Fırçalı DA Motorlarda Açık Veri Seti İle Makine Öğrenmesi Tabanlı Kestirimci Bakım ve Arıza Teşhisi

Öz

Fırçalı Doğru Akım (FDA) motorları, endüstriyel otomasyon sistemlerinde ve tahrik mekanizmalarında yaygın olarak kullanılmakta olup, fırça aşınması ve sargı deformasyonları beklenmedik duruşlara ve yüksek bakım maliyetlerine neden olabilmektedir. Bu çalışmada, fırçalı DA motorların durum teşhisi ve Kalan Faydalı Ömür (Remaining Useful Life, RUL) tahmini için veri odaklı, ağaç tabanlı toplu öğrenme yöntemleri önerilmiştir. F.A.I.R veri setinden çoklu sensörlerden elde edilen akım, gerilim, devir, motor-ortam sıcaklık farkı ve spektral titreşim öznitelikleri kullanılarak motor ömrü; Evre I (Sağlıklı), Evre II (Hafif Aşınma), Evre III (Orta Aşınma) ve Evre IV (Kritik/Arıza) olmak üzere dört sağlık evresine ayrılmıştır. Durum teşhisinde Random Forest algoritması %98,7 test doğruluğu ve %95,70 10 katlı çapraz doğrulama başarısı ile Extra Trees algoritmasına göre daha yüksek performans göstermiştir. Öznitelik önem analizi, motor sağlığını en güçlü şekilde temsil eden değişkenlerin net sıcaklık artışı, yüksek frekanslı titreşim bileşenleri (16–26 kHz) ve motor devri olduğunu ortaya koymuştur. RUL tahmininde ise Extra Trees Regressor modeli 0,9666 belirlenme katsayısı (R²) ve 21,49 zaman adımı RMSE değeri ile en başarılı sonucu elde ederken, Random Forest Regressor modeli 0,9603 R² ve 23,43 zaman adımı RMSE değerleriyle benzer düzeyde yüksek performans sergilemiştir. Elde edilen sonuçlar, önerilen yaklaşımın FDA motorlarda durum izleme ve öngörücü bakım uygulamaları için güvenilir ve etkili bir karar destek sistemi olarak kullanılabileceğini göstermektedir.

Anahtar Kelimeler

• Fırçalı DA motor
• Kalan faydalı ömür (RUL)
• Durum teşhisi
• Makine öğrenmesi
• Random forest
• Extra trees

Etik Beyan

Bu araştırmada hayvanlar ve insanlar üzerinde herhangi bir çalışma yapılmadığı için etik kurul onayı alınmamıştır.

Predictive Maintenance and Fault Diagnosis Based on Machine Learning with Open Data Set in Brushed DC Motors

Öz

Brushed DC motors are widely used in industrial automation systems and drive mechanisms, but brush wear and winding deformations can lead to unexpected shutdowns and high maintenance costs. This study proposes data-driven, tree-based ensemble learning methods for condition diagnosis and Remaining Useful Life (RUL) estimation of brushed DC motors. Using current, voltage, speed, motor-ambient temperature difference, and spectral vibration features obtained from multiple sensors in the F.A.I.R dataset, motor life was divided into four health stages: Stage I (Healthy), Stage II (Slight Wear), Stage III (Moderate Wear), and Stage IV (Critical/Failure). In condition diagnosis, the Random Forest algorithm showed higher performance than the Extra Trees algorithm with 98.7% test accuracy and 95.70% 10-fold cross-validation success. Feature significance analysis revealed that the variables most strongly representing motor health were net temperature rise, high-frequency vibration components (16–26 kHz), and motor speed. In RUL prediction, the Extra Trees Regressor model achieved the most successful result with a coefficient of determination (R²) of 0.9666 and a time-step RMSE value of 21.49, while the Random Forest Regressor model showed similarly high performance with an R² of 0.9603 and a time-step RMSE value of 23.43. The results indicate that the proposed approach can be used as a reliable and effective decision support system for condition monitoring and predictive maintenance applications in brushed DC motors.

Anahtar Kelimeler

• Brushed DC motor
• Remaining useful life (RUL)
• Condition diagnosis
• Machine learning
• Random forest
• Extra trees

Kaynakça

1. Barkas, D. A., Ioannidis, G. Ch., Psomopoulos, C. S., Kaminaris, D., & Vokas, G. A. (2020). Brushed DC Motor Drives for Industrial and Automobile Applications with Emphasis on Control Techniques: A Comprehensive Review. Electronics, 9(6), 887-887. https://doi.org/10.3390/electronics9060887
2. Bergstra, J., & Bengio, Y. (2011). Algorithms for hyper-parameter optimization. HAL (Le Centre pour la Communication Scientifique Directe). Centre National de la Recherche Scientifique. http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.476.5930
3. Breiman, L. (2001). Random Forests. Machine Learning, 45(1), 5-32. https://doi.org/10.1023/a:1010933404324
4. Cros, J., Sincero, G. C. R., & Viarouge, P. (2009). Design method for brush permanent magnet DA motors. 2009 IEEE International Electric Machines and Drives Conference, 1625-1632.
5. Ellefsen, A. L., Ushakov, S., Æsøy, V., & Zhang, H. (2019). Validation of Data-Driven Labeling Approaches Using a Novel Deep Network Structure for Remaining Useful Life Predictions. IEEE Access, 7, 71563-71575. https://doi.org/10.1109/access.2019.2920297
6. Filina, O., & Tsvetkov, A. N. (2019). Evaluation of the operational life of direct current motors. IOP Conference Series Materials Science and Engineering, 489, 12-16. https://doi.org/10.1088/1757-899x/489/1/012016
7. Filina, O., Martyushev, N. V., Malozyomov, B. V., Tynchenko, V. S., Kukartsev, V. A., Bashmur, K. A., Pavlov, P. P., & Panfilova, T. (2023). Increasing the efficiency of diagnostics in the brush-commutator assembly of a direct current electric motor. Energies, 17(1), 17-17. https://doi.org/10.3390/en17010017
8. Geurts, P., Ernst, D., & Wehenkel, L. (2006). Extremely randomized trees. Machine Learning, 63(1), 3-42. https://doi.org/10.1007/s10994-006-6226-1

9. Guyon, I., & Elisseeff, A. (2003). An introduction to variable and feature selection. Journal of machine learning research, 3(Mar), 1157-1182.
10. Hastie, T., Tibshirani, R., & Friedman, J. H. (2009). The Elements of Statistical Learning. Içinde Springer series in statistics. Springer Science+Business Media. https://doi.org/10.1007/978-0-387-84858-7
11. Ippolito, P. P. (2022). Hyperparameter tuning: the art of fine-tuning machine and deep learning models to improve metric results. İçinde Applied data science in tourism: Interdisciplinary approaches, methodologies, and applications (ss. 231-251). Springer International Publishing.
12. Jardine, A., Lin, D., & Banjević, D. (2005). A review on machinery diagnostics and prognostics implementing condition-based maintenance. Mechanical Systems and Signal Processing, 20(7), 1483-1510. https://doi.org/10.1016/j.ymssp.2005.09.012
13. Karakılıç, M., Hataş, H., & Pacal, İ. (2025, May). Open-Circuit Fault Detection in T-Type MLI Using XGBoost: A Machine Learning-Based Approach. In 2025 7th International Congress on Human-Computer Interaction, Optimization and Robotic Applications (ICHORA) (pp. 1-6). IEEE.
14. Lei, Y., Li, N., Guo, L., Li, N., Yan, T., & Lin, J. (2018). Machinery health prognostics: A systematic review from data acquisition to RUL prediction. Mechanical Systems and Signal Processing, 104, 799-834. https://doi.org/10.1016/j.ymssp.2017.11.016
15. Mahlich, C., Vente, T., & Beel, J. (2025). From theory to practice: Implementing and evaluating e-fold cross-validation. In J. Liu, A. Ghosh, C. W. Tan, & H. Zhao (Eds.), International Conference on Artificial Intelligence and Machine Learning Research (CAIMLR 2024) (Vol. 13635, Article 1363505). SPIE. https://doi.org/10.1117/12.3058568
16. Mohapatra, A. (2024). Generative AI for Predictive Maintenance: Predicting Equipment Failures and Optimizing Maintenance Schedules Using AI. International Journal of Scientific Research and Management (IJSRM), 12(11), 1648-1672. https://doi.org/10.18535/ijsrm/v12i11.ec03
17. Pham, V.-H., Le, D. B., & Nguyen, Pt. D. (2023). A Method of Assessing the Service Life and Reliability of Motor Brushes Based on Wear Under Variable Current and Environmental Conditions. Tribology in Industry, 45(1), 302-312. https://doi.org/10.24874/ti.1443.02.23.06
18. Rajabioun, R., Hataş, H., Afshar, M., Atan, Ö., & Akin, B. (2025). Optimal feature selection for distributed bearing fault detection using cost-effective, high-performance machine learning algorithms. IEEE Journal of Emerging and Selected Topics in Industrial Electronics, 7(1), 46-57.
19. Randall, R. B. (2021). Vibration-based condition monitoring: industrial, automotive and aerospace applications. John Wiley & Sons.
20. Reñones, A. (2020). F.A.I.R. open dataset of brushed DC motor faults for testing of AI algorithms (Version 1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.4314249
21. Teel-Jongebloed, C. C. (2020). Detection of permanent magnet DC motor failure due to brush wear using parameter estimation and statistical analysis. OakTrust (Texas A&M University Libraries). https://hdl.handle.net/1969.1/192653
22. Zheng, A., & Casari, A. (2018). Feature engineering for machine learning: principles and techniques for data scientists. O’Reilly Media, Inc