Artificial neural networks application for experimental data analysis of composite solid propellants combustion

DOI: 10.34759/vst-2020-1-164-170

Аuthors

Aver’kov I. S.^1*, Vlasov S. O.^1**, Raznoschikov V. V.^2***

1. Central Institute of Aviation Motors, CIAM, 2, Aviamotornaya St., Moscow, 111116, Russia
2. Central Institute of Aviation Motors named after P.I. Baranov, CIAM, 2, Aviamotornaya str., Moscow, 111116, Russia

*e-mail: averkov@ciam.ru
**e-mail: sergeyvlasovsaylar@yandex.ru
***e-mail: raznoschikov@ciam.ru

Abstract

While studying and solving the problems associated with a ramjet mathematical model developing, situations occur when a process model contains a complex mathematical formulation or a large number of assumptions. A number of experimental studies is being conducted in such cases, based on which corrections are being introduced to the model to increase accuracy of the obtained results.

The presented article regards the process of creating an electronic database of experimental studies on determination of the multicomponent combined solid propellant combustion rate, with their subsequent processing and analyzing with artificial neural networks. For gas generator and propellant consumption regulator of a ramjet operation modelling, information on combustion rate of a solid propellant is required.

Mass fractions of solid propellant components are included in the alterable variables vector. It is unreasonable to conduct experiments for all analyzed propellant compositions due to the complexity, expensiveness and long duration of their implementation. The authors suggest conducting experimental studies of particular compositions in the area under study and performing approximation by the obtained points. As the result, a function, reflecting the combustion rate behavior in dependence of the solid propellant composition and pressure is obtained.

There is a three-component propellant being a mixture of C₆H₂N₈O₄, ammonium perchlorate NH₄ClO₄ and a binder (rubber). The predicted parameter is the burning rate at various compositions and pressures.

The obtained topologies are built based on experimental research, and can be used later in formation of appearances of new ramjet engines.

When processing the obtained results, it is necessary to account for the fact that all experiments have certain error. The surfaces, obtained by neural networks allow identify the points at which random errors could reach high values, which is become noticeable by the function behavior.

1. Experimental data processing using neural networks allows forming a matrix of combustion rates database in specified intervals of alterable variables.

2. The burning rate topology analysis give grounds for analyzing the results obtained during the experiments, and, thus, to determine the experiments in which errors could be made.

Keywords:

approximation, fuel combustion, artificial neural network, experimental data processing, processes modeling

References

1. Krylov B.A., Onishchik I.I., Yun A.A. A simulation of mass and heat exchange within model combustion chambers. Aerospace MAI Journal, 2009, vol. 16, no. 1, pp. 27-30.

2. Bogdanov V.I. Research on realization of pulsating working processes in jet engines. Aerospace MAI Journal, 2017, vol. 24, no. 4, pp. 100-109.

3. Ryabov A.A., Romanov V.I., Kukanov S.S., Shmotin Yu.N., Gabov D.V. Numerical and experimental criterion of gas turbine engine hull dynamic strength in case of open rotor blade out. Aerospace MAI Journal, 2015, vol. 22, no. 3, pp. 76-84.

4. Dorofeev E.A., Dynnikov A.I., Kargopol’tsev A.V., Sviridenko Yu.N., Fadeev A.S. Uchenye zapiski TsAGI, 2007, vol. XXXVIII, no. 3-4, pp. 111-118.

5. Filatova T.V. Vestnik Tomskogo gosudarstvennogo universiteta, 2004, no. 284, pp. 121-125.

6. Reizlin V.I. Chislennye metody optimizatsii (Numerical optimization methods), Tomsk, Tomskii politekhnicheskii universitet, 2011, 105 p.

7. Matyushchenko N.S., Kopyrin A.S. Izvestiya Sochinskogo gosudarstvennogo universiteta, 2012, no. 3(21), pp. 51-62.

8. Aravin O.I. Rossiiskii zhurnal biomekhaniki, 2011, vol. 15, no. 3(53), pp. 45-51.

9. Abashev O.V., Kuprikov M.Yu. An application of artificial neural networks in aircraft design. Aerospace MAI Journal, 2008, vol. 15, no. 5, pp. 27-33.

10. Ignatyev D.I. Application of artificial neural networks for simulation of delta wing aerodynamic characteristics. Aerospace MAI Journal, 2010, vol. 17, no. 6, pp. 5-12.

11. Brusov V.S., Tiumentsev Yu.V. A synthesis of optimal neurocontroller ensemble for multiple-regime aircraft. Aerospace MAI Journal, 2006, vol. 13, no. 2, pp. 67-78.

12. Blinov A.O., Fralenko V.P. Multidimensional approximation for modeling and optimization problems. Automation and Remote Control, 2009, vol. 70, no. 4, pp. 652-662. DOI: 10.1134/ S0005117909040110

13. Korobkova S.V. Izvestiya TRTU, 2006, no. 3(58), pp. 121-127.

14. Borovikov V.P. Neironnye seti Statistica Neural Networks. Metodologiya i tekhnologii sovremennogo analiza dannykh (Neural networks. STATISTICA Neural Networks. Methodology and technologies of modern data analysis), Moscow, Telekom, 2008, 392 p.

15. Rudoi G.I. Mashinnoe obuchenie i analiz dannykh, 2011, vol. 1, no. 1, pp. 16-39.

16. Bishop C.M. Neural Networks for Pattern Recognition. Oxford University Press, USA, 1995, 502 p.

17. Gulakov K.V. Vestnik Bryanskogo gosudarstvennogo tekhnicheskogo universiteta, 2013, no. 2(38), pp. 95-105.

18. Belyaev M.G., Prikhod’ko P.V., Burnaev E.V., Bernshtein A.V. Materialy konferentsii molodykh uchenykh (Sankt–Peterburg, 14–17 April 2009) “Matematicheskoe modelirovanie i programmnoe obespechenie”, Moscow, ITMO, 2009, no. 4, pp. 46-51.

19. Kalatskaya L.V., Novikov V.A., Sadkov V.S. Organizatsiya i obuchenie iskusstvennykh neironnykh setei (Organization and training of artificial neural networks), Minsk, BGU, 2002, 76 p.

20. Sorokin V.A. (red.) Proektirovanie i otrabotka raketno– pryamotochnykh dvigatelei na tverdom toplive (Design and development of rocket-ramjet engines on solid fuel), Moscow, MGTU im. N.E. Baumana, 2016, pp. 9-63.

21. Haykin S.O. Neural Networks: A Comprehensive Foundation. Second Edition. Prentice Hall, New Jersey, 1999, 842 p.