Endüstriyel Uygulamalarda Piston Hata Tespitinde Hibrit Makine Öğrenimi Yaklaşımları

Öz

Bu çalışma, InceptionV3 mimarisi kullanılarak gerçekleştirilen derin özellik çıkarımı ile dört geleneksel makine öğrenimi sınıflandırıcısının (Gradient Boosting, Destek Vektör Makineleri, k-En Yakın Komşu ve Rastgele Orman) birleştirilmesiyle piston hatalarının tespiti için hibrit bir yaklaşım sunmaktadır. Modellerin performansı 10 katlı çapraz doğrulama yöntemiyle değerlendirilmiş ve Doğruluk (accuracy), kesinlik (precision), duyarlılık (recall) ve F1 Skoru açısından özetlenmiştir. Gradient Boosting, %95,31 doğruluk ve 0,9524 F1 Skoru ile en iyi performansı göstermiş, bunu Destek Vektör Makineleri izlemiştir. k-En Yakın Komşu ve Rastgele Orman ise biraz daha düşük ancak tutarlı sonuçlar vermiştir. Karışıklık matrisi analizleri, Gradient Boosting ve Destek Vektör Makineleri’nin dengeli sınıflandırma çıktıları sağladığını, k-En Yakın Komşu algoritmasının kesinliği ön plana çıkardığını, Rastgele Orman’ın ise nispeten daha yüksek duyarlılığı koruduğunu ortaya koymuştur. Sonuçlar, derin öğrenme tabanlı özellik çıkarımı ile klasik sınıflandırıcıları entegre eden hibrit yaklaşımların endüstriyel uygulamalarda hata tespitinde son derece etkili olduğunu ve hem doğruluk hem de sağlamlık sunduğunu doğrulamaktadır.

Anahtar Kelimeler

• inceptionV3
• Gradient Boosting
• Destek Vektör Makineleri
• k-En Yakın Komşu
• Rastgele Orman
• Piston

Hybrid machine learning approaches to piston defect detection in industrial applications

Öz

This study proposes a hybrid artificial intelligence approach for detecting surface defects in industrial piston components by combining deep learning based feature extraction with traditional machine learning classifiers. The experimental analysis was performed using the piston dataset, which includes both defected and perfect samples of industrial pistons. Four classification algorithms, namely Support Vector Machine, Artificial Neural Network, k Nearest Neighbors, and Random Forest, were implemented and compared based on their classification accuracy. The Support Vector Machine achieved the highest performance with an accuracy of 99.84%, demonstrating superior capability in distinguishing between defected and non-defected piston surfaces. The Artificial Neural Network followed closely with an accuracy of 99.69%, showing highly stable and consistent behavior. The k Nearest Neighbors model reached an accuracy of 98.75%, while the Random Forest achieved an accuracy of 94.84%, indicating a comparatively lower generalization performance. The results confirm that the hybrid combination of deep feature extraction and conventional classification methods significantly improves accuracy and robustness in defect detection. The proposed framework contributes to the industry 4.0 vision by providing a reliable, efficient, and intelligent quality control solution suitable for real-time manufacturing systems, supporting digital transformation in modern industrial environments.

Anahtar Kelimeler

• InceptionV3
• Gradient Boosting
• k-Nearest Neighbors
• Random Forest
• Piston
• Support Vector Machine

Kaynakça

1. C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, and Z. Wojna, "Rethinking the inception architecture for computer vision," Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2818-2826, 2016. doi: 10.1109/CVPR.2016.308.
2. X. Zhou et al., "Dense attention-guided cascaded network for salient object detection of strip steel surface defects," IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1-14, 2021. doi: 10.1109/TIM.2021.3132082.
3. M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, "Mobilenetv2: Inverted residuals and linear bottlenecks," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018. pp. 4510-4520, doi: 10.1109/CVPR.2018.00474.
4. D. Palma-Ramírez et al., "Deep convolutional neural network for weld defect classification in radiographic images," Heliyon, vol. 10, no. 9, 2024. doi: 10.1016/j.heliyon.2024.e30590.
5. S. Kumaresan, K. J. Aultrin, S. Kumar, and M. D. Anand, "Deep learning-based weld defect classification using VGG16 transfer learning adaptive fine-tuning," International Journal on Interactive Design and Manufacturing (IJIDeM), vol. 17, no. 6, pp. 2999-3010, 2023. doi: 10.1007/s12008-023-01327-3.
6. X. Li, W. Zhang, Q. Ding, and J.-Q. Sun, "Intelligent rotating machinery fault diagnosis based on deep learning using data augmentation," Journal of intelligent manufacturing, vol. 31, no. 2, pp. 433-452, 2020. doi: 10.1007/s10845-018-1456-1.
7. S. Woo, J. Park, J.-Y. Lee, and I. S. Kweon, "Cbam: Convolutional block attention module," Proceedings of the European conference on computer vision (ECCV), pp. 3-19, 2018. doi: 10.1007/978-3-030-01234-2_1.
8. Dosovitskiy et al., "An image is worth 16x16 words: Transformers for image recognition at scale," arXiv preprint arXiv:2010.11929, 2020. Available: https://openreview.net/forum?id=YicbFdNTTy.

9. M. M. Abagiu, D. Cojocaru, F. Manta, and A. Mariniuc, "Detecting machining defects inside engine piston chamber with computer vision and machine learning," Sensors, vol. 23, no. 2, p. 785, 2023. doi: 10.3390/s23020785.
10. M. S. Shams. "Piston Image Dataset." Kaggle. https://www.kaggle.com/datasets/mdsalmanshams/piston-image-dataset (accessed., 12,May, 2025)
11. O. Kilci and M. Koklu, "Machine Learning-Based Detection of Solar Panel Surface Defects Using Deep Features from InceptionV3," 2025.
12. Y. S. Taspinar, M. Koklu, and M. Altin, "Fire detection in images using framework based on image processing, motion detection and convolutional neural network," International Journal of Intelligent Systems and Applications in Engineering, 2021.
13. F. Chollet, "Xception: Deep learning with depthwise separable convolutions," in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1251-1258, 2017.
14. Lin, L. Li, W. Luo, K. C. Wang, and J. Guo, "Transfer learning based traffic sign recognition using inception-v3 model," Periodica Polytechnica Transportation Engineering, vol. 47, no. 3, pp. 242-250, 2019. doi: 10.3311/PPtr.11480.
15. T. A. Cengel, B. Gencturk, E. T. Yasin, M. B. Yildiz, I. Cinar, and M. Koklu, "Apple (malus domestica) quality evaluation based on analysis of features using machine learning techniques," Applied Fruit Science, vol. 66, no. 6, pp. 2123-2133, 2024. doi: 10.1007/s10341-024-01196-4.
16. I. Wahlang et al., "Brain magnetic resonance imaging classification using deep learning architectures with gender and age," Sensors, vol. 22, no. 5, p. 1766, 2022. doi: 10.3390/s22051766.
17. P. Liu, K.-K. R. Choo, L. Wang, and F. Huang, "SVM or deep learning? A comparative study on remote sensing image classification," Soft Computing, vol. 21, no. 23, pp. 7053-7065, 2017. doi: 10.1007/s00500-016-2247-2.
18. I. Cinar, Y. S. Taspinar, M. M. Saritas, and M. Koklu, "Feature extraction and recognition on traffic sign images," Selcuk University Journal of Engineering Sciences, vol. 19, no. 4, pp. 282-292, 2020.
19. S. K. S. Al-doori, Y. S. Taspinar, and M. Koklu, "Distracted driving detection with machine learning methods by cnn based feature extraction," International Journal of Applied Mathematics Electronics and Computers, vol. 9, no. 4, pp. 116-121, 2021.
20. M. M. Saritas and M. Koklu, "Classification Of Cauliflower Leaf Diseases Using Features Extracted From Squeezenet With Decision Tree And Random Forest," 2024.
21. Y. Kong and T. Yu, "A deep neural network model using random forest to extract feature representation for gene expression data classification," Scientific reports, vol. 8, no. 1, p. 16477, 2018. doi: 10.1038/s41598-018-34833-6.
22. I. Cinar, Y. S. Taspinar, and M. Koklu, "Development of early stage diabetes prediction model based on stacking approach," Tehnički glasnik, vol. 17, no. 2, pp. 153-159, 2023. doi: 10.31803/tg-20211119133806.
23. Jamma, O. Ahmed, S. Areibi, G. Grewal, and N. Molloy, "Design exploration of ASIP architectures for the K-nearest neighbor machine-learning algorithm," in 2016 28th international conference on microelectronics (ICM), 2016. IEEE, pp. 57-60, doi: 10.1109/ICM.2016.7847907.
24. M. Koklu and K. Sabanci, "Estimation of credit card customers payment status by using kNN and MLP," International Journal of Intelligent Systems and Applications in Engineering, vol. 4, no. Special Issue-1, pp. 249-251, 2016.
25. O. Kilci and M. Koklu, "Guava Fruit Disease Classification Using Deep Learning and Machine Learning Models," Research in Agricultural Sciences, vol. 56, no. 3, pp. 217-226, 2025. doi: 10.17097/agricultureatauni.1665941.
26. S. Smys, J. I. Z. Chen, and S. Shakya, "Survey on neural network architectures with deep learning," Journal of Soft Computing Paradigm (JSCP), vol. 2, no. 03, pp. 186-194, 2020. doi: 10.36548/jscp.2020.3.007
27. A. Yasar, I. Saritas, M. Sahman, and A. Cinar, "Classification of Parkinson disease data with artificial neural networks," in IOP conference series: materials science and engineering, 2019. vol. 675, no. 1: IOP Publishing, p. 012031, doi: 10.1088/1757-899X/675/1/012031.
28. O. Kilci and M. Koklu, "Classification Of Guava Diseases Using Features Extracted From Squeezenet With Adaboost And Gradient Boosting," 2024.
29. A. Golcuk and A. Yasar, "Classification of bread wheat genotypes by machine learning algorithms," Journal of Food Composition and Analysis, vol. 119, p. 105253, 2023. doi: 10.1016/j.jfca.2023.105253.
30. O. Kilci and M. Koklu, "Automated Classification of Biscuit Quality Using YOLOv8 Models in Food Industry," Food Analytical Methods, pp. 1-15, 2025. doi: 10.1007/s12161-025-02755-5.
31. Gholamiangonabadi, N. Kiselov, and K. Grolinger, "Deep neural networks for human activity recognition with wearable sensors: Leave-one-subject-out cross-validation for model selection," Ieee Access, vol. 8, pp. 133982-133994, 2020. doi: 10.1109/ACCESS.2020.3010715.
32. Y. S. Taspinar, I. Cinar, R. Kursun, and M. Koklu, "Monkeypox Skin Lesion Detection with Deep Learning Models and Development of Its Mobile Application," Public health, vol. 500, p. 5, 2024.
33. O. Kilci, M. B. Yildiz, and H. Y. Tukel, Deep Learning and Machine Learning Approaches for Coffee Leaf Disease Diagnosis, p. 273, 2024.
34. M. Koklu, H. Kahramanli, and N. Allahverdi, "A new accurate and efficient approach to extract classification rules," 2014.
35. O. Kilci, Y. Eryesil, and M. Koklu, "Classification of Biscuit Quality with Deep Learning Algorithms," Journal of Food Science, vol. 90, no. 7, p. e70379, 2025. doi: 10.1111/1750-3841.70379.