Leakage-Free Segmentation-Based ANN for Crack Localization and Depth Estimation in Cantilever Beams
DOI:
https://doi.org/10.59746/byt32350Keywords:
Structural health monitoring, cantilever beams, relative frequency shift, crack localization, crack depth estimation, artificial neural networksAbstract
This study develops a controlled and leakage-free learning framework for vibration-based crack characterization in a metallic cantilever beam using compact relative frequency shift (RFS) features. A fully numerical methodology is implemented in MATLAB based on a one-dimensional Euler–Bernoulli finite-element model, in which the cracked beam element is represented through a local stiffness matrix and statically condensed into an equivalent element for global assembly. A database of 1200 labelled damaged scenarios is generated by varying crack location over 100 beam elements and crack depth over 12 levels from 0.20 to 2.40 mm. Each scenario is represented by a six-dimensional RFS vector computed from the first six bending natural frequencies, while the outputs are the normalized crack location and crack-depth ratio. All models are trained and evaluated under a single leakage-free Master Split (70/15/15) to ensure fair and reproducible comparison. The baseline artificial neural network achieves test-set mean absolute errors of 2.992 mm for crack location and 0.0251 mm for crack depth. Under the same split, a segmentation-based coarse-to-fine strategy reduces these errors to 1.245 mm and 0.0211 mm, respectively. The main contribution of this work is a reproducible leakage-free segmentation-based ANN framework that improves RFS-based crack localization and depth estimation within a controlled numerical setting.
References
[1] A. A. S. R. de Sousa, J. da Silva Coelho, M. R. Machado, and M. Dutkiewicz, “Multiclass Supervised Machine Learning Algorithms Applied to Damage and Assessment Using Beam Dynamic Response,” J. Vib. Eng. Technol., vol. 11, no. 6, pp. 2709–2731, Sep. 2023, doi: 10.1007/s42417-023-01072-7.
[2] D. M. Ashigbi, M. N. Sackey, Y. A. K. Fiagbe, and J. Quaye-Ballard, “Vibration response based crack diagnosis in beam-like structures using fuzzy inference system,” Sci. African, vol. 14, p. e01051, Nov. 2021, doi: 10.1016/j.sciaf.2021.e01051.
[3] R. Katam, V. D. K. Pasupuleti, and P. Kalapatapu, “Machine learning-driven structural health monitoring: STFT-based feature extraction for damage detection,” Structures, vol. 78, no. 3, p. 109244, Aug. 2025, doi: 10.1016/j.istruc.2025.109244.
[4] M. Parziale, P. H. Silva, M. Giglio, and F. Cadini, “Explainability of convolutional neural networks for damage diagnosis using transmissibility functions,” Structures, vol. 69, no. January, p. 107583, Nov. 2024, doi: 10.1016/j.istruc.2024.107583.
[5] M. Moravvej and M. El-Badry, “Reference-Free Vibration-Based Damage Identification Techniques for Bridge Structural Health Monitoring—A Critical Review and Perspective,” Sensors, vol. 24, no. 3, p. 876, Jan. 2024, doi: 10.3390/s24030876.
[6] N. Gillich et al., “Beam Damage Assessment Using Natural Frequency Shift and Machine Learning,” Sensors, vol. 22, no. 3, p. 1118, Feb. 2022, doi: 10.3390/s22031118.
[7] S. K. Gupta and S. Das, “Multiple Damage Identification in a Beam Using Artificial Neural Network-Based Modified Mode Shape Curvature,” Arab. J. Sci. Eng., vol. 47, no. 4, pp. 4849–4864, Apr. 2022, doi: 10.1007/s13369-021-06267-2.
[8] M.-H. Tran et al., “A novel approach for identifying structural damage in beams with diverse boundary conditions using modal strain energy method combined with convolutional neural networks,” Structures, vol. 80, no. September, p. 110062, Oct. 2025, doi: 10.1016/j.istruc.2025.110062.
[9] S. M. H. Pooya and A. Massumi, “A novel and efficient method for damage detection in beam-like structures solely based on damaged structure data and using mode shape curvature estimation,” Appl. Math. Model., vol. 91, pp. 670–694, Mar. 2021, doi: 10.1016/j.apm.2020.09.012.
[10] Y. Zhang, Z. Xu, G. Li, and C. Xin, “An enhanced hybrid method for structural damage detection using DOG multi-scale analysis and adaptive fuzzy clustering,” Eng. Fail. Anal., vol. 179, no. April, p. 109785, Sep. 2025, doi: 10.1016/j.engfailanal.2025.109785.
[11] P. Saha and M. Yang, “A Neural Network Approach to Estimate the Frequency of a Cantilever Beam with Random Multiple Damages,” Sensors, vol. 23, no. 18, p. 7867, Sep. 2023, doi: 10.3390/s23187867.
[12] S. J. Seyed Hakim, M. J. Irwan, M. H. W. Ibrahim, and S. S. Ayop, “Structural damage identification employing hybrid intelligence using artificial neural networks and vibration-based methods,” J. Appl. Res. Technol., vol. 20, no. 2, pp. 221–236, May 2022, doi: 10.22201/icat.24486736e.2022.20.2.1233.
[13] M. Torzoni, M. Tezzele, S. Mariani, A. Manzoni, and K. E. Willcox, “A digital twin framework for civil engineering structures,” Comput. Methods Appl. Mech. Eng., vol. 418, no. PB, p. 116584, Jan. 2024, doi: 10.1016/j.cma.2023.116584.
[14] X. Zhang and Y. Lu, “A novel wavelet energy feature for damage identification with a digital twin considering measurement uncertainties,” J. Struct. Integr. Maint., vol. 10, no. 1, Jan. 2025, doi: 10.1080/24705314.2024.2440830.
[15] R. S. Battu, K. Agathos, J. M. Londoño Monsalve, K. Worden, and E. Papatheou, “Combining transfer learning and numerical modelling to deal with the lack of training data in data-based SHM,” J. Sound Vib., vol. 595, no. December 2023, p. 118710, Jan. 2025, doi: 10.1016/j.jsv.2024.118710.
[16] A. S. Yehia, D. K. Harris, and A. A. Aljundi, “What lies within: Utilizing graph neural networks for subsurface detection in finite element simulations,” Eng. Struct., vol. 341, no. June, p. 120842, Oct. 2025, doi: 10.1016/j.engstruct.2025.120842.
[17] M. Torzoni, A. Manzoni, and S. Mariani, “A multi-fidelity surrogate model for structural health monitoring exploiting model order reduction and artificial neural networks,” Mech. Syst. Signal Process., vol. 197, no. November 2022, p. 110376, Aug. 2023, doi: 10.1016/j.ymssp.2023.110376.
[18] C. V. Rusu, G.-R. Gillich, C. Tufisi, N. Gillich, T. H. Bui, and C. Ionut, “A Stacked Neural Network Model for Damage Localization,” Sensors, vol. 24, no. 21, p. 7019, Oct. 2024, doi: 10.3390/s24217019.
[19] J. Fish and T. Belytschko, A First Course in Finite Elements. Wiley, 2007. doi: 10.1002/9780470510858.
[20] D. Imamović and M. Skrinar, “Static bending analysis of a transversely cracked strip tapered footing on a two-parameter soil using a new beam finite element,” Contin. Mech. Thermodyn., vol. 36, no. 3, pp. 571–584, May 2024, doi: 10.1007/s00161-024-01283-7.
[21] A.-T. Aman, C. Tufisi, G.-R. Gillich, and Z. I. Praisach, “Detection of transverse cracks in steel beams using damage location coefficients and artificial neural networks,” Vibroengineering Procedia, vol. 50, pp. 42–48, Sep. 2023, doi: 10.21595/vp.2023.23432.
