Flight mission segment automatic recognition based on bidirectional recurrent neural network

doi:10.23919/JSEE.2026.000135

Abstract

Abstract:

Flight mission segments are essential components of aero-engine flight mission profiles. Current recognition methods rely on manual processes, leading to inefficiencies and inaccuracies in aero-engine load analysis. To solve this problem, this paper presents an automatic method for recognizing flight mission segments using bidirectional recurrent neural network (Bi-RNN). Initially, the decomposition of flight mission profiles, manual division of mission segments, labeling of mission segment, and the establishment of maneuver sequence table are introduced, respectively. Secondly, the automatic recognition models are constructed using bidirectional long short-term memory (Bi-LSTM) and bidirectional gated recurrent unit (Bi-GRU). By collecting measured load data from a third-generation aero-engine, including 50 low maneuvering and 50 complex maneuvering flight mission profiles, the dataset is set utilizing K-fold cross-validation and the sliding window method. The optimal parameters for the models are determined through training, and their recognition performance is assessed. The results indicate that the Bi-RNN model achieves over 80% accuracy in recognizing flight mission segments in both low and complex maneuvering flight mission profiles. As a result, the flight mission segment automatic recognition method presented in this study allows for more efficient and precise aero-engine load analysis, which is of great significance for aero-engine life evaluation research.

Key words: aero-engine loading spectrum, flight mission segment automatic recognition, maneuver recognition, bidirectional long short-term memory (Bi-LSTM), bidirectional gate recurrent unit (Bi-GRU)

Yu Jin, Yingdong Song, Xuming Niu, Zhigang Sun. Flight mission segment automatic recognition based on bidirectional recurrent neural network[J]. Journal of Systems Engineering and Electronics, 2026, 37(4): 1282-1296.

Figures/Tables 17

Fig 1

Fig 2

Table 1

Fig 3

Table 2

Table 3

Fig 4

Fig 5

Fig 6

Fig 7

Table 4

Fig 8

Fig 9

Table 5

Fig 10

Fig 11

Fig 12

References 31

1	Su Q Y, Kong R L, Chen X X, et al. Guidelines for determining the lifespan of major components of aviation turbojet and turbofan engines[M]. Beijing: Aviation Industry Press, 2004. (in Chinese)
2	Liu Z Y, Ma X B, Hong D P, et al Mission reliability assessment for battle-plane based on flight profile. Journal of Beijing University of Aeronautics and Astronautics, 2012, 38 (1): 59.
3	Jin Y, Sun Z G, Song Y D, et al Flight mission segment division of the whole aeroengine loading spectrum based on maneuvers. Chinese Journal of Aeronautics, 2022, 35 (2): 164. doi: 10.1016/j.cja.2021.06.015
4	Yang X Y, Ning X R, Shi H Q, et al Research on turbo shaft aero engine composite flight profiles. Journal of Mechanical Strength, 2006, (6): 909.
5	Wang H Life verification and plan of china military aeroengine. Aeroengine, 2023, 49 (4): 80.
6	Yang X Y. Aeroengine life control technology[M]. Beijing: The Science Publishing Company, 2018. (in Chinese)
7	Breitkopf G E. Basic approach in the development of TURBISTAN, a loading standard for fighter aircraft engine disks[R]. West Conshohocken: ASTM International, 1989: 14.
8	Bergmann J S T W. Standardised load sequence for hot turbine and compressor discs of military aircraft[R]. Ottobrunn: IABG, 1990: 2809.
9	Heuler P, Schu¨tz W. A review of standardized load-time histories for fatigue research and application[J]. International Journal of Fatigue, 2002, 24(2/4): 203.
10	Heuler P. Generation and use of standardized load spectra and load-time histories[J]. International Journal of Fatigue, 2005, 27(8): 877.
11	Song P C. Research on aeroengine load spectrum model and load combination[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2006 (in Chinese).
12	Niu X M, Lu Q, Sun Z G, et al A novel compilation method of comprehensive mission spectrum of aero-engine maneuvering load based on use-related flight mission segment. Chinese Journal of Aeronautics, 2023, 36, 161. doi: 10.1016/j.cja.2022.09.016
13	Samuel K, Gadepally V, Jacobs D, et al. Maneuver identification challenge[C]//IEEE High Performance Extreme Computing Conference, 2021. DOI: 10.1109/HPEC49654.2021.9622788.
14	Pang N Y, Guan D H, Yuan W W An interpretable real-time maneuver identification algorithm based on early time series classification. Computer Engineering & Science, 2024, 46 (2): 353.
15	Wasilefsky, Devin W, Yan Z, et al. AI enabled maneuver identification via the maneuver identification challenge[C]//Interservice/Industry Training, Simulation, and Education Conference, 2022: 22475.
16	Wei C, Lin L L. Research and simulation implementation of airplane target typical motion model[C]//International Conference on Computational and Information Sciences, 2023: 1068.
17	Benterki A, Boukhnifer M, Judalet V, et al. Artificial intelligence for vehicle behavior anticipation: hybrid approach based on maneuver classification and trajectory prediction[J]. IEEE Access, 2020, 8: 56992.
18	Feng A, Han C, Gong J, et al. Multi-scale learnable gabor transform for pedestrian trajectory prediction from different perspectives[J]. IEEE Trans. on Intelligent Transportation Systems, 2024, 25(10): 13253.
19	Wang Y J, Dong J, Liu X D, et al Identification and standardization of maneuvers based upon operational flight data. Chinese Journal of Aeronautics, 2015, 28 (1): 133. doi: 10.1016/j.cja.2014.12.026
20	KARLI, Efe M O, Sever H Extracting basic fighter maneuvers from actual flight data. Lecture Notes on Information Theory, 2017, (5): 1.
21	Rong H J, Jia Y X, Zhao G S Aircraft recognition using modular extreme learning machine. Neurocomputing, 2014, 128 (5): 166.
22	Huang H L, Huang J C, Feng Y H, et al. Aircraft type recognition based on target track[J]. Journal of Physics: Conference Series, 2018: 1061.
23	Li X D, Pan J D, Dezert J. Automatic aircraft recognition using DSmT and HMM[C]//17th International Conference on Information Fusion, 2014. DOI: 10.5281/ZENODO.22437.
24	Xu X M, Yang R N, Fu Y Situation assessment for air combat based on novel semi-supervised naive Bayes. Journal of Systems Engineering and Electronics, 2018, 29 (4): 768. doi: 10.21629/jsee.2018.04.11
25	Schreier M, Willert V, Adamy J An integrated approach to maneuver-based trajectory prediction and criticality assessment in arbitrary road environments. IEEE Trans. on Intelligent Transportation Systems, 2016, 17 (10): 2751. doi: 10.1109/TITS.2016.2522507
26	Rodin E Y, Amin M Maneuver prediction in air combat via artificial neural networks. Computers & Mathematics with Applications, 1992, 24 (3): 95. doi: 10.1016/0898-1221(92)90217-6
27	Toledo A, Toledo-moreo R Maneuver prediction for road vehicles based on a novel neuro-fuzzy dynamic architecture. Robotics and Autonomous Systems, 2010, 58 (12): 1316. doi: 10.1016/j.robot.2010.09.002
28	Fang W, Wang Y, Yan W J, et al Symbolized flight action recognition based on neural network. Journal of Systems Engineering and Electronics, 2022, 44 (3): 737.
29	Ren L Q, Wang H P. Small sample recognition method and device for complex flight actions based on dual one-dimensional convolution attention mechanism[P]. China: CN117076999A, 2023. (in Chinese)
30	Zhou D, Zhuang X, Zuo H F A hybrid deep neural network based on multi-time window convolutional bidirectional LSTM for civil aircraft APU hazard identification. Chinese Journal of Aeronautics, 2022, 35 (4): 344. doi: 10.1016/j.cja.2021.03.031
31	Yang K. Research on short-term power load forecasting based on CNN and BiGRU combination model[D]. Daqing: Northeast Petroleum University, 2023. (in Chinese)

Basic maneuver	Flight altitude	Bank angle	Path angle	Pitch angle
Flat flight	≈	≈	≈	≈
Ascending (climbing, jumping)	↑	≈	≈	≈
Descending (gliding, diving)	↓	≈	≈	≈
Horizontal turn	≈	↑/↓	↑/↓	≈
Horizontal roll	≈	↑/↓	≈	≈
Vertical roll	↑/↓	↑/↓	≈	≈
Ascending turn	↑	↑/↓	↑/↓	—
Descending turn	↓	↑/↓	↑/↓	—
Pitching	≈	≈	≈	↑/↓

Basic maneuver	Basic maneuver number
Flat flight	1
Ascending (climbing, jumping)	2
Descending (gliding, diving)	3
Horizontal turn	4
Horizontal roll	5
Vertical roll	6
Ascending turn	7
Descending turn	8
Pitching	9

Flight mission segment	Mission segment label
Climbing	1
Ascending turn	2
Ascending-descending turn	3
Horizontal turn	4
Descending turn	5
Gliding	6
Flat flight	7
Horizontal-descending turn	8
Jumping	9
Diving	10
Split-S	11
Combat half roll	12
Loop	13
Combat turn	14

Layer	Input size	Output size
Bi-RNN layer	$ {N}_\text{sample}\times {N}_\text{feature}\times {N}_{w} $	$ {N}_\text{sample}\times {S}_{\textBi{I}} $
Bi-RNN layer	$ {N}_\text{sample}\times {S}_{\textBi{I}} $	$ {N}_\text{sample}\times {S}_{\textBi{II}} $
Full connection layer	$ {N}_\text{sample}\times {S}_{\textBi{II}} $	$ {N}_\text{sample}\times {N}_\text{class} $
SoftMax layer	$ {N}_\text{sample}\times {N}_\text{class} $	$ {N}_\text{sample}\times {N}_\text{class} $
Classification layer	$ {N}_\text{sample}\times {N}_\text{class} $	$ {N}_\text{sample}\times {N}_\text{class} $

Model parameter	Low maneuvering		Complex maneuvering
Model parameter	Bi-LSTM	Bi-GRU	Bi-LSTM	Bi-GRU
Initial learn rate	0.0001
Max training epoch	200
Optimizer	Adam
Sliding window size	9	6	9	6
Minimum batch size	256	256	128	512