Rejim Duyarlı Transformer Mimarisi ile Havacılık Motorlarında Kalan Yararlı Ömür Kestirimi

Öz

Kalan Yararlı Ömür (Remaining Useful Life - RUL) tahmini endüstriyel sistemlerin güvenilirliğini artırmak amacıyla son yıllarda büyük önem kazanmıştır. Kestirimci bakım (Predictive Maintenance - PdM) kapsamında RUL tahmini, bakım karar verme süreçlerinin desteklenmesi ve bakım faaliyetlerinin daha etkin planlanması açısından kritik bir rol oynamaktadır. Havacılık alanında da PdM yöntemlerinin entegrasyonuna ihtiyaç duyulmaktadır. Ancak havacılık motorlarının sabit koşullar yerine farklı irtifalar ve hızlar gibi değişken çalışma koşulları altında faaliyet göstermesi, sensör ölçümlerinin yorumlanmasını zorlaştırmaktadır. Bu durum RUL tahmin performansını olumsuz etkilemektedir. Bu çalışmada, farklı çalışma koşulları altında çalışan havacılık motorları için ReACT adı verilen rejim farkındalıklı bir mimari önerilmektedir. Önerilen mimari iki ana aşamadan oluşmaktadır. İlk aşamada, K-means kümeleme yöntemi kullanılarak çalışma rejimleri belirlenmekte ve verilere rejime duyarlı normalizasyon uygulanmaktadır. İkinci aşamada ise Etkili Kanal Dikkati (Efficient Channel Attention) mekanizması ile iki aşamalı dikkat yapısını birleştiren bir Transformer mimarisi kullanılmaktadır. Bu yapı sayesinde zamansal bozunma örüntüleri ve sensörler arasındaki ilişkiler etkin biçimde modellenmektedir. NASA C-MAPSS veri seti üzerinde gerçekleştirilen deneysel sonuçlar, ReACT mimarisinin kalan yararlı ömür tahmin hatalarını önemli ölçüde azalttığını ve mevcut yöntemlere kıyasla RMSE ve NASA Skoru metriklerinde anlamlı iyileşmeler sağladığını göstermektedir.

Anahtar Kelimeler

• Kalan Yararlı Ömür
• Transformer
• Rejim Farkındalığı
• Derin Öğrenme
• C-MAPSS
• Kestirimci Bakım

Etik Beyan

Bu makalenin yazarı, bu çalışmada kullanılan materyaller ve yöntemlerin etik kurul izni ve/veya özel yasal izin gerektirmediğini beyan eder.

Teşekkür

Yazar, sağladığı araştırma ortamı ve teknik altyapı için Türk Havacılık ve Uzay Sanayii'ne (TUSAŞ) teşekkürlerini sunar.

Remaining Useful Life Estimation in Aero-Engines via Regime-Aware Transformer Architecture

Öz

Remaining Useful Life (RUL) estimation has gained significant importance in recent years to improve the reliability of industrial systems. Within the scope of Predictive Maintenance (PdM), RUL estimation plays a critical role in supporting maintenance decision-making processes and planning maintenance activities more effectively. The integration of PdM methods is also needed in the aviation sector. However, the fact that aircraft engines operate under variable conditions, such as different altitudes and speeds, rather than constant conditions, makes the interpretation of sensor measurements difficult. This negatively affects RUL estimation performance. In this study, a regime-aware architecture called ReACT is proposed for aircraft engines operating under different conditions. The proposed architecture consists of two main stages. In the first stage, operating regimes are determined using the K-means clustering method, and regime-aware normalization is applied to the data. In the second stage, a Transformer architecture combining a two-stage attention mechanism with the Efficient Channel Attention mechanism is used. This structure effectively models temporal degradation patterns and relationships between sensors. Experimental results on the NASA C-MAPSS dataset show that the ReACT architecture significantly reduces remaining useful life estimation errors and provides significant improvements in RMSE and NASA Score metrics compared to existing methods.

Anahtar Kelimeler

• Remaining Useful Life
• Transformer
• Regime Awareness
• Deep Learning
• C-MAPSS
• Predictive Maintenance

Etik Beyan

The author of this article declares that the materials and methods used in this study do not require ethical committee permission and/or legal-special permission.

Teşekkür

The author expresses acknowledgments to Turkish Aerospace (TUSAŞ) for the research environment and technical infrastructure provided.

Kaynakça

1. [1] Lee J., Bagheri B. ve Kao H.-A., “A cyber-physical systems architecture for industry 4.0-based manufacturing systems”, Manufacturing Letters, 3: 18-23, (2015).
2. [2] Jardine A. K., Lin D. ve Banjevic D., “A review on machinery diagnostics and prognostics implementing condition-based maintenance”, Mechanical Systems and Signal Processing, 20(7): 1483-1510, (2006).
3. [3] Stanton I., Munir K., Ikram A. ve El-Bakry M., “Predictive maintenance analytics and implementation for aircraft: Challenges and opportunities”, Systems Engineering, 26(2): 216-237, (2023).
4. [4] Berghout T. ve Benbouzid M., “A Systematic Guide for Predicting Remaining Useful Life with Machine Learning”, Electronics, 11(7): 1125, (2022).
5. [5] Saxena A., Goebel K., Simon D. ve Eholzer N., “Damage propagation modeling for aircraft engine run-to-failure simulation”, 2008 International Conference on Prognostics and Health Management, 1-9, (2008).
6. [6] Erdoğan M. ve Mercimek M., “A comparative study on data pre-processing techniques for remaining useful life prediction of turbofan engines”, International Journal of Materials Engineering Technologies, 6(2): 50-58, (2023).
7. [7] Ndao M. L., Youness G., Niang N. ve Saporta G., “Improving predictive maintenance: Evaluating the impact of preprocessing and model complexity on the effectiveness of explainable artificial intelligence methods”, Engineering Applications of Artificial Intelligence, 144: 110144, (2025).
8. [8] Khan S. ve Yairi T., “A review on the application of deep learning in system health management”, Mechanical Systems and Signal Processing, 107: 241-265, (2018).

9. [9] Wu F., Wu Q., Tan Y. ve Xu X., “Remaining useful life prediction based on deep learning: a survey”, Sensors, 24(11): 3454, (2024).
10. [10] Zhu J., Ge Z., Song Z. ve Gao F., “Review and big data perspectives on robust data mining approaches for industrial process modeling with outliers and missing data”, Annual Reviews in Control, 46: 107-133, (2018).
11. [11] Luo J., Pattipati K. R., Qiao L. ve Chigusa S., “Model-based prognostic techniques applied to a suspension system”, IEEE Transactions on Systems, Man, and Cybernetics Part A: Systems and Humans, 38(5): 1156-1168, (2008).
12. [12] Khelif R., Chebel-Morello B., Malinowski S., Laajili E., Fnaiech F. ve Zerhouni N., “Direct remaining useful life estimation based on support vector regression”, IEEE Transactions on Industrial Electronics, 64(3): 2276-2285, (2016).
13. [13] Li X., Ding Q. ve Sun J., “Remaining useful life estimation in prognostics using deep convolutional neural networks”, Reliability Engineering and System Safety, 172: 1-11, (2018).
14. [14] Wu Y., Yuan M., Dong S., Lin L. ve Liu Y., “Remaining useful life estimation of engineered systems using vanilla LSTM neural networks”, Neurocomputing, 275: 167-179, (2018).
15. [15] Ellefsen A. L., Bjørlykhaug E., Æsøy V., Ushakov S. ve Zhang H., “Remaining useful life predictions for turbofan engine degradation using semi-supervised deep architecture”, Reliability Engineering and System Safety, 183: 240-251, (2019).
16. [16] Babu G. S., Zhao P. ve Li X. L., “Deep convolutional neural network based regression approach for estimation of remaining useful life”, International Conference on Database Systems for Advanced Applications, 214-228, (2016).
17. [17] Zheng S., Ristovski K., Farahat A. ve Gupta C., “Long short-term memory network for remaining useful life estimation”, 2017 IEEE International Conference on Prognostics and Health Management, 88-95, (2017).
18. [18] Guilherme D. N. V. B., “Remaining Useful Life Prediction on the NASA CMAPSS Dataset Comparing LSTM and Transformer Models”, Master’s Thesis, Instituto Superior de Engenharia do Porto, (2025).
19. [19] Yan H., Zuo H., Sun J., Zhou D. ve Wang H., “Two-stage degradation assessment and prediction method for aircraft engine based on data fusion”, International Journal of Aerospace Engineering, 2021(1): 5876299, (2021).
20. [20] Chao M. A., Kulkarni C., Goebel K. ve Fink O., “Fusing physics-based and deep learning models for prognostics”, Reliability Engineering and System Safety, 217: 107961, (2022).
21. [21] Jiang F., Hou X. ve Xia M., “Spatio-temporal attention-based hidden physics-informed neural network for remaining useful life prediction”, Advanced Engineering Informatics, 63: 102958, (2025).
22. [22] Boujamza A. ve Elhaq S. L., “Attention-based LSTM for remaining useful life estimation of aircraft engines”, IFAC-PapersOnLine, 55(12): 450-455, (2022).
23. [23] Akkad K. ve He D., “A hybrid deep learning based approach for remaining useful life estimation”, 2019 IEEE International Conference on Prognostics and Health Management, 1-6, (2019).
24. [24] Dida M., Cheriet A. ve Belhadj M., “Remaining useful life prediction using attention-LSTM neural network of aircraft engines”, International Journal of Prognostics and Health Management, 16(2), (2025).
25. [25] Li H. ve Wang J., “Based on wavelet threshold denoising-LDA and BiLSTM aircraft engine life prediction”, Journal of Physics: Conference Series, 012019, (2022).
26. [26] Zhao R., Yan R., Wang J. ve Mao K., “Learning to monitor machine health with convolutional bi-directional LSTM networks”, Sensors, 17(2): 273, (2017).
27. [27] Star M. ve McKee K., “Remaining useful life estimation using neural ordinary differential equations”, International Journal of Prognostics and Health Management, 12(2), (2021).
28. [28] Vaswani A., Shazeer N., Parmar N., Uszkoreit J., Jones L. ve Polosukhin I., “Attention Is All You Need”, Advances in Neural Information Processing Systems, 30, (2017).
29. [29] Mo Y., Wu Q., Li X. ve Huang B., “Remaining useful life estimation via transformer encoder enhanced by a gated convolutional unit”, Journal of Intelligent Manufacturing, 32(7): 1997-2006, (2021).
30. [30] Wang L., Li Y., Liu H. ve Liu T., “A lightweight transformer edge intelligence model for RUL prediction classification”, Sensors, 25: 4224, (2025).
31. [31] Wang H., Lin T., Cui L., Ma B., Dong Z. ve Song L., “Multitask learning-based self-attention encoding atrous convolutional neural network for remaining useful life prediction”, IEEE Transactions on Instrumentation and Measurement, 71: 1-8, (2022).
32. [32] Li Z., Luo S., Liu H., Tang C. ve Miao J., “TTSNet: Transformer–Temporal Convolutional Network–Self-Attention with Feature Fusion for Prediction of Remaining Useful Life of Aircraft Engines”, Sensors, 25(2): 432, (2025).
33. [33] Wahid A., Breslin J. G. ve Intizar M. A., “TCRSCANet: Harnessing temporal convolutions and recurrent skip component for enhanced RUL estimation in mechanical systems”, Human-Centric Intelligent Systems, 4(1): 1-24, (2024).
34. [34] Chen X., “A novel transformer-based DL model enhanced by position-sensitive attention and gated hierarchical LSTM for aero-engine RUL prediction”, Scientific Reports, 14(1): 10061, (2024).
35. [35] Fan Z., Li W. ve Chang K. C., “A two-stage attention-based hierarchical transformer for turbofan engine remaining useful life prediction”, Sensors, 24(3): 824, (2024).
36. [36] Wang Y. vd., “Fully-connected spatial-temporal graph for multivariate time-series data”, Proceedings of the AAAI Conference on Artificial Intelligence, 15715-15724, (2024).
37. [37] Lu Z. vd., “A multi-scale cross-dimension interaction approach with adaptive dilated TCN for RUL prediction”, Scientific Reports, 15(1): 19229, (2025).
38. [38] Ramasso E. ve Saxena A., “Performance benchmarking and analysis of prognostic methods for CMAPSS datasets”, International Journal of Prognostics and Health Management, 5(2): 1-15, (2014).
39. [39] Li X., Li J., Zuo L., Zhu L. ve Shen H. T., “Domain adaptive remaining useful life prediction with transformer”, IEEE Transactions on Instrumentation and Measurement, 71: 1-13, (2022).
40. [40] Liu Q., Zhang Z., Guo P., Wang Y. ve Liang J., “Enhancing aircraft engine remaining useful life prediction via multiscale deep transfer learning with limited data”, Journal of Computational Design and Engineering, 11(1): 343-355, (2024).
41. [41] Hu W. ve Zhao S., “Remaining useful life prediction of lithium-ion batteries based on wavelet denoising and transformer neural network”, Frontiers in Energy Research, 10: 969168, (2022).
42. [42] Chen Z., Wu M., Zhao R., Guretno F., Yan R. ve Li X., “Machine remaining useful life prediction via an attention-based deep learning approach”, IEEE Transactions on Industrial Electronics, 68(3): 2521-2531, (2020).
43. [43] Hou G., Xu S., Zhou N., Yang L. ve Fu Q., “Remaining useful life estimation using deep convolutional generative adversarial networks based on an autoencoder scheme”, Computational Intelligence and Neuroscience, 2020(1): 9601389, (2020).
44. [44] Lei Y., Li N., Guo L., Yan T. ve Lin J., “Machinery health prognostics: A systematic review from data acquisition to RUL prediction”, Mechanical Systems and Signal Processing, 104: 799-834, (2018).
45. [45] Chaoub A., Voisin A., Cerisara C. ve Iung B., “Learning representations with end-to-end models for improved remaining useful life prognostics”, (2021).
46. [46] Hou C. ve Zheng L., “A multi-task joint learning model based on transformer and customized gate control for predicting remaining useful life and health status of tools”, Sensors, 24(13): 4117, (2024).
47. [47] Wang W., He Q., Cui Y. ve Li Z., “Joint prediction of remaining useful life and failure type of train wheelsets: Multitask learning approach”, Journal of Transportation Engineering Part A: Systems, 144(6): 04018016, (2018).
48. [48] Wang Q., Wu B., Zhu P., Li P., Zuo W. ve Hu Q., “ECA-Net: Efficient channel attention for deep convolutional neural networks”, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11534-11542, (2020).
49. [49] Xiang S., Qin Y., Luo J., Pu H. ve Tang B., “Multicellular LSTM-based deep learning model for aero-engine remaining useful life prediction”, Reliability Engineering and System Safety, 216: 107927, (2021).
50. [50] Kingma D. P., “Adam: A method for stochastic optimization”, (2014).
51. [51] Lundberg S. M. ve Lee S.-I., “A unified approach to interpreting model predictions”, Advances in Neural Information Processing Systems, 30, (2017).
52. [52] Jakubiak N., “Analysis of Explainable Artificial Intelligence on Time Series Data”, Master’s Thesis, Universitat Politècnica de Catalunya, (2022).
53. [53] Costa N. ve Sánchez L., “Variational encoding approach for interpretable assessment of remaining useful life estimation”, Reliability Engineering and System Safety, 222: 108353, (2022).
54. [54] Jutte A., Ahmed F., Linssen J. ve van Keulen M., “C-SHAP for time series: An approach to high-level temporal explanations”, (2025).
55. [55] Tibshirani R. J., Barber R. F., Candès E. J. ve Ramdas A., “Conformal prediction under covariate shift”, Advances in Neural Information Processing Systems, 32, (2019).
56. [56] Javanmardi A. ve Hüllermeier E., “Conformal prediction intervals for remaining useful lifetime estimation”, (2022).
57. [57] Deng S. ve Zhou J., “Prediction of remaining useful life of aero-engines based on CNN-LSTM-Attention”, International Journal of Computational Intelligence Systems, 17(1): 232, (2024).
58. [58] Wang H.-K., Cheng Y. ve Song K., “Remaining useful life estimation of aircraft engines using a joint deep learning model based on TCNN and transformer”, Computational Intelligence and Neuroscience, 2021(1): 5185938, (2021).
59. [59] Şener M. C. ve Karacan H., “Deep Learning Technologies in Dental Practice: Current Applications and Research Trends”, Politeknik Dergisi, 29(2): 1-17, (2026).
60. [60] Sözen Z. ve Barışçı N., “Derin Öğrenme ile Hücre Görüntülerinin Tespiti ve Sayımı”, Journal of Polytechnic, 28(3): 909-921, (2025).
61. [61] İşler B., “Deep Learning-Based Speech Emotion Recognition for Iot Edge Devices: A Comparative Study”, Journal of Polytechnic, 29(4): 1-13, (2026).