M.A. Ludagovskaya1, E.D. Larionova2
1, 2 Russian University of Transport (MIIT) (Moscow, Russia)

1 m.ludagovskaya@gmail.com, 2 Eva.sh.2363@gmail.com

The field of application of digital technologies in railway transport is actively expanding, while the volume of diagnostic information collected by measuring devices and sensors for further transmission to various automated control systems of railway transport is increasing. For processing large amounts of data, one of the effective directions is to use neural network models and tools, including for solving problems of analyzing data from track measuring wagons and for predicting changes in the technical condition of a railway track. The article discusses the possibilities of using hybrid neural networks to solve forecasting problems in transport. The purpose of the article is to develop an approach to the construction and analysis of neural network models for predicting changes in the technical condition of a railway track, taking into account further integration with automated monitoring and diagnostics systems for the upper structure of the track. The article discusses the issues of data analysis and neural network modeling in the diagnosis of the technical condition of the railway track. The possibilities of using hybrid neural networks for assessing and predicting the condition of the upper structure of the railway track have been described. A hybrid neural network model has been considered, combining the capabilities of recurrent and convolutional neural networks. Neural network algorithms based on the LSTM model and based on a combination of CNN and LSTM models have been proposed. The issue of data preparation in the case of boundary values of parameters for deviations from the standards has been considered. The results can be used in solving the problems of creating and improving automated systems for managing the technical condition of transport systems, tasks of neural network modeling of technical systems, and in creating tools for monitoring transport infrastructure elements.

Ludagovskaya M.A., Larionova E.D. Application of hybrid neural networks to predict changes in the technical condition of a railway track. Neurocomputers. 2025. V. 27. № 2. P. 62–68. DOI: https://doi.org/10.18127/j19998554-202502-07 (in Russian)

1. Dulhare U.N., Bin Ahmad K.A., Ahmad K. Machine learning and big data: Concepts, algorithms, tools and applications. Wiley. 2020.
2. Hoseinzade E., Haratizadeh S. CNNpred: CNN-based stock market prediction using a diverse set of variables. Expert System with Applications. 2019. V. 129. P. 273–285.
3. Platonov E.N., Prosvirin K.V. Prognozirovanie defektov verkhnego stroeniya zheleznodorozhnogo puti metodami mashinnogo obucheniya. Vestnik komp'yuternykh i informatsionnykh tekhnologij. 2022. T. 19. № 2. S. 8–18. (in Russian)
4. Druzhinina O.V., Lyudagovskaya M.A. Analiz dannykh i nejrosetevoe modelirovanie v diagnostike tekhnicheskogo sostoyaniya zhelezno­dorozhnogo puti. Transport: nauka, tekhnika, upravlenie. 2022. № 6. S. 19–25. (in Russian)
5. Instruktsiya po tekushchemu soderzhaniyu zheleznodorozhnogo puti, utverzhdennaya rasporyazheniem OAO «RZhD» ot 14.11.2016 № 2288r. Tekhekspert [Elektronnyj resurs]. URL: https://docs.cntd.ru/document/456052478 (data obrashcheniya: 19.11.2024). (in Russian)
6. Lyudagovskaya M.A. Algoritmy analiza vremennykh ryadov v sistemakh diagnostiki i udalennogo monitoringa zheleznodorozhnogo puti. Sb. materialov Vseross. nauch.-praktich. konf. aspirantov i molodykh uchenykh, posvyashchennoj Dnyu aspiranta «Nauchnye issledovaniya i razrabotki molodykh uchenykh». Ul'yanovsk. 2024. S. 52–56. (in Russian)
7. Lyudagovskaya M.A. Vozmozhnosti primeneniya gibridnykh nejrosetej na zheleznodorozhnom transporte. Studencheskij vestnik: Aktual'nye voprosy nauki i obrazovaniya. Elets: Izd-vo EGU im. I.A. Bunina. 2023. S. 54–57. (in Russian)
8. Mahmoud A.A., Ammar M. Survey on deep learning for time-series forecasting. Machine Learning and Big Data Analytics Paradigms: Analysis, Applications and Challenges. 2021. P. 365–392.
9. Yang B., Sun S., Li J., Lin X., Tian Y. Traffic flow prediction using LSTM with feature enhancement. Neurocomputing. 2019. V. 332. P. 320–327.
10. Korol'kov E.P., Druzhinina O.V., Larionova E.D. Analiticheskoe i chislennoe issledovanie matematicheskoj modeli dlya kontrolya shiriny rel'sovoj kolei pri otstupleniyakh ot normativnykh parametrov. Transport: nauka, tekhnika, upravlenie. 2023. № 12. S. 32–37. DOI: 10.36535/0236-1914-2023-12-4. (in Russian)