Rüzgar Türbini Kanat Arızalarının Tespiti için Akıllı Bir Yaklaşım

Abstract

Rüzgâr türbini kanatları, sürekli mekanik gerilime maruz kalmaları ve zorlu çevresel koşullarda çalışmaları nedeniyle yenilenebilir enerji sistemlerinde en fazla arıza riski taşıyan bileşenler arasındadır. Kanat arızalarının erken ve doğru bir şekilde tespiti, bakım maliyetlerinin azaltılması ve rüzgâr enerjisi üretiminde güvenliğin ve verimliliğin sağlanması açısından kritik öneme sahiptir. Geleneksel yöntemler iç yapısal hasarların tespitinde yetersiz kalırken, titreşim tabanlı durum izleme sistemleri ve derin öğrenmedeki son gelişmeler arıza teşhisi için yeni olanaklar sunmaktadır. Bu çalışma, rüzgâr türbini kanat arızalarının otomatik tespiti için zaman-frekans analizi ve derin öğrenme tabanlı görüntü sınıflandırmayı birleştiren yenilikçi bir yöntem önermektedir. Açık erişimli bir veri kümesinden elde edilen titreşim sinyalleri, yerel özelliklerin korunması için zaman pencerelerine bölünmekte ve daha sonra sinyallerin zaman-frekans özelliklerini görselleştirmek amacıyla Güç Spektral Yoğunluğu (PSD) görüntülerine dönüştürülmektedir. Elde edilen görüntüler, hiyerarşik özellik çıkarımı için ConvNeXt ve MaxViT gibi gelişmiş derin öğrenme mimaralarına aktarılmakta ve bu derin özellikler Destek Vektör Makineleri (SVM) kullanılarak belirli arıza sınıflarına ayrılmaktadır. Önerilen yöntem, 10 katlı çapraz doğrulama ile test edilmiş ve farklı ölçütlerde tutarlı performans sergileyerek etkinliğini ve güvenilirliğini kanıtlamıştır. Ortalama olarak model, %81,38 doğruluk (accuracy) ve %81,80 kesinlik (precision) değerine ulaşmış; bu sonuçlarla uyumlu bir F1 skoru elde etmiştir. Ayrıca ortalama Matthews Korelasyon Katsayısı (MCC) 0,7226 ve Cohen’s Kappa değeri 0,7208 olarak hesaplanmış, bu da tahmin edilen etiketler ile gerçek etiketler arasında yüksek düzeyde bir uyumu göstermektedir. Bu çerçeve, titreşim verilerine dayalı, tamamen otomatik ve ölçeklenebilir bir arıza tespit sistemi sunarak rüzgâr enerjisi sistemlerinde akıllı arıza teşhisi alanına önemli bir katkı sağlamaktadır.

Keywords

• Rüzgâr Türbini Arızaları
• MaxViT
• ConvNeXt
• Çoklu Sınıflandırma
• Sinyal Işleme
• Destek Vektör Makineleri

An Intelligent Framework for Wind Turbine Blade Fault Detection

Abstract

Wind turbine blades are among the most failure-prone components in renewable energy systems due to their constant mechanical stress and exposure to harsh environmental conditions. Early and accurate detection of blade failures is critical to reducing maintenance costs and ensuring safety and efficiency in wind energy production. While traditional methods are insufficient in detecting internal structural damage, recent advances in vibration-based condition monitoring systems and deep learning offer new opportunities for fault diagnosis. This study proposes an innovative method combining time-frequency analysis and deep learning-based image classification for the automatic detection of wind turbine blade faults. Vibration signals obtained from an open access dataset are divided into time windows to preserve local features and then converted into Power Spectral Density (PSD) images to visualize the time-frequency characteristics of the signals. The resulting images are fed to advanced deep learning architectures such as ConvNeXt and MaxViT for hierarchical feature extraction, and these deep features are classified into specific fault classes using Support Vector Machines (SVM). The proposed method has been tested with 10-fold cross-validation and has proven its effectiveness and reliability by demonstrating consistent performance across multiple metrics. On average, the model achieved an accuracy of 81.38% and a precision of 81.80%, while maintaining balanced classification performance with an F1 score consistent with these values. Furthermore, it obtained an average Matthews Correlation Coefficient (MCC) of 0.7226 and a Cohen’s Kappa value of 0.7208, indicating substantial agreement between predicted and actual labels. This framework contributes to the field of intelligent fault diagnosis in wind energy systems by providing a fully automatic and scalable fault detection system based on vibration data.

Keywords

• Wind Turbine Faults
• MaxViT
• ConvNeXt
• Multi-Classification
• Signal Processing
• Support Vector Machines

References

1. Liu H, Wang Y, Zeng T, Wang H, Chan SC, Ran L. Wind turbine generator failure analysis and fault diagnosis: A review. IET Renewable Power Generation 2024;18(15):3127-3148. doi:10.1049/rpg2.13104.
2. RHassan IU, Panduru K, Walsh J. An in-depth study of vibration sensors for condition monitoring. Sensors 2024;24(3):740. doi:10.3390/s24030740.
3. Ogaili AAF, Hamzah MN, Jaber AA. Enhanced fault detection of wind turbine using extreme gradient boosting technique based on nonstationary vibration analysis. Journal of Failure Analysis and Prevention 2024;24(2):877-895. doi:10.1007/s11668-024-01894-x.
4. Yang C, Liu X, Zhou H, Ke Y, See J. Towards accurate image stitching for drone-based wind turbine blade inspection. Renewable Energy 2023;203:267-279. doi:10.1016/j.renene.2022.12.063.
5. Leon-Medina JX, Anaya M, Parés N, Tibaduiza DA, Pozo F. Structural damage classification in a Jacket-type wind-turbine foundation using principal component analysis and extreme gradient boosting. Sensors 2021;21(8):2748. doi:10.3390/s21082748.
6. Rahimilarki R, Gao Z, Jin N, Zhang A. Convolutional neural network fault classification based on time-series analysis for benchmark wind turbine machine. Renewable Energy 2022;185:916-931. doi:10.1016/j.renene.2021.12.056.
7. Sethi MR, Subba AB, Faisal M, Sahoo S, Raju DK. Fault diagnosis of wind turbine blades with continuous wavelet transform based deep learning model using vibration signal. Engineering Applications of Artificial Intelligence 2024;138:109372. doi:10.1016/j.engappai.2024.109372.
8. Aranizadeh A, Shad H, Vahidi B, Khorsandi A. A novel small-scale wind-turbine blade failure detection according to monitored-data. Results in Engineering 2025;25:103809. doi:10.1016/j.rineng.2024.103809.

9. Ogaili AAF, Jaber AA, Hamzah MN. A methodological approach for detecting multiple faults in wind turbine blades based on vibration signals and machine learning. Curved and Layered Structures 2023;10(1):20220214.
10. Yang Y, Lyu Y, Li Y, Fang L, Luo Y, Liu W. A two-stage data-driven method to identify blade faults for wind turbines using vibration data. In: 7th International Conference on Energy, Electrical and Power Engineering (CEEPE); 2024.
11. Ali W, El-Thalji I, Giljarhus KET, Delimitis A. Classification analytics for wind turbine blade faults: integrated signal analysis and machine learning approach. Energies 2024;17(23):5856. doi:10.3390/en17235856.
12. Tuerxun W, Xu C, Haderbieke M, Guo L, Cheng Z. A wind turbine fault classification model using broad learning system optimized by improved pelican optimization algorithm. Machines 2022;10(5):407. doi:10.3390/machines10050407.
13. Li H, Wang Z. Anomaly identification of wind turbine blades based on Mel-Spectrogram Difference feature of aerodynamic noise. Measurement 2025;240:115428. doi:10.1016/j.measurement.2024.115428.
14. Lee CY, Maceren EDC. Wind energy system fault classification using deep CNN and improved PSO-tuned extreme gradient boosting. IET Renewable Power Generation 2024;18(14):2496-2511. doi:10.1049/rpg2.13091.
15. Zhang J, Cosma G, Watkins J. Image enhanced mask R-CNN: A deep learning pipeline with new evaluation measures for wind turbine blade defect detection and classification. Journal of Imaging 2021;7(3):4. doi:10.3390/jimaging7030046.
16. Ruiz M, Mujica LE, Alférez S, Acho L, Tutivén C, Vidal Y, et al. Wind turbine fault detection and classification by means of image texture analysis. Mechanical Systems and Signal Processing 2018;107:149-167. doi:10.1016/j.ymssp.2017.12.035.
17. Reddy A, Indragandhi V, Ravi L, Subramaniyaswamy V. Detection of Cracks and damage in wind turbine blades using artificial intelligence-based image analytics. Measurement 2019;147:106823. doi:10.1016/j.measurement.2019.07.051.
18. Ogaili AAF, Jaber AA, Hamzah MN. Wind turbine blades fault diagnosis based on vibration dataset analysis. Data in Brief 2023;49:109414. doi:10.1016/j.dib.2023.109414.
19. Okumuş H, Ergün E. Development Of A Ternary Levels Emotion Classification Model Utilizing Electroencephalography Data Set. Konya Journal of Engineering Sciences 2025;13(2):607-623. doi:10.36306/konjes.1649691.
20. Okumus H, Ergun E. Broken Rotor Bar Fault Detection in Induction Motors through Power Spectral Density to Image Method. Sigma J Eng Nat Sci 2025. doi:10.14744/sigma.2024.00153.
21. Liu Z, Mao H, Wu CY, Feichtenhofer C, Darrell T, Xie S. A convnet for the 2020s. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition; 2022.
22. Tu Z, Talebi H, Zhang H, Yang F, Milanfar P, Bovik A, Li Y. Maxvit: Multi-axis vision transformer. In: European conference on computer vision; 2022.
23. Ergün E. Artificial intelligence approaches for accurate assessment of insulator cleanliness in high-voltage electrical systems. Electrical Engineering 2025;107(3):2969-2982. doi:10.1007/s00202-024-02691-3.
24. Ramu K, Patthi S, Prajapati YN, Ramesh JVN, Banerjee S, Rao KB, Alzahrani SI. Hybrid CNN-SVM model for enhanced early detection of Chronic kidney disease. Biomedical Signal Processing and Control 2025;100:107084. doi:10.1016/j.bspc.2024.107084.