Modeling of a mechanical device for well cleaning from asphalt-resin-paraffin deposits based on a multilayer perceptron neural network

UDK: 622.276.53.054.004.5

DOI: 10.24887/0028-2448-2025-9-73-79

Key words: asphalt resin paraffin deposits (ARPD), oil well, tubing, scraper, digital twin, multilayer perceptron

Authors: A.N. Krasnov(Ufa State Petroleum Technological University, RF, Ufa); S.N. Fedorov (Ufa State Petroleum Technological University, RF, Ufa); M.Yu. Prakhova (Ufa State Petroleum Technological University, RF, Ufa); D.V. Kalashnik (Ufa State Petroleum Technological University, RF, Ufa); D.V. Usikov (NEDRA LLC, RF, Saint Petersburg); V.D. Gulyaev (Research and Education Center Gazprom Neft – UGNTU, RF, Ufa); A.V. Ryzhikov (Association «Digital Technologies in Industry», RF, Saint Petersburg)

One of the challenges complicating the operation of oil production wells and oilfield equipment is the formation of asphalt-resin-paraffin deposits (ARPD). To combat ARPD, oil-producing companies employ both preventive measures and removal of already formed deposits. The primary method for cleaning the internal surface of tubing (tubing strings) from paraffin is mechanical, involving the lowering and lifting of scrapers using manual winches. This operation is performed in each well at regular intervals, determined by the well's production rate, ARPD content, temperature and pressure, and is set by the field's engineering service. The main drawback of such scrapers is their low mechanical reliability. Scrapers often cause emergency situations due to jamming inside the tubing or wire breakage. One possible way to improve the ARPD removal system is to implement a digital twin for diagnostics – a virtual analog of the tubing cleaning device. This enables real-time monitoring of the cleaning system's status and the anticipation of emergency situations, such as a potential winch cable break. This paper proposes a digital twin based on a multilayer perceptron neural network for diagnosing the operation of the well cleaning device. The studies showed that the most accurate model includes two hidden layers with 20 neurons in the first hidden layer and 10 in the second, trained using the Bayesian regularization algorithm. The recognition accuracy of possible emergency situations on the test sample reached no less than 85 %, which is a sufficient level for practical use of the model.

References

1. Krysanov D., II v neftyanoy industrii: kak tekhnologii menyayut geologorazvedku i dobychu (AI in the oil industry: How technology is changing exploration and production), URL: https://trends.rbc.ru/trends/industry/678f9f5f9a79477d9725513e

2. Blyablyas A.N., Application of artificial intelligence approach to the well complication control (In Russ.), Gazovaya promyshlennost', 2018, no. 4(767), pp. 16–21.

3. Belkina S.A., Nagaeva S.N., The education reasons the asfaltapitchesparaffin of deposits in PCP (In Russ.), Vestnik Yugorskogo gosudarstvennogo universiteta, 2016, V. 42, no. 3, pp. 7–11.

4. Korobov G.Yu., Parfenov D.V., Mechanisms of formation of asphaltene-resin-paraffin deposits: research methods (In Russ.), Neftegaz.RU, 2022, no. 8, pp. 22–31.

5. Gryaznova E.S., Main causes & factors affecting formation of asphalt-resin-paraffin deposits (In Russ.), Vestnik nauki, 2022, V. 1, no. 12(57), pp. 359–362.

6. Fedorov S.N., Kolovertnov G.Yu., Krasnov A.N., Prakhova M.Yu., Digital twins in the oil and gas industry: experience and prospects of use, Proceedings of the International Science Conference “Science. Education. Practice”, 2025, March 26, pp. 116–123, DOI: https://doi.org/10.34660/INF.2025.52.92.052

7. Krasnov A.N., Khoroshavina E.A., Prakhova M.Yu., Preventing paraffination of pumping equipment of oil wells, Advances in Engineering Research, 2017, V. 133,

pp. 370–375, DOI: https://doi.org/10.2991/aime-17.2017.60

8. Fedorov S.N., Kolovertnov G.Yu., Krasnov A.N., Prakhova M.Yu., Diagnostic diagram of state of well mechanical cleaning unit (In Russ.), Elektrotekhnicheskie i informatsionnye kompleksy i sistemy, 2025, V. 21, no. 2, pp. 98–110, DOI: http://doi.org/10.17122/1999-5458-2025-21-2-98-110.

9. Fedorov S.N., Kolovertnov G.Yu., Krasnov A.N., Prahova M.Yu., Digital twin of the device for cleaning pump-compressor pipes, HMMOCS, 2025, pp. 404–412,

DOI: https://doi.org/10.1007/978-3-031-95649-2_35

10. Kobzar O., Mosyagin G., Gudilov M. et al., A new approach to creating a digital twin of well for production monitoring in western Siberia fields, SPE-216731-MS, 2023, DOI: https://doi.org/10.2118/216731-MS

11. Andrianova A.M., Yudin E.V., Ganeev T.A. et al., Application of intelligent methods for analysis high-frequency production data for solving oil engineering challenges

(In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2021, no. 9, pp. 70–75, DOI: https://doi.org/10.24887/0028-2448-2021-9-70-75

12. Eremin N.A., Eremin A.N., Digital twin in the oil and gas production (In Russ.), Neft’. Gaz. Novatsii, 2018, no. 12, pp. 14–17.

13. Ilushin P., Vyatkin K., Menshikov A., Development of a methodology and software package for predicting the formation of organic deposits based on the results of laboratory studies, Fluids, 2021, no. 6, pp. 446-450, DOI: https://doi.org/10.3390/fluids6120446

14. Bykova V.N., Kim E., Gadzhialiev M.R. et al., Application of a digital twin in the oil and gas industry (In Russ.), Actual Problems of Oil and Gas, 2020, no. 1(28),

pp. 8–15, DOI: https://doi.org/10.29222/ipng.2078-5712.2020-28.art8

15. Yudin E., Kovaleva M., Shevchenko V. et al., Maintaining ESP operational efficiency through machine learning-based anomaly detection, Geoenergy Science and Engineering, 2025, V. 251, DOI: https://doi.org/10.1016/j.geoen.2025.213864

16. E. Yudin et al., Modelling of a gas-lift well operation with an automated gas-lift gas supply control system, SPE-196816-MS, 2019, DOI: https://doi.org/10.2118/196816-MS

17. Yudin E.V., Andrianova A.M., Ganeev T.A. et al., Production monitoring using a virtual flow meter for an unstable operating well stock (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2023, no. 8, pp. 82–87, DOI: https://doi.org/10.24887/0028-2448-2023-8-82-87

18. Yudin E. et al., Using deep learning algorithms to monitor well performance and restore well rate dynamics, SPE-217526-MS, 2023,

DOI: https://doi.org/10.2118/217526-MS

19. Neyronnaya set’: raskrytie vozmozhnostey iskusstvennogo intellekta (Neural Network: Unlocking the power of artificial intelligence),

URL: https://www.easiio.com/ru/neural-network-hidden-layer/

20. Chornyy A.V., Kozhemyakina I.A., Churanova N.Yu. et al., Analysis of wells interaction based on algorithms of complexing geological and field data (In Russ.), Neftyanoe khozyaystvo = Oil Industry, 2019, no. 1, pp. 36–39, DOI: https://doi.org/10.24887/0028-2448-2019-1-36-39

21. Nazhimova N.A., Naumova E.G., Formation of skills for building neural networks in various software environments (In Russ.), Sovremennye problemy nauki i obrazovaniya, 2024, no. 3, pp. 98–105, URL: https://doi.org/10.17513/spno.33522

22. De Luka G., Arkhitektura neyronnoy seti: kriterii vybora kolichestva i razmera skrytykh sloev (Neural network architecture: Criteria for selecting the number and size of hidden layers), URL: https://www.baeldung.com/cs/neural-networks-hidden-layers-criteria