Makine öğrenmesi yöntemleri ile raylı sistem araçlarında apleti arızalarının operasyonel ve meteorolojik veriler kullanılarak tahmini

Öz

Raylı sistemlerde tekerlek yüzeyinde oluşan apleti arızaları, dinamik darbe yüklerini artırarak işletme güvenilirliğini ve bakım süreçlerini olumsuz etkilemektedir. Bu çalışmada, operasyonel ve meteorolojik değişkenler kullanılarak apleti arızalarının erken tahminine yönelik açıklanabilir bir makine öğrenmesi yaklaşımı geliştirilmiştir. Kayseri raylı sistem işletmesinden 2015–2024 döneminde elde edilen veriler birleştirilerek 1562 gözlemden oluşan bir veri seti hazırlanmıştır. Lojistik Regresyon, Karar Ağacı, Rastgele Orman, Destek Vektör Makineleri, K-En Yakın Komşu ve XGBoost algoritmaları karşılaştırılmıştır. Sınıf dengesizliğinin etkisini değerlendirmek amacıyla SMOTE, Tomek Bağları ve rastgele aşağı örnekleme yöntemleri uygulanmıştır. Optuna ile hiperparametre optimizasyonu gerçekleştirilmiş, veri sızıntısını önlemek için örnekleme işlemleri iç içe çapraz doğrulamanın yalnızca eğitim katlarında yürütülmüştür. Tekil modeller arasında Rastgele Orman, 0,8758 doğruluk ve 0,8118 F1-skor ile öne çıkmıştır. Rastgele Orman ve XGBoost modellerini birleştiren katı oylama yaklaşımı 0,8828 doğruluk, 0,8984 kesinlik ve 0,8182 F1-skor sağlamıştır. SHAP analizi, yılın günü, teker çapı değişimi ve yılın haftasının tahminlerde belirleyici olduğunu göstermiştir. Bulgular, mevcut işletme kayıtlarının açıklanabilir arıza tahmini ve bakım kararlarının desteklenmesi amacıyla kullanılabileceğini ortaya koymaktadır. Böylece ek sensör altyapısına bağımlı olmayan, saha verilerine dayalı ve bakım planlamasına uyarlanabilecek sürdürülebilir bir karar destek çerçevesi sunulmaktadır.

Anahtar Kelimeler

• Raylı sistemler
• kestirimci bakım
• apleti arızası
• makine öğrenmesi
• topluluk öğrenmesi

Destekleyen Kurum

Erciyes Üniversitesi Bilimsel Araştırma Projeleri (BAP) Koordinasyon Birimi

Proje Numarası

FDK-2023-13350, FDK-2026-15932

Etik Beyan

Bu çalışma bilimsel araştırma ve yayın etiğine uygun olarak yürütülmüştür. Çalışmada herhangi bir etik kurul onayı gerektirecek insan ya da hayvan deneyi yapılmamıştır. Herhangi bir çıkar çatışması bulunmamaktadır.

Teşekkür

Bu çalışmada kullanılan apleti verilerinin incelenmesine olanak sağlayan Kayseri Ulaşım A.Ş.’ye teşekkür ederiz.

Prediction of wheel flat failures in railway vehicles using machine learning methods with operational and meteorological data

Öz

Wheel flat defects occurring on the wheel surfaces of railway vehicles increase dynamic impact loads, adversely affecting operational reliability and maintenance processes. This study develops an explainable machine learning approach for the early prediction of wheel flat failures using operational and meteorological variables. A dataset comprising 1,562 observations was constructed by integrating data collected from the Kayseri railway system during the 2015–2024 period. Six machine learning algorithms, namely Logistic Regression, Decision Tree, Random Forest, Support Vector Machines, K-Nearest Neighbors, and XGBoost, were comparatively evaluated. To assess the effects of class imbalance, SMOTE, Tomek Links, and random undersampling techniques were applied. Hyperparameter optimization was performed using Optuna, while resampling procedures were restricted to the training folds within nested cross-validation to prevent data leakage. Among the individual models, Random Forest achieved the best performance, with an accuracy of 0.8758 and an F1-score of 0.8118. The hard-voting ensemble combining Random Forest and XGBoost achieved an accuracy of 0.8828, precision of 0.8984, and an F1-score of 0.8182. SHAP analysis revealed that day of the year, wheel diameter change, and week of the year were the most influential predictors. The findings demonstrate that existing operational records can be utilized for explainable failure prediction and maintenance decision support. Thus, a sustainable, field-data-driven decision support framework is proposed that does not rely on additional sensor infrastructure and can be integrated into maintenance planning processes.

Anahtar Kelimeler

• Railway systems
• predictive maintenance
• wheel flat defect
• machine learning
• ensemble learning

Destekleyen Kurum

Erciyes University Scientific Research Projects (BAP) Coordination Unit

Proje Numarası

FDK-2023-13350, FDK-2026-15932

Etik Beyan

This study was conducted in accordance with scientific research and publication ethics. No human or animal experiments requiring ethics committee approval were performed in this study. There are no conflicts of interest.

Teşekkür

We would like to thank Kayseri Transportation Inc. for providing us with the opportunity to analyze the Applet data used in this study.

Kaynakça

1. 1. Cui, Y., Kim, E., Yan, S., & Yu, Q., Wheel flat detection using long short-term memory and transformer models with a 1: 10 scale railway test rig, Advances in Mechanical Engineering, 17 (1), 16878132251314988, 2025.
2. 2. Mosleh, A., Meixedo, A., Ribeiro, D., Montenegro, P., Calçada, R., Early wheel flat detection: an automatic data-driven wavelet-based approach for railways, Vehicle System Dynamics, 61 (6), 1644-1673,2023.
3. 3. Iwnicki, S., Nielsen, J. C., Tao, G., Out-of-round railway wheels and polygonisation, Vehicle system dynamics, 61 (7), 1787-1830, 2023.
4. 4. Fu, W., He, Q., Feng, Q., Li, J., Zheng, F., Zhang, B., Recent advances in wayside railway wheel flat detection techniques: A review, Sensors, 23 (8), 3916, 2023.
5. 5. Xu, D., Wang, J., Yang, J., Wu, Y., Wen, X., Effect of Wheel Polygonalization on the Dynamic Characteristics of Gear-Transmission Systems of Urban Railway Vehicles, Machines, 13 (4), 323, 2025.
6. 6. Salehi, M., Bagherzadeh, S. A., Fakhari, M., Experimental detection of train wheel defects using wayside vibration signal processing, Structural Health Monitoring, 22 (5), 3286-3301, 2023.
7. 7. Mosleh, A., Ghiasi, R., Gordan, M., Ribeiro, D., Malekjafarian, A., 2026, Wayside monitoring for railway wheel flat identification: A multiclass supervised learning approach, Proceedings of the Institution of Mechanical Engineers, Part F: Journal of Rail and Rapid Transit, 240 (1), 112-129, 2026.
8. 8. Ye, Y., Zhu, B., Huang, P., & Peng, B., OORNet: A deep learning model for on-board condition monitoring and fault diagnosis of out-of-round wheels of high-speed trains, Measurement, 199, 111268, 2022.

9. 9. Shi, D., Ye, Y., Gillwald, M., & Hecht, M., Designing a lightweight 1D convolutional neural network with Bayesian optimization for wheel flat detection using carbody accelerations. International Journal of Rail Transportation, 9 (4), 311-341, 2021.
10. 10. Bianchi, G., Fanelli, C., Freddi, F., Giuliani, F., & La Placa, A., Systematic review railway infrastructure monitoring: From classic techniques to predictive maintenance. Advances in Mechanical Engineering, 17 (1), 16878132241285631, 2025.
11. 11. Wang, X., Liu, X., & Bian, Z., A machine learning based methodology for broken rail prediction on freight railroads: A case study in the United States, Construction and Building Materials, 346, 128353, 2022.
12. 12. Shaikh, K., Hussain, I., & Chowdhry, B. S., Wheel defect detection using a hybrid deep learning approach, Sensors, 23 (14), 6248, 2023.
13. 13. Bridgelall, R., & Tolliver, D. D., Railroad accident analysis using extreme gradient boosting, Accident Analysis & Prevention, 156, 106126, 2021.
14. 14. Dong, Y., Han, C., Qu, S., Zhao, C., Madan, A., Fu, Y., Yang, Y., DAS-Accelerometer Data Fusion With Semi-Supervised Graph Variational Autoencoder for In-Service Train Wheel Flat Detection, IEEE Transactions on Intelligent Transportation Systems, 27 (1), 593-608, 2025.
15. 15. Fu, W., He, Q., Liu, S., Feng, Q., Gao, R., Detection of Railway Wheel-Flat Based on CBAM-Enhanced ResNet for Imbalanced Data, IEEE Sensors Journal, 2025.
16. 16. Kılıçarslan, S., Gögebakan, M., & Közkurt, C., Cervical cancer prediction using SMOTE algorithm and machine learning approaches, Journal of the Institute of Science and Technology, 13 (2), 747-759, 2023.
17. 17. Ghofrani, F., Sun, H., & He, Q., Analyzing risk of service failures in heavy haul rail lines: A hybrid approach for imbalanced data, Risk analysis, 42 (8), 1852-1871, 2022.
18. 18. Swana, E. F., Doorsamy, W., Bokoro, P., Tomek link and SMOTE approaches for machine fault classification with an imbalanced dataset, Sensors, 2 2(9), 3246, 2022.
19. 19. Azizoglu, G., Toprak, A. N., MELD3: Integrating multi-task ensemble learning for driver distraction detection, IEEE Access, 12, 186022-186034, 2024.
20. 20. Ou, J., Zhang, J., Li, H., Duan, B., Road damage prediction and intelligent maintenance methods based on stacking ensemble learning, Advanced Engineering Informatics, 66, 103466, 2025.
21. 21. Lundberg, S. M., Lee, S. I., A unified approach to interpreting model predictions, Advances in neural information processing systems, 30, 2017.
22. 22. Liu, D., Qin, Y., Zhao, Y., Yang, W., Hu, H., Yang, N., Liu, B., Fault early warning model for high-speed railway train based on feature contribution and causal inference, Sensors, 22 (23), 9184, 2022.
23. 23. García-Méndez, S., de Arriba-Pérez, F., Leal, F., Veloso, B., Malheiro, B., Burguillo-Rial, J. C., An explainable machine learning framework for railway predictive maintenance using data streams from the metro operator of Portugal. Scientific Reports, 15 (1), 27495, 2025.
24. 24. Zoralioğlu, Y., Gül, M. F., Azizoğlu, F., Azizoğlu, G., & Toprak, A. N., Predicting academic performance of students using machine learning techniques, In Innovations in Intelligent Systems and Applications Conference (ASYU), 1-6, 11-13 October, 2023.
25. 25. Chawla, N. V., Bowyer, K. W., Hall, L. O., Kegelmeyer, W. P., SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 16, 321-357, 2002.
26. 26. Alkan, A., Sevli, O., A classification study on predicting academic success using machine learning with oversampling and undersampling, Journal of the Faculty of Engineering and Architecture of Gazi University, 40 (4), 2191-2204, 2025.
27. 27. Tomek, I., Two Modifications of CNN, IEEE Transactions on Systems Man and Communications SMC-6, 769-772, 1976.
28. 28. He, H., Garcia, E. A., Learning from imbalanced data, IEEE Transactions on knowledge and data engineering, 21 (9), 1263-1284, 2009.
29. 29. Cox, D. R., The regression analysis of binary sequences. Journal of the Royal Statistical Society Series B: Statistical Methodology, 20 (2), 215-232, 1958.
30. 30. Breiman, L., Friedman, J., Olshen, R. A., Stone, C. J., Classification and regression trees, Chapman and Hall/CRC, 2017.
31. 31. Breiman, L., Random forests, Machine learning, 45 (1), 5-32, 2001.
32. 32. Cortes, C., Vapnik, V., Support-vector networks, Machine learning, 20 (3), 273-297, 1995.
33. 33. Baş, S., Kiliçarslan, S., Elen, A., Közkurt, C., Adaptive arctan kernel: a generalized kernel for support vector machine, Sādhanā, 49 (2), 120, 2024.
34. 34. Cover, T., Hart, P., Nearest neighbor pattern classification, IEEE transactions on information theory, 13 (1), 21-27, 1967.
35. 35. Chen, T., Guestrin, C., Xgboost: A scalable tree boosting system, In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785-794, 13 - 17 August, 2016.
36. 36. Fawcett, T., An introduction to ROC analysis, Pattern recognition letters, 27 (8), 861-874, 2006.
37. 37. Davis, J., & Goadrich, M., The relationship between Precision-Recall and ROC curves, In Proceedings of the 23rd international conference on Machine learning, 233-240, 2006.
38. 38. Matthews, B. W., Comparison of the predicted and observed secondary structure of T4 phage lysozyme, Biochimica et Biophysica Acta (BBA)-Protein Structure, 405 (2), 442-451, 1975.
39. 39. Cohen, J., A coefficient of agreement for nominal scales, Educational and psychological measurement, 20 (1), 37-46, 1960.
40. 40. Friedman, M., The use of ranks to avoid the assumption of normality implicit in the analysis of variance, Journal of the american statistical association, 32 (200), 675-701, 1937.
41. 41. Wilcoxon, F., Individual comparisons by ranking methods, Biometrics bulletin, 1 (6), 80-83, 1945.
42. 42. Holm, S., A simple sequentially rejective multiple test procedure, Scandinavian journal of statistics, 6 (2), 65-70, 1979.