering and intelligent science: the endless frontier in AI+ era.Sci Technol Rev2025;43:93-109

[9] Huang K,Zhao J.Weakly supervised complex texture defect detection based on nested U-Net architecture.J Nondestruct Eval2025;44:24

[10] Abdulrahman Y,Ayyad A,Zweiri Y.Aero-engine blade defect detection: a systematic review of deep learning models.IEEE Access2023;11:53048-61

[11] Hu S,Zhang Y.Application of self-supervised learning in steel surface defect detection.J Mater Inf2025;5:44

[12] Wu Z,Kruzic JJ.Critical damage events of 3D printed AlSi10Mg alloy via in situ synchrotron X-ray tomography.Acta Mater2025;282:120464

[13] Li D,Xie Q,Yu Z.Tiny defect detection in high-resolution aero-engine blade images via a coarse-to-fine framework.IEEE Trans Instrum Meas2021;70:1-12

[14] Li X,Ju H.Surface defect detection model for aero-engine components based on improved YOLOv5.Appl Sci2022;12:7235

[15] Song K,Ma S.Surface defect detection of aeroengine blades based on cross-layer semantic guidance.IEEE Trans Instrum Meas2023;72:1-11

[16] Shang H,Sun C,Yan R.Superpixel perception graph neural network for intelligent defect detection of aero-engine blade.J Manuf Syst2024;77:112-26

[17] Sun X,Wen X,Yan Y.SDD-DETR: surface defect detection for no-service aero-engine blades with detection transformer.IEEE Trans Automat Sci Eng2025;22:6984-97

[18] Zong Y,Wang H.An intelligent and automated 3D surface defect detection system for quantitative 3D estimation and feature classification of material surface defects.Opt Lasers Eng2021;144:106633

[19] Aggour KS,Ruscitto D.Artificial intelligence/machine learning in manufacturing and inspection: a GE perspective.MRS Bull2019;44:545-58

[20] Zhang S,Wang X.Large language models enabled intelligent microstructure optimization and defects classification of welded titanium alloys.J Mater Inf2024;4:34

[21] Wang Y,Wang W.High-efficient view planning for surface inspection based on parallel deep reinforcement learning.Adv Eng Inform2023;55:101849

[22] Du Z,Li X.A new contrastive GAN with data augmentation for surface defect recognition under limited data.IEEE Trans Instrum Meas2023;72:1-13

[23] Liu L,Wu Y.Defective samples simulation through adversarial training for automatic surface inspection.Neurocomputing2019;360:230-45

[24] Ren W,Chen C.DD-Aug: a knowledge-to-image synthetic data augmentation pipeline for industrial defect detection.IEEE Trans Ind Inf2025;21:2284-93

[25] He R,Yu X. Is synthetic data from generative models ready for image recognition? arXiv 2023, arXiv:2210.07574. Available online: https://doi.org/10.48550/arXiv.2210.07574. (accessed 4 Mar 2026)

[26] Singh AR,Hazra S.Generating synthetic training images to detect split defects in stamped components.IEEE Trans Ind Inf2024;20:4816-27

[27] Bosnar L,Gospodnetic P.Procedural defect modeling for virtual surface inspection environments.IEEE Comput Graph Appl2023;43:13-22

[28] Manyar OM,Levine R,Barbič J.Physics informed synthetic image generation for deep learning-based detection of wrinkles and folds.J Comput Inform Sci Eng2023;23:030903

[29] Abou Akar C,Khalil J.SORDI.ai: large-scale synthetic object recognition dataset generation for industries.Multimed Tools Appl2024;84:18263-304

[30] Gutierrez P,Cordier A,Schappert M. Synthetic training data generation for deep learning based quality inspection. arXiv 2021, arXiv:2104.02980. Available online: https://doi.org/10.48550/arXiv.2104.02980. (accessed 4 Mar 2026)

[31] Fulir J,Hagen H.Synthetic data for defect segmentation on complex metal surfaces. In 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Vancouver, Canada, June 17-24, 2023. IEEE; 2023. pp. 4424-34.

[32] Schraml D.Synthetic training data in AI-driven quality inspection: the significance of camera, lighting, and noise parameters.Sensors2024;24:649 PMCID:PMC10820774

[33] Bosch C,Mérillou S.A physically-based model for rendering realistic scratches.Comput Graph Forum2004;23:361-70

[34] He L,Hou W.Objective image quality assessment: a survey.Int J Comput Math2013;91:2374-88

[35] Xie E,Yu Z,Alvarez JM. SegFormer: simple and efficient design for semantic segmentation with transformers. arXiv 2021, arXiv:2105.15203. Available online: https://doi.org/10.48550/arXiv.2105.15203. (accessed 4 Mar 2026)

[36] Weng W.INet: convolutional networks for biomedical image segmentation.IEEE Access2021;9:16591-603

[37] Chen LC,Papandreou G,Adam H.Encoder-decoder with atrous separable convolution for semantic image segmentation. In: Ferrari V, Hebert M, Sminchisescu C, Weiss Y, Editors. Computer Vision - ECCV 2018. Cham: Springer International Publishing; 2018. pp. 833-51.

[38] Badrinarayanan V,Cipolla R.SegNet: a deep convolutional encoder-decoder architecture for image segmentation.IEEE Trans Pattern Anal Mach Intell2017;39:2481-95

[39] Mirza M. Conditional generative adversarial nets. arXiv 2014, arXiv:1411.1784. Available online: https://doi.org/10.48550/arXiv.1411.1784. (accessed 4 Mar 2026)

[40] Niu T,Li W,Niu S.Positive-sample-based surface defect detection using memory-augmented adversarial autoencoders.IEEE/ASME Trans Mechatron2022;27:46-57