E.O. Deryugina1, L.V. Tsarev2, A.A. Davidovich3

1-3 Kaluga Branch of Bauman Moscow State Technical University (Kaluga, Russia)

1 deryugina_eo@bmstu.ru, 2 tsarevlv@student.bmstu.ru, 3 davidovichaa@student.bmstu.ru

In modern industrial systems, reliable equipment operation is critical to minimizing downtime and reducing maintenance costs. Predictive analytics of equipment failures using neural networks enables early detection of potential faults. However, the effectiveness of such models is significantly limited by the lack of high-quality training data, which results in reduced forecasting accuracy.

The purpose of this research is to develop a methodology based on the application of neural networks and generative models for data synthesis, aimed at compensating for the lack of information and improving the accuracy of predictive equipment failure analytics.

Experimental results demonstrate that the proposed methodology enables the construction of more accurate and robust failure prediction models, capable of identifying various types of faults that may occur in industrial systems. This has direct practical relevance for reducing the risk of emergency situations, optimizing maintenance schedules, and improving the overall efficiency of manufacturing processes. The integration of such approaches can significantly reduce maintenance and repair costs and improve product quality. Moreover, the use of synthetic data to train models opens new horizons in machine learning and predictive analytics, allowing enterprises to adapt to rapidly changing market conditions and technological requirements. Thus, this study makes a significant contribution to the advancement of predictive maintenance technologies and enhances the competitiveness of industrial enterprises.

Deryugina E.O., Tsarev L.V., Davidovich A.A. Application of neural networks in predictive analytics of equipment failures: issue of data deficit and methods for their synthesis. Information-measuring and Control Systems. 2026. V. 24. № 2. P. 80−88. DOI: https://doi.org/10.18127/j20700814-202602-10 (in Russian)

1. Akimova A.A., Dadykin V.S. Prediktivnaya analitika kak metod povysheniya effektivnosti deyatelnosti predpriyatiya. Materialy V Respublikansko i nauchno-prakticheskoi konf. "Resursosberezhenie. Effektivnost. Razvitie".  Donetsk. 30 oktyabrya 2020. Otv. redaktor A.V. Yaroshenko. Donetsk: Donetskii natsionalnyi tekhnicheskii universitet. 2020. S. 9−14. EDN LZRERR. (in Russian)
2. Mustafina S.I., Zhilyakov S.A. Primenenie neironnykh setei dlya diagnostiki i prognozirovaniya tekhnicheskogo sostoyaniya oborudovaniya. Avtomatizatsiya i informatizatsiya TEK. 2023. № 7(600). S. 13−21. DOI 10.33285/2782-604X-2023-7(600)-13-21. EDN XPTCQK. (in Russian)
3. Bryukhova A.A., Kostyunina T.N. Problemy predektivnoi diagnostiki i analitiki. Izvestiya Tulskogo gosudarstvennogo universiteta. Tekhnicheskie nauki. 2023. № 7. S. 335−338. DOI 10.24412/2071-6168-2023-7-335-336. EDN NBMEMD. (in Russian)
4. Manakova E.I., Khachetlov B.A. Analiz generativnykh modelei iskusstvennogo intellekta: preimushchestva i perspektivy primeneniya. Sb. nauchnykh statei po materialam XII Mezhdunar. nauchno-prakticheskoi konf. "Aktualnye voprosy nauki i praktiki". Ufa. 20 iyunya 2023. Ufa: Obshchestvo s ogranichennoi otvetstvennostyu "Nauchno-izdatelskii tsentr "Vestnik nauki". 2023. Chast 2. S. 247−262. EDN HTRGBD. (in Russian)
5. Baghirov E. Building robust malware detection through conditional Generative Adversarial Network-based data augmentation. Program Systems: Theory and Applications. 2024. V. 15. № 4(63). P. 97−110. DOI 10.25209/2079-3316-2024-15-4-97-110. EDN DWLNAR.
6. Mazur A.E. Modelirovanie i podgonka vremennykh ryadov c tyazhelymi khvostami raspredelenii i silnoi vremennoi zavisimostyu posredstvom gaussovskikh ryadov: spetsialnost 01.01.05 "Teoriya veroyatnostei i matematicheskaya statistika". Dis … kand. fiz.-mat. nauk. 2021. 105 s. EDN QDMMFS. (in Russian)
7. Kolesnikov I.N., Finogeev A.G. Proaktivnyi monitoring sobytii na osnove prediktivnogo analiza vremennykh ryadov. Modeli, sistemy, seti v ekonomike, tekhnike, prirode i obshchestve. 2020. № 1(33). S. 111−125. DOI 10.21685/2227-8486-2020-1-9. EDN XGPJLA. (in Russian)
8. Electronic resource. Python. https://www.python.org/ (25.02.2025).
9. Electronic resource. Keras. https://keras.io (30.02.2025).
10. Electronic resource. TensorFlow. https://www.tensorflow.org/ (30.02.2025).
11. Electronic resource. Keras Tuner. https://keras-team.github.io/keras-tuner/ ( 15.02.2025).