«Automation and Informatization of the fuel and energy complex»

Development of a predictive model for defects occurence in gas pipelines using deep learning algorithms

UDC: 620.19:621.43
DOI: -

Authors:

¹ Gazprom Diagnostics, St. Petersburg, Russia
² National University of Oil and Gas "Gubkin University", Moscow, Russia

Keywords: defects, in-pipe flaw detection, neural networks, fully connected networks

Annotation:

The authors of the article consider the building and practical use of the neural network model for solving the actual problem of predicting stress-corrosion defects occurrence in high pressure pipelines. The information obtained in the course of a long-lasing survey of the gas transportation system and kept in the corporate base of "Infotech" data base was chosen as a target indicator for teaching the model. The data from 17 sections of different gas pipelines the total length of which is more than 1000 km were chosen for the model teaching. The probability of the model real prediction after teaching made 82,3 %. The common principals of the model operation are described, the process of its building and teaching as well as methods of the data processing and normalizing are described. The results were analyzed and the prospects for improving the work with data in order of raising the prediction results stability as well as widening the model possibilities were outlined. It is noted that probabilistic statistical and neural network algorithms of predicting defects for increasing the reliability of the technical state of the oil and gas sector equipment, significantly decreasing the labor intensity of field works. The prospect for this direction is the synthesis of an algorithm for predicting the technical state of equipment, based on the accumulated array of information, simultaneously using both robust methods and deep learning methods.

Bibliography:

1. Aksyutin O.E. Innovatsii – neobkhodimost' ili dan' mode? // Nauka i tekhnika v gazovoy prom-sti. – 2022. – № 4(92). – S. 3–14.
2. Zavyalov A.P., Zhuchkov K.N., Vasilchenko M. Process Pipeline Strength Calculation Methodology Enhancement Using Finite-Element Method // J. of Pipeline Systems Engineering and Practice. – 2023. – Vol. 14, No. 2. – Article No. 1401. – DOI: 10.1061/JPSEA2.PSENG-1401
3. Gus'kov M.A., Fomina E.E., Mogilo A.A. Analiz prichin avariy na opasnykh proizvodstvennykh ob"ektakh magistral'nogo truboprovodnogo transporta gaza // Zashchita okruzhayushchey sredy v neftegazovom komplekse. – 2024. – № 4(319). – S. 62–69.
4. Mustafina S.I., Zhilyakov S.A. Primenenie neyronnykh setey 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
5. Zav'yalov A.P., Dyachenko I.F. Ob izmenenii trebovaniy k tekhnicheskoy diagnostike dlya vnedreniya metodov prediktivnoy analitiki pri perekhode k risk-orientirovannoy sisteme ekspluatatsii oborudovaniya toplivno-energeticheskogo kompleksa // Oborudovanie i tekhnologii dlya neftegazovogo kompleksa. – 2025. – № 1(145). – S. 68–71.
6. Vasil'ev G.G., Butovka A.N. Issledovanie protsessov stareniya svoystv metalla nadzemnykh magistral'nykh gazoprovodov // Oborudovanie i tekhnologii dlya neftegazovogo kompleksa. – 2025. – № 1(145). – S. 34–41.
7. The use of machine learning for various predictive models of the occurrence of pipe defects / K. Zhuchkov, A. Zavyalov, A. Lopatin [et al.] // Insight – Civil Engineering. – 2024. – Vol. 7, No. 2. – P. 646. – DOI: 10.18282/ice.v7i2.646
8. Postroenie prognoznykh modeley na osnove iskusstvennogo intellekta dlya resheniya zadachi predskazaniya poyavleniya defektov truboprovodov / A.V. Shibanov, D.S. Pochikeev, F.A. Kochubey [i dr.] // Avtomatizatsiya i informatizatsiya TEK. – 2023. – № 10(603). – S. 38–47. – DOI: 10.33285/2782-604X-2023-10(603)-38-47
9. Kere K.J., Huang Qindan. Development of a Probabilistic Interaction Rule and Failure Pressure Model for Pipelines with a Colony of Corrosion Defects // J. of Pipeline Systems Engineering and Practice. – 2025. – Vol. 16, No. 2. – Article No. 1714. – DOI: 10.1061/jpsea2.pseng-1714
10. Bazin N.R. Issledovanie effektivnosti mnogosloynoy polnosvyaznoy neyronnoy seti v zadachakh regressii // Reshetnevskie chteniya: materialy XXVIII Mezhdunar. nauch.-prakt. konf., posvyashch. 100-letiyu so Dnya rozhdeniya gener. konstruktora raketno-kosmicheskikh sistem akad. Mikhaila Fedorovicha Reshetneva, Krasnoyarsk, 18–22 noyab. 2024 g.: v 2 ch. Ch. 2. – Krasnoyarsk: Sib. gos. un-t nauki i tekhnologiy im. akad. M.F. Reshetneva, 2024. – S. 40–42.
11. Terekhov V.I., Chernen'kiy I.M., Minakova S.V. Issledovanie i razrabotka metodov nastroyki chisla neyronov v skrytom sloe polnosvyaznoy neyronnoy seti // Neyrokomp'yutery: razrabotka, primenenie. – 2017. – № 3. – S. 52–62.
12. Anan'eva A.A., Lantsov V.N. Osobennosti bibliotek programmnogo kompleksa Keras dlya sozdaniya iskusstvennykh neyronnykh setey // Prikladnaya matematika i informatika: sovremennye issled. v oblasti estestvennykh i tekhnicheskikh nauk: sb. materialov IX Mezhdunar. nauch.-prakt. konf. (shk.-seminara) molodykh uchenykh, Tol'yatti, 18–20 apr. 2023 g. – Tol'yatti: Tol'yattinskiy gos. un-t, 2023. – S. 386–390.
13. Vasilchenko M., Zavyalov A.P., Zhuchkov K.N. Increasing the Stability of a Spatially Distributed Information System using a Robust Algorithm for Filtering Anomalous Measurements // Information Technology in Industry. – 2020. – Vol. 8, No. 3. – P. 1–7.
14. Identifying flow defects in amorphous alloys using machine learning outlier detection methods / Tian Liang, Fan Yue, Li Lin, N. Mousseau // Scripta Materialia. – 2020. – Vol. 186. – P. 185–189. – DOI: 10.1016/j.scriptamat.2020.05.038
15. A Feature Selection–Based Intelligent Framework for Predicting Maximum Depth of Corroded Pipeline Defects / Lu Hongfang, Peng Haoyan, Xu Zhao-Dong [et al.] // J. of Performance of Constructed Facilities. – 2022. – Vol. 36, No. 5. – DOI: 10.1061/(asce)cf.1943-5509.0001753