RULMAN ARIZA TEŞHİSİ İÇİN CNN-LSTM VE CNN-GRU HİBRİT DERİN ÖĞRENME MODELLERİNİN KARŞILAŞTIRMALI BİR ÇALIŞMASI

Abstract

Rulman arıza teşhisi, endüstriyel döner makinelerin güvenli ve verimli çalışmasının sağlanması açısından kritik bir öneme sahiptir. Bu çalışmada, rulman arızalarının sınıflandırılması amacıyla iki farklı hibrit derin öğrenme modeli olan Evrişimsel Sinir Ağı–Uzun Kısa Süreli Bellek (CNN-LSTM) ve Evrişimsel Sinir Ağı–Geçitli Tekrarlayan Birim (CNN-GRU) modelleri önerilmiş ve karşılaştırmalı olarak analiz edilmiştir. Çalışmada, Case Western Reserve University (CWRU) rulman veri setinden elde edilen titreşim sinyalleri kullanılmıştır. CNN yapısı, ham sinyallerden otomatik özellik çıkarımı için kullanılırken, LSTM ve GRU katmanları ise verilerin zamansal bağımlılıklarını modellemek amacıyla kullanılmıştır. Önerilen modellerin performansı doğruluk (accuracy), kesinlik (precision), duyarlılık (recall) ve F1-skoru metrikleri kullanılarak değerlendirilmiştir. Elde edilen sonuçlara göre, CNN-LSTM modeli %96.32 test doğruluğu elde ederken, CNN-GRU modeli %97.79 doğruluk ile daha yüksek bir performans sergilemiştir. Karışıklık matrisleri ve sınıf bazlı analizler, CNN-GRU modelinin özellikle zor ayrıştırılan arıza sınıflarında daha kararlı ve başarılı sonuçlar ürettiğini göstermektedir. Sonuç olarak, hibrit derin öğrenme modellerinin rulman arıza teşhisi problemlerinde oldukça etkili olduğu ve CNN-GRU modelinin daha basit yapısı ve güçlü zamansal öğrenme yeteneği sayesinde daha üstün performans sağladığı belirlenmiştir. Bu çalışma, endüstriyel durum izleme sistemlerinde güvenilir ve akıllı arıza teşhis yöntemlerinin geliştirilmesine katkı sunmaktadır.

Keywords

• Rulman Arıza Teşhisi
• Derin Öğrenme
• CNN-LSTM
• CNN-GRU
• Titreşim Analizi

A COMPARATİVE STUDY OF CNN-LSTM AND CNN-GRU HYBRİD DEEP LEARNİNG MODELS FOR BEARİNG FAULT DİAGNOSİS

Abstract

Bearing fault diagnosis is of critical importance for ensuring the safe and efficient operation of industrial rotating machinery. In this study, two different hybrid deep learning models—Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) and Convolutional Neural Network–Gated Recurrent Unit (CNN-GRU)—are proposed and comparatively analyzed for the classification of bearing faults. Vibration signals obtained from the Case Western Reserve University (CWRU) bearing dataset are utilized in the study. While the CNN structure is employed for automatic feature extraction from raw signals, the LSTM and GRU layers are used to model the temporal dependencies of the data. The performance of the proposed models is evaluated using accuracy, precision, recall, and F1-score metrics. According to the obtained results, the CNN-LSTM model achieves a test accuracy of 96.32%, whereas the CNN-GRU model demonstrates superior performance with an accuracy of 97.79%. Confusion matrices and class-wise analyses indicate that the CNN-GRU model produces more stable and accurate results, particularly for fault classes that are difficult to distinguish. In conclusion, hybrid deep learning models are highly effective for bearing fault diagnosis problems, and the CNN-GRU model provides better performance due to its simpler architecture and strong capability in modeling temporal dependencies. This study contributes to the development of reliable and intelligent fault diagnosis methods for industrial condition monitoring systems.

Keywords

• Bearing Fault Diagnosis
• Deep Learning
• CNN-LSTM
• CNN-GRU
• Vibration Analysis

References

1. Huo, C., Jiang, Q., Shen, Y., Lin, X., Zhu, Q., & Zhang, Q. (2023). A class-level matching unsupervised transfer learning network for rolling bearing fault diagnosis under various working conditions. Applied Soft Computing, 146, 110739.
2. Wu, Z., Jiang, H., Zhu, H., & Wang, X. (2023). A knowledge dynamic matching unit-guided multi-source domain adaptation network with attention mechanism for rolling bearing fault diagnosis. Mechanical Systems and Signal Processing, 189, 110098.
3. Li, F., Wang, L., Wang, D., Wu, J., & Zhao, H. (2023). An adaptive multiscale fully convolutional network for bearing fault diagnosis under noisy environments. Measurement, 216, 112993.
4. Zhang, J., Chen, J., Deng, H., & Hu, W. (2023). A novel framework based on adaptive Multi-Task learning for bearing fault diagnosis. Energy Reports, 9, 522-531.
5. Wang, P., Xiong, H., & He, H. (2023). Bearing fault diagnosis under various conditions using an incremental learning-based multi-task shared classifier. Knowledge-Based Systems, 266, 110395.
6. Hou, Y., Wang, J., Chen, Z., Ma, J., & Li, T. (2023). Diagnosisformer: An efficient rolling bearing fault diagnosis method based on improved Transformer. Engineering Applications of Artificial Intelligence, 124, 106507.
7. Gu, X., Yu, Y., Guo, L., Gao, H., & Luo, M. (2023). CSWGAN-GP: A new method for bearing fault diagnosis under imbalanced condition. Measurement, 217, 113014.
8. Tang, G., Yi, C., Liu, L., Yang, X., Xu, D., Zhou, Q., & Lin, J. (2023). A novel transfer learning network with adaptive input length selection and lightweight structure for bearing fault diagnosis. Engineering Applications of Artificial Intelligence, 123, 106395.

9. Zhao, K., Xiao, J., Li, C., Xu, Z., & Yue, M. (2023). Fault diagnosis of rolling bearing using CNN and PCA fractal based feature extraction. Measurement, 113754.
10. Yi, H., Li, D., Lu, Z., Jin, Y., Duan, H., Hou, L., Duraihem, F. Z., Awwad, E. M., & Saeed, N. A. (2025). VibrMamba: A lightweight Mamba based fault diagnosis of rotating machinery using vibration signal. Measurement, 249, 116881.
11. Noussis, A., Saif, A., Khatab, A., & Diallo, C. (2024). Using a Minimalist Bi-LSTM for Multi-Faceted Bearing Fault Detection. IFAC PapersOnLine, 58(19), 1258–1263.
12. Wang, J., Song, P., & Zhao, C. (2024). Towards consensual representation: Model-agnostic knowledge extraction for dual heterogeneous federated fault diagnosis. Neural Networks, 179, 106618.
13. Liu, Y., & Li, W. (2025). Time-frequency ensemble network for wind turbine gearbox fault diagnosis under variable working conditions. Engineering Science and Technology, an International Journal, 38, 102410.
14. Yu, W., Zhang, Q., Feng, X., Wang, K., Cai, Y., & Wu, D. (2025). Threshold-optimized and features-fused semi-supervised domain adaptive bearing fault diagnosis via chaotic sparrow search and deep belief network. Neurocomputing, 633, 128992.
15. Wang, H., Lv, D., Lin, T., Han, C., & Song, L. (2025). Task-adaptive unbiased regularization meta-learning for few-shot cross-domain fault diagnosis. Engineering Applications of Artificial Intelligence, 144, 110200.
16. O’Neill, L. E., Izenson, M. G., Kharangate, C. R., Pirnstill, L., & Li, J. (2025). Steady boiling and transient quenching of cryogenic fluids: new experiments to enable advanced propellant tank design and operations planning. Cryogenics, 151, 104177.
17. Gao, T., Xie, X., Wei, N., Shu, M., & Geng, Y. (2025). Advancing few-shot rolling element bearing diagnostics with fine-grained similarity learning across variable working environments. Neurocomputing, 131093.
18. Akcan, E. (2025). Detection of Bearing Faults from Vibration Signals. Balkan Journal of Electrical and Computer Engineering, 13(3), 295-306.
19. Ayadi, W., Elhamzi, W., Charfi, I., & Atri, M. (2021). Deep CNN for brain tumor classification. Neural processing letters, 53(1), 671-700.
20. Lei, X., Pan, H., & Huang, X. (2019). A dilated CNN model for image classification. IEEE access, 7, 124087-124095.
21. Sharma, A., & Phonsa, G. (2021). Image classification using CNN. In Proceedings of the international conference on innovative computing & communication (ICICC).
22. González-Planet, I. C., & Juez, C. (2025). Hydrological insights from a comparative evaluation of LSTM and MC-LSTM networks in the Ebro River basin (Spain). Journal of Hydrology: Regional Studies, 61, 102719.
23. Pai, A., Mishra, K. K., Jeyan, J. M. L., & Sayal, A. (2025). Enhanced household energy consumption forecasting using multivariate long short-term memory (LSTM) networks with weather data integration. Results in Engineering, 106512.
24. Minh, T. Q., Matos, J. C., Sousa, H. S., Ngoc, S. D., Van, T. N., Nguyen, H. X., & Nguyễn, Q. (2025). Data reconstruction leverages one-dimensional convolutional neural networks (1DCNN) combined with long short-term memory (LSTM) networks for structural health monitoring (SHM). Measurement, 253, 117810.
25. Zhang, N., Li, J., Ma, Y., & Wu, K. (2025). Lithium-Ion Batteries state of health estimation based on optimized TCN-GRU-WNN. Energy Reports, 13, 2502-2515.
26. Yunita, A., Pratama, M. I., Almuzakki, M. Z., Ramadhan, H., Akhir, E. A. P., Mansur, A. B. F., & Basori, A. H. (2025). Performance analysis of neural network architectures for time series forecasting: A comparative study of RNN, LSTM, GRU, and hybrid models. MethodsX, 15, 103462.