Гибридные нейросетевые модели в задаче мультиклассовой классификации данных телеметрической информации малых космических аппаратов
Ключевые слова:
гибридные нейросетевые модели, анализ данных, классификация, телеметрическая информация, полносвязные нейронные сети/слои, одномерные сверточные нейронные сети/слои, рекуррентные нейронные сети/слои
Аннотация
В статье представлены решения актуальной задачи интеллектуального анализа данных телеметрической информации (ТМИ) бортовой аппаратуры (БА) малых космических аппаратов (МКА) с целью определения их технических состояний. Исследованы и разработаны гибридные нейросетевые модели на основе современных архитектур глубокого обучения для решения задачи мультиклассовой классификации данных телеметрической информации, позволяющие определять штатное и нештатные состояния функционирования БА МКА. Предложенная гибридная нейросетевая модель представляет собой соединение трех блоков слоев: сверточного 1D CNN, рекуррентного GRU и итогового полносвязного блока-классификатора, с применением слоя агрегации AveragePooling и метода проброса остаточных связей семейства архитектур ResNet. Проведен компьютерный анализ на данных ТМИ БА МКА группировки АИСТ Самарского национального исследовательского университета имени академика С.П. Королёва., позволивший оценить качество разработанной модели на этапах обучения, валидации и тестирования. Для рассматриваемой задачи достигнута точность классификации на этапах валидации и тестирования более 98 %. Выполнен сравнительный анализ полученной гибридной нейросетевой модели (ГНМ) с широко распространенными глубокими нейросетевыми классификаторами, показавший преимущество полученного решения по точности классификации на этапе тестирования для всех сравниваемых моделей на 1–7 %. При этом по времени одной эпохи обучения и валидации почти для всех сравниваемых моделей получено улучшение выходных характеристик ГНМ в 1.5–4 раза. Результаты данного анализа подтвердили преимущество гибридной нейросетевой модели, реализующей методологию квалиметрии моделей, над «чистыми» сверточными, рекуррентными, полносвязными нейронными сетями и широко известными гибридными нейросетевыми моделями.
Скачивания
Данные скачивания пока не доступны.
Литература
1. Okhtilev M. Yu. [et al.] (2014) Koncepcija proaktivnogo upravlenija slozhnymi ob’ektami: teoreticheskie i tehnologicheskie osnovy [The concept of proactive management of complex objects: theoretical and technological foundations]. Journal of instrument engineering. 57(11). P. 7–14. (in Russian)
2. GOST RO 1410-002-2010. Raketno-kosmicheskaja tehnika. Sistema informacii o tehnicheskom sostojanii i nadezhnosti kosmicheskih kompleksov i vhodjashhih v ih sostav izdelij [Rocket and space technology. Information system on the technical condition and reliability of space complexes and their constituent products]. Introduced 2011-01-01. Moscow: Izd-vo standartov, 2011. 49 p. (in Russian)
3. Proekt strategii informacionnyh tehnologij Goskorporacii «Roskosmos» [Draft information technology strategy of the State Corporation Roscosmos.]. URL
4. Zelentsov V. A. [et al.] (2017) Service-oriented distributed software package for evaluating and multi-criteria analysis of indicators of reliability and survivability of onboard equipment of small spacecraft: Russian and Belarusian segments. In: Actual problems of rocket and space technology “(V Kozlov readings): Proceedings of V All-Russian scientific and technical conference with international participation (Samara, September 11–15, 2017). Samara, 2017, P. 45–56. (in Russian)
5. Zelentsov V. A. [et al.] (2018) Distributed software package for evaluating and analyzing indicators of reliability and survivability of onboard equipment of small spacecraft. In: Actual problems of creating space systems for remote sensing of the Earth: abstracts of the Sixth International Scientific and Technical Conference (Moscow, May 24, 2018). Moscow, 2018, 146 p. (in Russian)
6. Zelentsov V. A. [et al.] (2017) Servis-orientirovannyj raspredeljonnyj programmnyj kompleks dlja ocenivanija i mnogokriterial’nogo analiza pokazatelej nadjozhnosti i zhivuchesti bortovoj apparatury malyh kosmicheskih apparatov: rossijskij i belorusskij segment [Service-oriented distributed software package for evaluating and multi-criteria analysis of indicators of reliability and survivability of onboard equipment of small spacecraft: Russian and Belarusian segments]. Vestnik of Samara University. Aerospace and Mechanical Engineering. 16(4). P. 118–129. DOI
7. Skobtsov V. Yu. and Novoselova N. A. (2020) Issledovanie algoritmov potokovoj klasterizacii v reshenii zadachi analiza dannyh telemetrii malyh kosmicheskih apparatov [Investigation of stream clustering algorithms in solving the problem of analyzing telemetry data of small space vehicles]. Journal of instrument engineering. 63(11). P. 1003–1011. DOI
8. Skobtsov V. [et al.] (2017) Intelligent Telemetry Data Analysis of Small Satellites. In: Silhavy R., Senkerik R., Kominkova Oplatkova Z., Prokopova Z., Silhavy P. (eds) Cybernetics and Mathematics Applications in Intelligent Systems. CSOC 2017. Advances in Intelligent Systems and Computing – Springer International Publishing Switzerland. 574, P. 351–361. DOI
9. Skobtsov V. Yu. and Arkhipov V. I. (2021) Nejrosetevoj analiz dannyh telemetrii bortovoj apparatury kosmicheskih apparatov [Neural network analysis of telemetry data of spacecraft onboard equipment]. Space engineering and technology. 3(34). P. 111–124. (in Russian)
10. Skobtsov V. Yu. (2021) Binary classification of small satellites telemetry data based on deep learning approach. Applied Aspects of Information Technology. 4(4). P. 299–310.
11. Gamboa B. and Cristian J. Deep learning for time-series analysis. URL
12. Rajkomar A. [et al] (2018) Scalable and accurate deep learning for electronic health records. NPJ Digital Medicine. May, 1:18. URL DOI
13. Nweke H. F. (2018) Deep learning algorithms for human activity recognition using mobile and wearable sensor networks: state of the art and research challenges. Expert Systems with Applications. 105. P. 233–261. DOI
14. New T. L., Dat T. H. and Ma B. (217) Convolutional neural network with multi-task learning scheme for acoustic scene classification. Annual Summit and Conference 2017: proceedings of APSIPA, P. 1347–1350. DOI
15. Susto G. A. [et al.] (2018) Time-series classification methods: review and applications to power systems data. Big data application in power systems, Elsevier Inc., Chapter 9. P. 179–220. DOI
16. Yang, Q. and Wu X. (2006) 10 challenging problems in data mining research. International Journal of Information Technology & Decision Making. 05(04). P. 597–604. DOI
17. Fawaz H. I. [et al.] (2019) Muller Deep learning for time series classification: a review. Data Mining and Knowledge Discovery. 33. P. 917–963. DOI
18. Silva D. F. [et al.] (2018) Speeding up similarity search under dynamic time warping by pruning unpromising alignments. Data Mining and Knowledge Discovery. 32(4). P. 988–1016. DOI
19. Bagnall A. [et al.] (2017) The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances. Data Mining and Knowledge Discovery. 31(3). P. 606–660. DOI
20. Bagnall A. [et al.] (2016) Time-series classification with COTE: the collective of transformation-based ensembles. 2016 IEEE 32nd International Conference on Data Engineering (ICDE): proceedings (May 2016). P. 1548–1549. DOI
21. Hills J. [et al.] (2014) Classification of time series by shapelet transformation. Data Mining and Knowledge Discovery. 28(4). P. 851–881. DOI
22. Bostrom A. and Bagnall A. (2015) Binary shapelet transform for multiclass time series classification. Conference on Big Data Analytics and Knowledge Discovery: proceedings of the International Conference. P. 257–269. DOI
23. Baydogan M. G. Multivariate time series classification datasets. – URL
24. Lines J., Taylor S. and Bagnall A. (2016) HIVE-COTE: the hierarchical vote collective of transformation-based ensembles for time series classification. 2016 IEEE 16th International Conference on Data Mining (ICDM): proceedings of the international conference (December 2016). P. 1041–1046. DOI
25. Lines J., Taylor S. and Bagnall A. (2018) Time series classification with HIVE-COTE: the hierarchical vote collective of transformation-based ensembles. ACM Trans Knowl Discov Data. 2(5), Article No 52. P. 1–35. DOI
26. LeCun Y., Bengio Y. and Hinton G. (2015) Deep learning. Nature. 521. P. 436–444. DOI
27. Wang Z., Yan W. and Oates T. (2017) Time series classification from scratch with deep neural networks: a strong baseline. 2017 International Joint Conference on Neural Networks (IJCNN): proceedings of the international joint conference (May 2017). P. 1578–1585. DOI
28. Gorban’ A. N., Dunin-Barkovskij V. L., Kirdin A. N. [et al.] (1998) Nejroinformatika [Neuroinformatics]. Novosibirsk, Science. (in Russian)
29. Tsaregorodsev V. G. (1998) Neurosimulator NEUROPRO. Neuroinformatics and its applications: abstract. report VI All-Russian Seminar (October 2–5, 1998) / ed. Gorban’ A. N. Krasnoyarsk, 207 p. (in Russian)
30. Nikolenko S., Kadurin A. and Arhangel’skaja E. (2018) Glubokoe obuchenie [Deep Learning]. Saint-Petersburg, Piter. (in Russian)
31. Sholle F. (2018) Glubokoe obuchenie na Python [Deep Learning in Python]. Saint-Petersburg, Piter. (in Russian)
32. Cho Kyunghyun. (2014) Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. Empirical Methods in Natural Language Processing (EMNLP): proceedings of the 2014 Conference. arXiv:1406.1078. DOI
33. Lin Tao, Guo Tian and Aberer Karl. (2017) Hybrid Neural Networks for Learning the Trend in Time Series. Artificial Intelligence (IJCAI-17): proceedings of the Twenty-Sixth International Joint Conference. P. 2273–2279. DOI
34. Chen Haoze and Zhang Zhijie. (2020) Hybrid neural network based on novel audio feature for vehicle type identification. 2020 IEEE International Instrumentation and Measurement Technology Conference (I2MTC): proceedings of the international conference (25–28 May 2020). DOI
35. Zhao N. (2021) Combination of Convolutional Neural Network and Gated Recurrent Unit for Aspect-Based Sentiment Analysis. EEE Access. 9. P. 15561–15569. DOI
36. He Kaiming [et al.] (2016) Deep Residual Learning for Image Recognition. Conference on Computer Vision and Pattern Recognition: proceedings of the 2016 IEEE Conference. URL DOI
37. Volgin S. S. (2019) Rezul’taty obrabotki dannyh telemetricheskih izmerenij, postupajushhih ot gruppirovki malyh kosmicheskih apparatov «AIST» [The results of processing data from telemetry measurements coming from the grouping of small spacecraft “AIST”]. Kosmonavtika i raketostroenie. (1). P. 80–91. (in Russian)
38. Repository of online seminars «Machine Learning Tokyo – Democratizing Machine Learning». URL
39. Mikoni S. V., Sokolov B. V. and Yusupov R. M. (2018). Kvalimetrya modeley i polymodelnyx komplexov. Moscow : RAN. 314 p.
2. GOST RO 1410-002-2010. Raketno-kosmicheskaja tehnika. Sistema informacii o tehnicheskom sostojanii i nadezhnosti kosmicheskih kompleksov i vhodjashhih v ih sostav izdelij [Rocket and space technology. Information system on the technical condition and reliability of space complexes and their constituent products]. Introduced 2011-01-01. Moscow: Izd-vo standartov, 2011. 49 p. (in Russian)
3. Proekt strategii informacionnyh tehnologij Goskorporacii «Roskosmos» [Draft information technology strategy of the State Corporation Roscosmos.]. URL
4. Zelentsov V. A. [et al.] (2017) Service-oriented distributed software package for evaluating and multi-criteria analysis of indicators of reliability and survivability of onboard equipment of small spacecraft: Russian and Belarusian segments. In: Actual problems of rocket and space technology “(V Kozlov readings): Proceedings of V All-Russian scientific and technical conference with international participation (Samara, September 11–15, 2017). Samara, 2017, P. 45–56. (in Russian)
5. Zelentsov V. A. [et al.] (2018) Distributed software package for evaluating and analyzing indicators of reliability and survivability of onboard equipment of small spacecraft. In: Actual problems of creating space systems for remote sensing of the Earth: abstracts of the Sixth International Scientific and Technical Conference (Moscow, May 24, 2018). Moscow, 2018, 146 p. (in Russian)
6. Zelentsov V. A. [et al.] (2017) Servis-orientirovannyj raspredeljonnyj programmnyj kompleks dlja ocenivanija i mnogokriterial’nogo analiza pokazatelej nadjozhnosti i zhivuchesti bortovoj apparatury malyh kosmicheskih apparatov: rossijskij i belorusskij segment [Service-oriented distributed software package for evaluating and multi-criteria analysis of indicators of reliability and survivability of onboard equipment of small spacecraft: Russian and Belarusian segments]. Vestnik of Samara University. Aerospace and Mechanical Engineering. 16(4). P. 118–129. DOI
7. Skobtsov V. Yu. and Novoselova N. A. (2020) Issledovanie algoritmov potokovoj klasterizacii v reshenii zadachi analiza dannyh telemetrii malyh kosmicheskih apparatov [Investigation of stream clustering algorithms in solving the problem of analyzing telemetry data of small space vehicles]. Journal of instrument engineering. 63(11). P. 1003–1011. DOI
8. Skobtsov V. [et al.] (2017) Intelligent Telemetry Data Analysis of Small Satellites. In: Silhavy R., Senkerik R., Kominkova Oplatkova Z., Prokopova Z., Silhavy P. (eds) Cybernetics and Mathematics Applications in Intelligent Systems. CSOC 2017. Advances in Intelligent Systems and Computing – Springer International Publishing Switzerland. 574, P. 351–361. DOI
9. Skobtsov V. Yu. and Arkhipov V. I. (2021) Nejrosetevoj analiz dannyh telemetrii bortovoj apparatury kosmicheskih apparatov [Neural network analysis of telemetry data of spacecraft onboard equipment]. Space engineering and technology. 3(34). P. 111–124. (in Russian)
10. Skobtsov V. Yu. (2021) Binary classification of small satellites telemetry data based on deep learning approach. Applied Aspects of Information Technology. 4(4). P. 299–310.
11. Gamboa B. and Cristian J. Deep learning for time-series analysis. URL
12. Rajkomar A. [et al] (2018) Scalable and accurate deep learning for electronic health records. NPJ Digital Medicine. May, 1:18. URL DOI
13. Nweke H. F. (2018) Deep learning algorithms for human activity recognition using mobile and wearable sensor networks: state of the art and research challenges. Expert Systems with Applications. 105. P. 233–261. DOI
14. New T. L., Dat T. H. and Ma B. (217) Convolutional neural network with multi-task learning scheme for acoustic scene classification. Annual Summit and Conference 2017: proceedings of APSIPA, P. 1347–1350. DOI
15. Susto G. A. [et al.] (2018) Time-series classification methods: review and applications to power systems data. Big data application in power systems, Elsevier Inc., Chapter 9. P. 179–220. DOI
16. Yang, Q. and Wu X. (2006) 10 challenging problems in data mining research. International Journal of Information Technology & Decision Making. 05(04). P. 597–604. DOI
17. Fawaz H. I. [et al.] (2019) Muller Deep learning for time series classification: a review. Data Mining and Knowledge Discovery. 33. P. 917–963. DOI
18. Silva D. F. [et al.] (2018) Speeding up similarity search under dynamic time warping by pruning unpromising alignments. Data Mining and Knowledge Discovery. 32(4). P. 988–1016. DOI
19. Bagnall A. [et al.] (2017) The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances. Data Mining and Knowledge Discovery. 31(3). P. 606–660. DOI
20. Bagnall A. [et al.] (2016) Time-series classification with COTE: the collective of transformation-based ensembles. 2016 IEEE 32nd International Conference on Data Engineering (ICDE): proceedings (May 2016). P. 1548–1549. DOI
21. Hills J. [et al.] (2014) Classification of time series by shapelet transformation. Data Mining and Knowledge Discovery. 28(4). P. 851–881. DOI
22. Bostrom A. and Bagnall A. (2015) Binary shapelet transform for multiclass time series classification. Conference on Big Data Analytics and Knowledge Discovery: proceedings of the International Conference. P. 257–269. DOI
23. Baydogan M. G. Multivariate time series classification datasets. – URL
24. Lines J., Taylor S. and Bagnall A. (2016) HIVE-COTE: the hierarchical vote collective of transformation-based ensembles for time series classification. 2016 IEEE 16th International Conference on Data Mining (ICDM): proceedings of the international conference (December 2016). P. 1041–1046. DOI
25. Lines J., Taylor S. and Bagnall A. (2018) Time series classification with HIVE-COTE: the hierarchical vote collective of transformation-based ensembles. ACM Trans Knowl Discov Data. 2(5), Article No 52. P. 1–35. DOI
26. LeCun Y., Bengio Y. and Hinton G. (2015) Deep learning. Nature. 521. P. 436–444. DOI
27. Wang Z., Yan W. and Oates T. (2017) Time series classification from scratch with deep neural networks: a strong baseline. 2017 International Joint Conference on Neural Networks (IJCNN): proceedings of the international joint conference (May 2017). P. 1578–1585. DOI
28. Gorban’ A. N., Dunin-Barkovskij V. L., Kirdin A. N. [et al.] (1998) Nejroinformatika [Neuroinformatics]. Novosibirsk, Science. (in Russian)
29. Tsaregorodsev V. G. (1998) Neurosimulator NEUROPRO. Neuroinformatics and its applications: abstract. report VI All-Russian Seminar (October 2–5, 1998) / ed. Gorban’ A. N. Krasnoyarsk, 207 p. (in Russian)
30. Nikolenko S., Kadurin A. and Arhangel’skaja E. (2018) Glubokoe obuchenie [Deep Learning]. Saint-Petersburg, Piter. (in Russian)
31. Sholle F. (2018) Glubokoe obuchenie na Python [Deep Learning in Python]. Saint-Petersburg, Piter. (in Russian)
32. Cho Kyunghyun. (2014) Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation. Empirical Methods in Natural Language Processing (EMNLP): proceedings of the 2014 Conference. arXiv:1406.1078. DOI
33. Lin Tao, Guo Tian and Aberer Karl. (2017) Hybrid Neural Networks for Learning the Trend in Time Series. Artificial Intelligence (IJCAI-17): proceedings of the Twenty-Sixth International Joint Conference. P. 2273–2279. DOI
34. Chen Haoze and Zhang Zhijie. (2020) Hybrid neural network based on novel audio feature for vehicle type identification. 2020 IEEE International Instrumentation and Measurement Technology Conference (I2MTC): proceedings of the international conference (25–28 May 2020). DOI
35. Zhao N. (2021) Combination of Convolutional Neural Network and Gated Recurrent Unit for Aspect-Based Sentiment Analysis. EEE Access. 9. P. 15561–15569. DOI
36. He Kaiming [et al.] (2016) Deep Residual Learning for Image Recognition. Conference on Computer Vision and Pattern Recognition: proceedings of the 2016 IEEE Conference. URL DOI
37. Volgin S. S. (2019) Rezul’taty obrabotki dannyh telemetricheskih izmerenij, postupajushhih ot gruppirovki malyh kosmicheskih apparatov «AIST» [The results of processing data from telemetry measurements coming from the grouping of small spacecraft “AIST”]. Kosmonavtika i raketostroenie. (1). P. 80–91. (in Russian)
38. Repository of online seminars «Machine Learning Tokyo – Democratizing Machine Learning». URL
39. Mikoni S. V., Sokolov B. V. and Yusupov R. M. (2018). Kvalimetrya modeley i polymodelnyx komplexov. Moscow : RAN. 314 p.
Опубликован
2022-11-09
Как цитировать
Скобцов, В. Ю., & Соколов, Б. В. (2022). Гибридные нейросетевые модели в задаче мультиклассовой классификации данных телеметрической информации малых космических аппаратов. Вестник ВГУ. Серия: Системный анализ и информационные технологии, (3), 99-114. https://doi.org/10.17308/sait/1995-5499/2022/3/99-114
Выпуск
Раздел
Интеллектуальные системы, анализ данных и машинное обучение
- Авторы сохраняют за собой авторские права и предоставляют журналу право первой публикации работы, которая по истечении 6 месяцев после публикации автоматически лицензируется на условиях Creative Commons Attribution License , которая позволяет другим распространять данную работу с обязательным сохранением ссылок на авторов оригинальной работы и оригинальную публикацию в этом журнале.
- Авторы имеют право размещать их работу в сети Интернет (например в институтском хранилище или персональном сайте) до и во время процесса рассмотрения ее данным журналом, так как это может привести к продуктивному обсуждению и большему количеству ссылок на данную работу (См. The Effect of Open Access).
