Modeling of artificial intelligence system for early detection of emergency situations at vital facilities

The article presents the results of modeling an artificial intelligence system for early detection of undesirable situations of various types at objects of particular national economic importance. Pipeline transport or any other production system, in which continuous monitoring of operability parameters of critical components and mechanisms is carried out, can be specified as such object. This model can be applied by various oil and gas production companies. The results of modeling and subsequent development of the information system will provide the basis for industrial implementation of highly effective systems of accident detection and prevention in reliance on neural network analysis of continuously received streaming data. As a part of this research, the possibility of using modern neural network architectures for the problem under consideration is examined, namely, convolutional neural networks – TCN, direct propagation neural networks – MLP, recurrent neural networks – LSTM. It was proposed to abandon the activation function for LSTM which helps to provide the neural network with "long-term memory" of stored values, which is crucial to this problem. In addition, a cross-comparison of the error reduction rate during network training was performed to detect an architecture capable of "self-learning". All models were tested with the aid of the training data from the "Vostochny kupol" wells. Acceptable coincidence of test and extrapolation data was obtained for all models.

1. Pavlov S.G. Development of technology for assessing the geoecological safety of gas pipelines in emergency situations: dis. ... Candidate of Technical Sciences. M.: 2009. Available by: http://miigaik.ru/science/councils/congregation / 09_05_21_pavlov.pdf (accessed: 10/30/2021). (In Russ.)

2. Komashinsky V.I., Smirnov D.A. Neural networks and their application in control and communication systems. M.: Hotline – Telecom; 2002. 94 p. (In Russ.)

3. Podolsky A.K. Application of artificial intelligence methods in the oil and gas industry. Modern science. 2016;(3):33–36. (In Russ.)

4. The neural network will be able to assess the risks of accidents at oil industry enterprises in the Arctic. Available by: https://minobrnauki.gov.ru/press-center/news/?ELEMENT_ID=40904 (accessed: 11/15/2021). (In Russ.)

5. Malkina V.D., Kukarenko A.V. The OMEGA fiber-optic pipeline monitoring system with the use of artificial intelligence. Ecological Bulletin of Russia. 2020. Available by: https://omega.transneft.ru/press/articles/?id=36023 (accessed: 15.11.2021). (In Russ.)

6. Automatic control system for emergency emissions at chemically hazardous facilities (ASKAV). Available by: http://cit-avtomatika.ru/catalogue/avtomatizaciya/askav/ (date of reference: 01.12.2021). (In Russ.)

7. Vyalyshev A.I., Dobrov V.M., Dolgov A.A., Zinoviev S.V., Fayzulin T.S. Emergency control system at oil and gas production facilities in the waters of the Arctic seas. Technologies of civil safety. 2013;37(3):16–21. (In Russ.)

8. Multilayer perceptron (Multilayered perceptron). Available by: https://wiki.loginom.ru/articles/multilayered-perceptron.html (accessed: 11/18/2021). (In Russ.)

9. Long short-term memory. Available by: https://neerc.ifmo.ru/wiki/index.php?title=Long-term memory (accessed: 11/18/2021). (In Russ.)

10. The device regulating the executive UR. Available by: https://p-sr.ru/products/ustrojstva-zapornye-krany-sharovye/ustrojstva-reguliruyushhie-ur (accessed: 15.11.2021). (In Russ.)

11. Kataev K.A. Hydrate formation in natural gas pipelines. All-Russian Journal of Scientific publications. 2011;2(1). (In Russ.)

12. Dorman V.N., Baskakova N.T. Equipment condition analysis as a tool to reduce repair costs. Economic analysis: theory and practice. 2011;(39). (In Russ.)

13. Naumenko A.P. Theory and methods of monitoring and diagnostics: Materials of lectures. Omsk: OmSTU; 2017. 154 p. (In Russ.)

14. Dietterich T. Machine Learning for Sequential Data: A Review. Structural, Syntactic, and Statistical Pattern Recognition. 2002;15–30.

15. Fox M. Industrial Applications of Artificial Intelligence. Robotics. 1986;301–311.

16. Temporary convolutional networks and forecasting. Available by: https://ichi.pro/ru/vremennye-svertocnye-seti-i-prognozirovanie-102124506152164 (accessed: 11/18/2021).

17. A Method for Stochastic Optimization. Available by: https://arxiv.org/abs/1412.6980 (accessed: 04.12.2021).