Structural Health Monitoring of Railway Tracks from On-Board Vibration Measurements Using Deep Learning Techniques
DOI:
https://doi.org/10.63595/vetor.v36i1.18345Keywords:
Structural Health Monitoring, Variational Autoencoder, Hotelling T² Control Chart, Railway MaintenanceAbstract
Railway infrastructure maintenance presents a constant challenge due to natural wear and tear and damage resulting from continuous use, which can compromise the safety and efficiency of this means of transport. This study proposes the use of Structural Health Monitoring (SHM) through the application of a Variational Autoencoder (VAE) for the detection of irregularities in railway tracks. The VAE is trained to learn to represent the normal behavior of the track and, together with the Hotelling T² Control Chart, is capable of detecting structural anomalies. The proposed methodology includes the optimization of the VAE hyperparameters using the Optuna library, seeking to maximize the model's performance. The results indicate that the combination of the VAE with the Hotelling T² Control Chart is effective in detecting defects, offering a more economical and less invasive solution for preventive maintenance of tracks. Although the method does not yet quantify anomalies, it has great potential to improve operational reliability.
Downloads
References
[1] H. Souza e W. Courage, “Value of information of a pro-active shm-based tool supported by advanced fe models and bayesian statistics towards efficiency in bridge management,” Structure and Infrastructure Engineering, vol. 18, no. 4, pp. 554–572, 2021. Disponível em: https://doi.org/10.1080/15732479.2021.1978507
[2] R. P. Finotti, F. Barbosa, A. A. Cury, e R. L. Pimentel, “Numerical and experimental evaluation of structural changes using sparse auto-encoders and svm applied to dynamic responses,” Applied Sciences, vol. 11, no. 24, 2021. Disponível em: https://doi.org/10.3390/app112411965
[3] M. Spínola Neto, R. Finotti, F. Barbosa, e A. Cury, “Structural damage identification using autoencoders: A comparative study,” Buildings, vol. 14, no. 7, 2024. Disponível em: https://doi.org/10.3390/buildings14072014
[4] F. J. Carrión, J. A. Quintana, e S. E. Crespo, “Techno-economical and practical considerations for SHM systems,” Journal of Civil Structural Health Monitoring, vol. 7, no. 2, pp. 207–215, 2017. Disponível em: https://doi.org/10.1007/s13349-017-0215-x
[5] I. Goodfellow, Y. Bengio, e A. Courville, Deep Learning. Cambridge, Massachusetts, EUA: MIT Press, 2016. Disponível em: http://www.deeplearningbook.org
[6] R. P. Finotti, C. F. da Silva, P. H. E. Oliveira, F. Barbosa, A. A. Cury, e R. C. Silva, “Novelty detection on a laboratory benchmark slender structure using an unsupervised deep learning algorithm,” Latin American Journal of Solids and Structures, vol. 20, no. 9, 2023. Disponível em: https://doi.org/10.1590/1679-78257591
[7] D. C. Montgomery, Introduction to Statistical Quality Control. Nova York, EUA: John Wiley & Sons, 2009. Disponível em: chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://www.uaar.edu.pk/fs/books/12.pdf
[8] J. Magalhães, T. Jorge, R. Silva, A. Guedes, D. Ribeiro, A. Meixedo, M. Araliya, C. Vale, P. Montenegro, e A. A. Cury, “A strategy for out-of-roundness damage wheels identification in railway vehicles based on sparse autoencoders,” Railway Engineering Science, vol. 32, pp. 421–443, 2024. Disponível em: https://doi.org/10.1007/s40534-024-00338-4
[9] The, MathWorks, e Inc., “Matlab version: (r2018a),” Natick, Massachusetts, United States, 2018. Disponível em: https://www.mathworks.com
[10] P. A. Montenegro, S. G. M. Neves, R. Calçada, M. Tanabe, e M. Sogabe, “Wheel–rail contact formulation for analyzing the lateral train–structure dynamic interaction,” Computers & Structures, vol. 152, pp. 200–214, 2015. Disponível em: https://doi.org/10.1016/j.compstruc.2015.01.004
[11] P. A. Montenegro e R. Calçada, “Wheel–rail contact model for railway vehicle–structure interaction applications: development and validation,” Railway Engineering Science, vol. 31, no. 3, pp. 181–206, 2023. Disponível em: https://doi.org/10.1007/s40534-023-00306-4
[12] A. Guedes, R. Silva, D. Ribeiro, C. Vale, M. Araliya, P. Montenegro, e A. Meixedo, “Detection of wheel polygonization based on wayside monitoring and artificial intelligence,” Sensors, vol. 23, no. 4, 2023. Disponível em: https://doi.org/10.3390/s23042188
[13] R. Silva, A. Guedes, D. Ribeiro, C. Vale, A. Meixedo, A. Mosleh, e P. Montenegro, “Early identification of unbalanced freight traffic loads based on wayside monitoring and artificial intelligence,” Sensors, vol. 23, no. 3, 2023. Disponível em: https://doi.org/10.3390/s23031544
[14] A. Mosleh, A. Meixedo, D. Ribeiro, P. Montenegro, e R. Calçada, “Early wheel flat detection: an automatic data-driven wavelet-based approach for railways,” Vehicle System Dynamics, vol. 61, no. 6, pp. 1664–1673, 2022. Disponível em: https://doi.org/10.1080/00423114.2022.2103436
[15] H. Hertz, Ueber die Berührung fester elastischer Körper. Germany: Journal für die reine und angewandte Mathematik, 1882. Disponível em: http://eudml.org/doc/148490
[16] J. J. Kalker, Book of Tables for the Hertzian Creep-force Law. Netherlands: Faculty of Technical Mathematics and Informatics, Delft University of Technology, 1996. Disponível em: https://books.google.com.br/books?id=Yrf5GgAACAAJ
[17] T. Akiba, S. Sano, T. Yanase, T. Ohta, e M. Koyama, “Optuna: A next-generation hyperparameter optimization framework,” em KDD ’19: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, A. for Computing Machinery, Ed., Ankur Teredesai and Vipin Kumar. Anchorage, EUA: 25 July 2019, 2019, pp. 2623 – 2631. Disponível em: https://doi.org/10.1145/3292500.3330701
Downloads
Published
How to Cite
Issue
Section
