V.A. Kanarsky1

1 Far Eastern State University of Railway Transport (Khabarovsk, Russia)

1 jizzierose@yahoo.com

Problem setting: Railway infrastructure is a set of complex technical systems. The railway switches used at stations are mostly maintained in a planned manner, which does not guarantee the occurrence of faults between regular inspections. The existing railway automation monitoring systems are not capable of providing recommendations on impending breakdowns. They only record oscillograms of electrical parameters from the power supply circuits of the switches. These oscillograms can be an indicator of one or another pre-failure state.

Target: to develop a state classifier of a switch based on a recurrent neural network capable of determining the state of a switch in automatic mode by the active power graph.

Results: analyzed the technical possibilities of providing data from the power supply circuits of the switch actuator; considered the existing solutions for processing numerical sequences; selected the topology of recurrent neural network; we proposed the classifier architecture based on Elman networks and Gated Recurrent Unit; made a comparative analysis of learning curves of each architecture. In the process of training the classifier based on the controlled recurrent block, high classification accuracy is achieved.

Practical significance: the proposed method of automatic determination of the state of the switch can be used by companies-manufacturers of railway automation monitoring systems for processing analog telemetry measurements coming from the engine of the switch mechanism.  The use of neural network classifier will allow interpreting active power graphs in terms of existing faults of the switch, thus providing support to maintenance electromechanics when making a decision on repair. In turn, this approach will prevent accidents, unplanned downtime and train delays at the station.

Kanarsky V.A. Determination of railway switch state using a recurrent neural network classifier. Neurocomputers. 2024. V. 26. № 1. Р. 23-31. DOI: https://doi.org/10.18127/j19998554-202401-03 (In Russian)

1. Veselova A.S. Assessment of the quality of technical operation of railway automation and telemechanics systems: dis. ... candidate of technical sciences. Moscow: Russian University of Transport. 2020. 270 p. (In Russian)
2. Cherezov G.A. The current state of diagnostics of railway automation and telemechanics infrastructure facilities. Bulletin of transport of the Volga region. 2017. № 2(62). P. 62–66. (In Russian)
3. Kanarsky V.A., Khaliman V.V. Modern approaches to diagnostics of railway switches in the global practice. Transport of the Asia-Pacific region. 2023. № 4(37). P. 37–42. (In Russian)
4. Abdullaev R.B. Implementation of the subsystem of diagnostic information gathering in continuous monitoring systems of railway automation facilities based on programmable logic controllers. Automation in transport. 2020. V. 6. № 3. P. 309–331. DOI 10.20295/ 2412-9186-2020-6-3-309-331. (In Russian)
5. Belozerova I.G., Serova D.S. The general course of railways: Textbook. Khabarovsk: DVGUPS. 2020. 115 p. (In Russian)
6. Kanarsky V.A. Classification of railway switch states using discrete fourier transform. Automation, communications, informatics. 2023. № 12. P. 6–10. DOI 10.34649/AT.2023.12.12.002. (In Russian)
7. Buryak S.Yu., Gavrilyuk V.I., Gololobova O.A., Kovrigin M.A. Remote diagnostics of turnouts state on timing and spectral composition in current curve. Science and progress of transport. 2015. № 2(56). P. 39–57. (In Russian)
8. Belousov S.V. Methods and algorithms for troubleshooting arrows with AC electric motors: dis. ... candidate of technical sciences. St. Petersburg: PGUPS. 2019. 128 p. (In Russian)
9. Li M., Hei X., Ji W., Zhu L., Wang Y., Qiu Y. A Fault-Diagnosis Method for Railway Turnout Systems Based on Improved Autoencoder and Data Augmentation. Sensors. 2022. V. 22. № 23. P. 9438. DOI 10.3390/s22239438.
10. B¨ohm T. Remaining Useful Life Prediction for Railway Switch Engines Using Classification Techniques. International Journal of Prognostics and Health Management. 2017. V. 8. № 3. DOI 136001/ijphm.2017.v8i3.2666.
11. The device for monitoring and controlling switch drives (UMK SP): a brief user's guide. [Electronic resource] – Access mode: http://vast-arp.spb.ru/files/umksp_manual.pdf, date of application 27.01.2024. (In Russian)
12. Efanov D.V., Bogdanov N.A. Control of parameters of switch electric drives. Problems of safety and reliability of microprocessor complexes. St. Petersburg: St. Petersburg State University of Railways of Emperor Alexander I. 2015. P. 118–128. (In Russian)
13. Efanov D.V., Basalaev E.V., Alekseev V.G. Functional diagnostics of ac electric switch mechanisms. Transport of the Urals. 2012. № 4(35). P. 26–29. (In Russian)
14. Blagoveshchenskaya E.A., Bochkarev S.V., Gruzdev N.V., Zuev D.V. The use of self-learning systems to solve the problems of finding failures on the railway. Intelligent technologies in transport. 2020. № 1(21). P. 47–55. (In Russian)
15. Kanarsky V.A. Detection of the pre-failure state of the switch by the active power diagram. Bulletin of the Siberian State University of Railway Engineering. 2023. № 4(67). P. 40–46. DOI 10.52170/1815-9265_2023_67_40. (In Russian)
16. Kanarsky V.A. Usage of recurrent neural networks for diagnosis of temporal characteristics of a pointer drive. Materials of the II International Scientific and Practical Conference: Intelligent transport systems. Moscow: Russian University of Transport. 2023. P. 578–583. DOI 10.30932/9785002182794-2023-578-583. (In Russian)
17. Ivanova G.S., Martynyuk P.A. Analysis of neural network language models for solving problems of text data processing. Neurocomputers. 2023. V. 25. № 2. Р. 5–20. DOI 10.18127/j19998554-202302-01. (In Russian)
18. Introduction to Recurrent Neural Network. [Electronic resource] – Access mode: https://www.geeksforgeeks.org/introduction-to-recurrent-neural-network/, date of application 27.11.2023.
19. Pointer I. Programming PyTorch for Deep Learning: Creating and deploying deep learning applications. O'Reilly Media. 2019. 220 p.
20. Kingma D.P., Ba J. Adam: A Method for Stochastic Optimization. ArXiv:1412.6980. DOI 10.48550/arXiv.1412.6980.
21. Cho K., van Merriënboer B., Gulcehre C., Bahdanau D., Bougares F., Schwenk H., Bengio Y. Learning Phrase Representations Using RNN Encoder-Decoder for Statistical Machine Translation. Proceedings of the Conference on Empirical Methods in Natural Language Processing. Doha, Qatar. 2014. P. 1724–1734. DOI 10.3115/v1/d14-1179.
22. The site of Bolshaya Luga – Slyudyanka (Transsib). [Electronic resource] – Access mode: http://apkdk.ru/news/92-uchastok-bolshoy-lug-slyudyanka-transsib.html, date of application 27.11.2023.