Application of the direct propagation neural network for localization of the impact site of microparticles on the surface of the spacecraft

DOI: 10.34759/trd-2021-118-10

Аuthors

Voronov K. E.^*, Grigoriev D. P.^**, Telegin A. M.^***

Samara National Research University named after Academician S.P. Korolev, 34, Moskovskoye shosse, Samara, 443086, Russia

*e-mail: voronov.ke@ssau.ru
**e-mail: dan-22225@yandex.ru
***e-mail: talex85@mail.ru

Abstract

The purpose of this article is to demonstrate an experimental method for determining the impact region of microparticles in the surface of a spacecraft, through a neural network, with information about time delays in the data set. The article briefly describes the main types of neural network architectures that are widely used in various tasks. The theory of operation of the architecture of the neural network of direct propagation with mathematical explanations is given. The fundamental operations of neural network training, such as the method of error back propagation, gradient descent of the loss function, the training coefficient and its optimization, are also considered. The task of detecting the impact site of microparticles on the surface of the spacecraft body is set. As input data for training and testing the neural network, we used the results of an experiment to measure time delays on an experimental model with four piezo sensors. The output data was the numbers of the areas of the plate that were subjected to a simulated impact with a steel ball. The neural network model itself was written in the python programming language, using the Keras library and TensorFlow. This article also provides a detailed method for constructing a neural network model in python. The neural network obtained in the course of the study showed good results in terms of predicting the impact area of cosmic particles. The accuracy reached almost 90-100%. These results, as well as the advantages, disadvantages and prospects of the considered method, are given at the end of the article.

Keywords:

space debris, spacecraft, feed forward neural network, keras, tensorflow, impact localization

References

1. Barkova M.E. Trudy MAI, 2018, no. 103. URL: http://trudymai.ru/eng/published.php?ID=100712

2. Efimov E.N., Shevgunov T.Ya. Trudy MAI, 2015, no. 82. URL: http://trudymai.ru/eng/published.php?ID=58786

3. Voronov K.E., Grigor’ev D.P., Telegin A.M. Aviakosmicheskoe priborostroenie, 2021, no. 1, pp. 40 — 50. DOI: 10.25791/aviakosmos.1.2021.1200

4. Barskii A.B. Neironnye seti: raspoznavanie, upravlenie, prinyatie reshenii (Neural networks: recognition, management, decision-making), Moscow, Finansy i statistika, 2004, 176 p.

5. Rashid T. Sozdaem neironnuyu set’ (Creating a neural network), Saint Petersburg, Al’fa-kniga, 2017, 272 p.

6. Anikeev M.V., Babenko L.K., Makarevich O.B. Radioelektronika, informatika, upravlinnya, 2001, no. 1, pp. 48 — 56.

7. Akinina N.V., Akinin M.V., Sokolova A.V., Nikiforov M.B., Taganov A.I. Izvestiya TulGU. Tekhnicheskie nauki, 2016, no. 9, pp. 3 — 12.

8. Zemlevskii A.D. Vestnik nauki i obrazovaniya, 2017, vol. 2, no. 6 (30), pp. 36 — 43.

9. Bredikhin A.I. Vestnik Yugorskogo gosudarstvennogo universiteta, 2019, no. 1 (52), pp. 41 — 54. DOI: 10.17816/byusu20190141-54

10. Nikolenko S.I. Glubokoe obuchenie (Deep Learning), Saint Petersburg, Piter, 2018, 480 p.

11. Mean Square Error. URL: https://en.wikipedia.org/wiki/Mean_squared_error

12. Shumskikh I.Yu., Piganov M.N. Regional’naya nauchno-prakticheskaya konferentsiya, posvyashchennaya 50-letiyu pervogo poleta cheloveka v kosmos: tezisy dokladov, Samara, SGAU im. S.P. Koroleva, 2011, pp. 205 — 207.

13. A complex open-source machine learning platform. URL: https://www.tensorflow.org/

14. Keras: the python deep learning API. URL: https://keras.io/

15. Diederik Kingma, Jimmy Ba. Adam: A Method for Stochastic Optimization, Cornell University, 2017. URL: https://arxiv.org/abs/1412.6980

16. Lei Qi et al. Research on Leakage Location of Spacecraft in orbit Based on Frequency Weighting Matrix Beam forming Algorithm by Lamb Waves, Applied sciences, 2020, no. 10 (4), pp 1201. URL: https://doi.org/10.3390/app10041201

17. Ciampa F., Michele Meo. A new algorithm for acoustic emission localization and flexural group velocity determination in anisotropic structures, Journal Composites Part A Applied Science and Manufacturing, 2010, no. 41, pp. 1777 — 1786.

18. Tobias A. Acoustic-emission source location in two dimensions by an array of three sensors, Non-destructive testing, 1976, vol. 9, no. 1, pp. 9 — 12.

19. Ebrahimkhanlou A. A probabilistic framework for single-sensor acoustic emission source localization in thin metallic plates, Smart Materials and Structures, 2017, no. 26, pp. 1 — 19. DOI: 10.1088/1361-665X/aa78de

20. Ebrahimkhanlou A., Salamone S. Single-Sensor Acoustic Emission Source Localization in Plate-Like Structures Using Deep Learning, Aerospace, 2018, pp. 5 — 50. DOI: 10.3390/aerospace5020050

    
    Download