creexpCrede Experto: транспорт, общество, образование, языкCrede Experto: transport, society, education, language2312-1327Иркутский филиал ФГБОУ ВО «МГТУ ГА»10.51955/2312-1327_2025_3_71creexp-70Research ArticleЛЕТАТЕЛЬНЫЕ АППАРАТЫ, АВИАЦИОННЫЕ ДВИГАТЕЛИ И МЕТОДЫ ИХ ЭКСПЛУАТАЦИИК вопросу экспериментальной постановки апробации алгоритма обработки диагностических параметров авиационных ГТД на основе многослойных нейронных сетейOn the experimental setup for approbation of an algorithm for processing diagnostic parameters of aircraft Gas Turbine Engine based on multilayer neural networkshttps://orcid.org/0009-0002-9280-6361ГусейновГ.HuseynovH.

Гусейн Гусейнов, аспирант

Кронштадтский бульвар, д. 20, Москва, 125493

Huseyn Huseynov, Postgraduate student

20, Kronshtadtsky blvd, Moscow, 125493

khuseyn.21@gmail.comhttps://orcid.org/0009-0004-8099-5198МашошинО. Ф.MashoshinO. F.

Олег Федорович Машошин, доктор технических наук, профессор

Кронштадтский бульвар, д. 20, Москва, 125493

Oleg F. Mashoshin, Doctor of Technical Sciences, Professor

20, Kronshtadtsky blvd, Moscow, 125493

o.mashoshin@mstuca.ruМосковский государственный технический университет гражданской авиацииРоссияMoscow State Technical University of Civil AviationRussian Federation202527112025037185Copyright © Гусейнов Г., Машошин О.Ф., 20252025Гусейнов Г., Машошин О.Ф.Huseynov H., Mashoshin O.F.This work is licensed under a Creative Commons Attribution 4.0 License.https://ce.if-mstuca.ru/jour/article/view/70

В работе представлены экспериментально обоснованные табличные данные для настройки гиперпараметров многослойных нейронных сетей в задачах диагностики авиационных газотурбинных двигателей. Предложены семь оригинальных алгоритмов адаптивной настройки параметров обучения, включающих методы динамической адаптации скорости обучения, стратегии изменения архитектуры сети в зависимости от режима работы двигателя и адаптивные подходы к регуляризации. Диапазоны параметров охватывают значения от 10–5 до 103, что обеспечивает практическую применимость для различных архитектур и типов данных. Научная новизна заключается в создании адаптивных алгоритмов, учитывающих специфику диагностических параметров компонентов ГТД и их временную динамику.

The paper presents experimentally substantiated tabular data for hyperparameter tuning of multilayer neural networks in aviation gas turbine engine diagnostics. The authors propose seven original algorithms for adaptive training parameter tuning, including methods for dynamic adaptation of the learning rate, strategies for changing the network architecture depending on the engine operating mode, and adaptive approaches to regularization. The parameter ranges cover values from 10-5 to 103, which ensures practical applicability for various architectures and data types. The scientific novelty lies in the creation of adaptive algorithms that take into account the specifics of the diagnostic parameters of gas turbine engine components and their time dynamics.

многослойные нейронные сетидиагностика авиационных двигателейгиперпараметрыадаптивная оптимизациягазотурбинные двигателимашинное обучениевременные рядыmultilayer neural networksaviation engine diagnosticshyperparametersadaptive optimizationgas turbine enginesmachine learningtime seriesReferencesКозлов В. М. Адаптивные алгоритмы настройки гиперпараметров нейронных сетей / В. М. Козлов, Е. С. Иванова // Известия РАН. Теория и системы управления. 2023. № 2. С. 78-89.Bai M., Liu J., Ma Y., Zhao X., Long Z., Yu D. (2021). Long short-term memory network-based normal pattern group for fault detection of three-shaft marine gas turbine. Energies. 14(1): 13. DOI 10.3390/en14010013.Машошин О. Ф. Разработка комплексного алгоритма обработки диагностических параметров авиационных ГТД на основе многослойных нейронных сетей / О. Ф. Машошин, Г. Гусейнов // Контроль. Диагностика. 2025. Т. 28, № 7. С. 41-54. DOI 10.14489/td.2025.07.pp.041-054.Bischl B., Binder M., Lang M., Pielok T., Richter J., Coors S., Thomas J., Ullmann T., Becker M., Boulesteix A.-L., Deng D., Lindauer M. (2023). Hyperparameter optimization: Foundations, algorithms, best practices, and open challenges. Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery. 13(2). DOI 10.1002/widm.1484.A Deep Learning Approach for Trajectory Control of Tilt-Rotor UAV / Ja. Sembiring, R. A. Sasongko, E. I. Bastian [et al.] // Aerospace. 2024. Vol. 11, № 1. P. 96. DOI 10.3390/aerospace11010096. EDN CDJAEF.Cruz Y. J., Castaño F., Haber R. E., Yarens J. Cruz, Fernando Castaño, Rodolfo E. Haber, Villalonga A., Ejsmont K., Gladysz B., Flores Á., Alemany P. (2024). Self-reconfiguration for smart manufacturing based on artificial intelligence: A review and case study. Artificial Intelligence in Manufacturing. Springer. 121-144. DOI 10.1007/978-3-031-46452-2-8.A review on gas turbine gas-path diagnostics: State-of-the-art methods, challenges and opportunities / A. D. Fentaye, A. T. Baheta, S. I. Gilani, K. G. Kyprianidis // Aerospace. 2019. Vol. 6, № 7. P. 83. DOI 10.3390/aerospace6070083. EDN NIFUIL.Hashmi M. B., Mansouri M., Fentaye A. D., Ahsan S., Kyprianidis K. (2024). An artificial neural network-based fault diagnostics approach for hydrogen-fueled micro gas turbines. Energies. 17(3): 719. DOI 10.3390/en17030719.An Artificial Neural Network-Based Fault Diagnostics Approach for Hydrogen-Fueled Micro Gas Turbines / M. B. Hashmi, M. Mansouri, A. D. Fentaye [et al.] // Energies. 2024. Vol. 17, № 3. P. 719. DOI 10.3390/en17030719. EDN SYQIEH.He X., Zhao K., Chu X. (2021). AutoML: A survey of the state-of-the-art. Knowledge-Based Systems. 212: 106622. DOI 10.1016/j.knosys.2020.106622.Deep transfer learning strategy in intelligent fault diagnosis of gas turbines based on the Koopman operator / F. N. Irani, M. Soleimani, M. Yadegar, N. Meskin // Applied Energy. 2024. Vol. 365. P. 123256. DOI 10.1016/j.apenergy.2024.123256. EDN GALUKY.Irani F. N., Soleimani M., Yadegar M., Meskin N. (2024). Deep transfer learning strategy in intelligent fault diagnosis of gas turbines based on the Koopman operator. Applied Energy. 365: 123256. DOI 10.1016/j.apenergy.2024.123256.Dynamic Temporal Denoise Neural Network with Multi-Head Attention for Fault Diagnosis Under Noise Background / Zh. Li, R. Fan, J. Ma [et al.] // Sensors. 2024. Vol. 24, № 21. P. 6813. DOI 10.3390/s24216813. EDN ERSVJI.Jin Y., Hou L., Chen Y. (2022). A time series transformer based method for the rotating machinery fault diagnosis. Neurocomputing. 494: 379-395. DOI 10.1016/j.neucom.2022.04.111.Fault diagnosis of gas turbine based on partly interpretable convolutional neural networks / D. Zhou, Q. Yao, H. Wu [et al.] // Energy. 2020. Vol. 200. P. 117467. DOI 10.1016/j.energy.2020.117467. EDN MBPRPN.Kandasamy K., Vysyaraju K. R., Neiswanger W., Paria B., Collins C. R., Schneider J., Poczos B., Xing E. P. (2020). Tuning hyperparameters without grad students: Scalable and robust bayesian optimisation with dragonfly. The Journal of Machine Learning Research. 21(1): 3098-3124.He X. AutoML: A survey of the state-of-the-art / X. He, K. Zhao, X. Chu // Knowledge-Based Systems. 2021. Vol. 212. P. 106622. DOI 10.1016/j.knosys.2020.106622. EDN HGRDWI.Kozlov V. M, Ivanova E. S. (2023). Adaptive algorithms for tuning hyperparameters of neural networks. Izvestia RAN. Theory and control systems. 2: 78-89. (In Russian)HELP: An LSTM-based approach to hyperparameter exploration in neural network learning / W. Li, W. W. Y Ng, T. Wang [et al.] // Neurocomputing. 2021. Vol. 442. P. 161-172. DOI 10.1016/j.neucom.2020.12.133. EDN CPLZSW.Li J. Y., Zhan Z. H., Wang C. (2021). Evolutionary Deep Learning Survey. Neurocomputing. 442: 89-109.Hyperparameter optimization: Foundations, algorithms, best practices, and open challenges / B. Bischl, M. Binder, M. Lang [et al.] // Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery. 2023. Vol. 13, № 2. DOI 10.1002/widm.1484. EDN AWVZRK.Li Y., Dong J., Jiang H., Su D. (2023). Multi-head spatio-temporal attention based parallel GRU architecture: a novel multi-sensor fusion method for mechanical fault diagnosis. Measurement Science and Technology. 35. DOI 10.1088/1361-6501/acf89e.Jin Yu. A Time Series Transformer based method for the rotating machinery fault diagnosis / Yu. Jin, L. Hou, Yu. Chen // Neurocomputing. 2022. Vol. 494. P. 379-395. DOI 10.1016/j.neucom.2022.04.111. EDN JLRUHL.Li Z., Fan R., Ma J., Ai J., Dong Y. (2024). Dynamic temporal denoise neural network with multihead attention for fault diagnosis under noise background. Sensors. 24(21): 6813. DOI 10.3390/s24216813.Li J. Y. Evolutionary Deep Learning Survey / J. Y. Li, Z. H. Zhan, C. Wang // Neurocomputing. 2021. № 442. Pp. 89-109.Liu X., Wu J., Zhou Z. (2021). Hyperparameter Exploration LSTM-Predictor (HELP). Neurocomputing. 442: 161-172.Long short-term memory network-based normal pattern group for fault detection of three-shaft marine gas turbine / M. Bai, J. Liu, Y. Ma, X. Zhao, Z. Long, D. Yu // Energies. 2021. № 14(1). Pp. 13. DOI 10.3390/en14010013.Mashoshin O. F., Huseynov H. (2025). Development of an Integrated Algorithm for Processing Aircraft GTE Diagnostic Parameters Using Multilayer Neural Networks. Kontrol'. Diagnostika. 28 (07): 41-54. (In Russian). DOI 10.14489/td.2025.07. pp.041-054.Multi-head spatio-temporal attention based parallel GRU architecture: a novel multi-sensor fusion method for mechanical fault diagnosis / Y. Li, J. Dong, H. Jiang, D. Su // Measurement Science and Technology. 2023. № 35. DOI 10.1088/1361-6501/acf89e.Pei X., Zheng X., Wu J. (2021). Rotating machinery fault diagnosis through a transformer convolution network subjected to transfer learning. IEEE Transactions on Instrumentation and Measurement. 70: 1-11.Pei X. Rotating machinery fault diagnosis through a transformer convolution network subjected to transfer learning / X. Pei, X. Zheng, J. Wu // IEEE Transactions on Instrumentation and Measurement. 2021. № 70. Pp. 1-11. DOI 10.1109/TIM.2021.3119137.Sembiring J., Sasongko R. A., Bastian E. I., Raditya B. A., Limansubroto R. E. (2024). A deep learning approach for trajectory control of tilt-rotor UAV. Aerospace. 11(1): 96.Performance-based health monitoring, diagnostics and prognostics for condition-based maintenance of gas turbines: A review / M. Tahan, M. Muhammad, Z. A. Abdul Karim, E. Tsoutsanis // Applied Energy. 2017. Vol. 198. P. 122-144. DOI 10.1016/j.apenergy.2017.04.048. EDN YDANFB.Smith L. N. Cyclical learning rates for training neural networks. (2017). IEEE Winter Conference on Applications of Computer Vision. 464-472.Self-reconfiguration for smart manufacturing based on artificial intelligence: A review and case study / Y. J. Cruz, F. Castaño, R. E. Haber [et al.] // Artificial Intelligence in Manufacturing. Springer. 2024. Pp. 121-144. DOI 10.1007/978-3-031-46452-2-8.Tahan M., Tsoutsanis E., Muhammad M., Karim Z. A. A. (2017). Performance-based health monitoring, diagnostics and prognostics for condition-based maintenance of gas turbines: A review. Applied Energy. 198: 122-144.Smith L. N. Cyclical learning rates for training neural networks // IEEE Winter Conference on Applications of Computer Vision. 2017.Pp. 464-472. DOI 10.1109/WACV.2017.58.Tsoutsanis E., Meskin N., Benammar M., Khorasani K. (2019). A review on gas turbine gas-path diagnostics: state-of-the-art methods, challenges and opportunities. Aerospace. 6(7): 83.Tuning hyperparameters without grad students: Scalable and robust bayesian optimisation with dragonfly / K. Kandasamy, K. R. Vysyaraju, W. Neiswanger [et al.] // Journal of Machine Learning Research. 2020. Vol. 21. EDN YRTFCN.Zhou D., Yao Q., Wu H., Ma S., Zhang H. (2020). Fault diagnosis of gas turbine based on partly interpretable convolutional neural networks. Energy. 200. DOI 10.1016/j.energy.2020.117467.

The authors declare that there are no conflicts of interest present.