O.V. Druzhinina1, E.R. Korepanov2, A.A. Petrov3, I.V. Makarenkova4, V.V. Maksimova5

1,2,4,5 Federal Research Center «Computer Science and Control» of Russian Academy of Sciences (Moscow, Russia)

3 Bunin Yelets State University (Yelets, Russia)

1 ovdruzh@ipiran.ru, 2 ekorepanov@ipiran.ru, 3 xeal91@yandex.ru, 4 imakarenkova@ipiran.ru, 5 vmaksimova@mail.ru

The problems related to the development of intelligent instrumental and methodological support for diagnosing the condition of technical transport systems are important and relevant. Among such problems, it is necessary to highlight the problems of neural network, fuzzy and stochastic modeling, the solving of which is aimed at identifying malfunctions in conditions of incomplete information and data uncertainty. Methods and algorithms for solving these tasks should focus on the possible use of the domestic hardware and software environment, both in the learning process and in the process of application for data processing. The purpose of the article is to develop algorithms for neural network, fuzzy and stochastic modeling to identify malfunctions of elements and nodes of transport systems, as well as the implementation of modeling algorithms to solve specific problems of diagnosing the technical condition of axle boxes of railway wagons. The results are aimed at improving the efficiency of assessing the technical state of the axle boxes of railway wagons using data mining. The main modules of the software package have been developed to assess the technical condition of transport system nodes using stochastic and fuzzy algorithms for generating the initial data of a neural network model. The results of computer experiments aimed at stochastic modeling of temperature characteristics data and subsequent generation of initial data for solving classification problems in diagnosing the condition of axle boxes are presented. The implementation of a data generation model based on fuzzy modeling made it possible to create synthetic datasets for subsequent processing by neural network algorithms. The choice of neural network architecture for technical troubleshooting of axle box nodes has been made. The values of hyperparameters of the neural network are determined. The structure of the software package for assessing the technical condition of axle boxes is described. The modules of the software package are designed taking into account the adaptation and use of the capabilities of the domestic hardware and software platform Elbrus 801-RS. The results can be used in technical diagnostics, taking into account the incompleteness of information and the involvement of expert knowledge. It is possible to use the obtained results to solve practical diagnostic problems with a pre-trained specialized neural network, as well as to develop and improve intelligent monitoring systems and digital twin systems in transport.

1. Gorobchenko S.L., Shifrin B.M., Alekseeva S.V., Gogolevsky A.S., Krivonogova A.S., Pushkov Yu.L., Voynash S.A. Current status of application and development of artificial intelligence methods in industrial controllers and intelligent control systems. Proceedings of Tula State University. Technical sciences. 2023. № 3. P. 106–112. DOI 10.24412/2071-6168-2023-3-106-112. (In Russian)
2. Pushkarev D.O. Application of neural network expert systems for monitoring, diagnostics and forecasting of technical condition of aircraft engines. Information Technology Bulletin. 2020. № 3(25). P. 51–59. (In Russian)
3. Ryumin M.G. Application of fuzzy logic algorithms for predicting the pipe-culvert technical condition on railways. Russian Journal of Transport Engineering. 2023. V. 10. № 3. DOI 10.15862/12SATS323. (in Russian)
4. Orlova E.V. System engineering of digital twins of organizational and technical systems using methods of intellectual analysis. Software engineering. 2022. V. 13. № 9. P. 425–439. (in Russian)
5. Epikhin A.I., Kondratev S.I., Khekert E.V. Application of neural networks based on a multi-layer perceptron using fuzzy logic for technical diagnostics of ship technical equipme. Operation of marine transport. 2020. № 3(96). P. 111-119. DOI 10.34046/aumsu-omt96/15. (in Russian)
6. Dikarev P.V., Shilin A.A., Yudin S.Yu. System for recognition of emergency modes of overhead power lines using fuzzy logic. Energy and resource conservation: Industry and Transport. 2022. № 1(38). P. 6–12. (in Russian)
7. Kataev A.S., Akhatova C.F. Hybrid neuro fuzzy model of data mining for the formation of knowledge bases of soft expert diagnostic systems. Science and Education: scientific edition of Bauman Moscow State Technical University. 2012. № 12. P. 34. (in Russian)
8. Mironov A.A., Pavlyukov A.E., Saltykov D.N. Complex of computational models for the study of control processes of rolling stock units by infrared radiation. The world of dimensions. 2014. № 6. P. 21–27. (in Russian)
9. Mironov A.A., Pavlyukov A.E. Means of implementing an automated control and monitoring system for heating axle boxes. Control Engineering Russia. 2016. № 3(63). P. 53–59. (in Russian)
10. Belousov V.V., Druzhinina O.V., Korepanov E.R., Makarenkova I.V., Maksimova V.V. An approach to assessing the technical condition of elements and nodes of transport systems using neural network modeling methods and digital twin technology. Neurocomputers. 2021. V. 23. № 5. 2021. P. 5−20. DOI 10.18127/j19998554-202105-01. (In Russian)
11. Belousov V.V., Druzhinina O.V., Korepanov E.R., Makarenkova I.V., Maksimova V.V. The use of neural networks for solving classification problems in diagnosing malfunctions of transport systems. Neurocomputers. 2022. V. 24. № 4. Р. 18–27. DOI 10.18127/ j19998554-202204-02 (in Russian)
12. Druzhinina O.V., Korepanov E.R., Petrov A.A., Makarenkova I.V., Maksimova V.V. Design of a data generation model for solving classification problems in the diagnosis of transport system malfunctions. Nonlinear World. 2023. V. 21. № 3. P. 16–26. DOI 10.18127/ j20700970-202303-02. (in Russian)
13. Elbrus 801-RS workstation. [[Electronic resource] – Access mode: http://www.mcst.ru/elbrus_801-pc/, date of reference 21.12.2023. (In Russian)
14. Guide to effective programming on the Elbrus platform. [Electronic resource] – Access mode: http://www.mcst.ru/doc/el­brus_prog/elbrus-prog-1.2_2024-02-28.pdf, date of reference 21.12.2023. (In Russian)
15. Druzhinina O.V., Korepanov E.R., Belousov V.V., Masina O.N., Petrov A.A. Development of the instrumental support of the domestic computing platform "Elbrus 801-PC" in the problems of neural network modeling of nonlinear dynamic systems. Nonlinear World. 2021. V. 19. № 1. 2021. P. 15−28. DOI 10.18127/j20700970-202101-02. (In Russian)
16. Cherepov O.V., Panov V.A. Information technologies and systems for integrated control of the technical condition of wagons. Method. instructions. Yekaterinburg: UrGUPS. 2016. (In Russian)
17. Yarushkina N.G. Fundamentals of the theory of odd and hybrid systems. M.: Finance and Statistics. 2004. 320 p. ISBN 5-279-02776-6. (In Russian)