Comparison between image based and tabular data-based inclusion class categorization
-
S. R. Babu
Postdoctoral researcher at the Christian Doppler Laboratory for Inclusion Metallurgy in Advanced Steelmaking at Montanuniversität Loeben, Austria. Completed Ph.D. in Materials Engineering at University of Oulu, Finland and Masters in Material Science and Simulation at the Ruhr Universität Bochum.
Student assistant at the Christian Doppler Laboratory for Inclusion Metallurgy in Advanced Steelmaking at Montanuniversität Loeben, Austria. Recently completed Master’s degree in Metallurgy at Montanuniversität Leoben, Austria.
Abstract
Non-metallic inclusions (NMI) have a significant impact on the final properties of steel products. As of today, the scanning electron microscope equipped with energy-dispersive spectroscopy (SEM-EDS) serves as the state of art characterization tool to study NMIs in steel. The automated 2D analysis method with the SEM-EDS allows for a comprehensive analysis of all the inclusions observed within a selected area of the sample. The drawback of this method is the time taken to complete the analysis. Therefore, machine learning methods have been introduced which can potentially replace the usage of EDS for obtaining chemical information of the inclusion by making quick categorizations of the inclusion classes and types. The machine learning methods can be developed by either training it directly with labeled backscattered electron (BSE) images or by tabular data consisting of image features input such as morphology and mean gray value obtained from the BSE images. The current paper compares both these methods using two steel grades. The advantages and the disadvantages have been documented. The paper will also compare the usage of shallow and deep learning methods to classify the steels and discuss the outlook of the existing machine learning methods to efficiently categorize the NMIs in steel.
Kurzfassung
Nicht-metallische Einschlüsse (NMEs) haben einen starken Einfluss auf die finalen Eigenschaften von Stahlerzeugnissen. Aktuell ist das Rasterelektronenmikroskop (REM-EDX) mit energiedispersiver Spektroskopie in der Untersuchung von NMEs in Stahl Stand der Technik. Mit dem automatisierten 2D-Analyseverfahren mittels REM-EDX lassen sich alle in einem ausgewählten Bereich der Probe beobachteten Einschlüsse umfassend analysieren. Nachteil dieses Verfahrens ist der für die Ausführung der Analyse erforderliche Zeitaufwand. Daher wurden auf maschinellem Lernen (Machine Learning) basierenden Verfahren eingesetzt, die mit ihrer schnellen Kategorisierung von Einschlussklassen und -arten die EDX zur Beschaffung chemischer Informationen über den Einschluss möglicherweise ersetzen können. Machine Learning-Verfahren können entweder durch direktes Trainieren mit gelabelten (gekennzeichneten) Rückstreuelektronen-(RE)-Bildern oder mittels tabellarischer Daten entwickelt werden, die aus den RE-Bildern gewonnene Informationen zu Bildmerkmalen wie beispielsweise die Morphologie und den mittleren Grauwert enthalten. Die beiden Verfahren werden im vorliegenden Beitrag am Beispiel zweier Stahlsorten verglichen. Vorteile und Nachteile wurden dokumentiert. Im Beitrag werden auch der Einsatz von Shallow und Deep Learning-Verfahren zur Klassifizierung der Stähle verglichen und Perspektiven der bestehenden Machine Learning-Verfahren für eine effiziente Kategorisierung der NMEs in Stahl erörtert.
About the authors
Postdoctoral researcher at the Christian Doppler Laboratory for Inclusion Metallurgy in Advanced Steelmaking at Montanuniversität Loeben, Austria. Completed Ph.D. in Materials Engineering at University of Oulu, Finland and Masters in Material Science and Simulation at the Ruhr Universität Bochum.
Student assistant at the Christian Doppler Laboratory for Inclusion Metallurgy in Advanced Steelmaking at Montanuniversität Loeben, Austria. Recently completed Master’s degree in Metallurgy at Montanuniversität Leoben, Austria.
5 Acknowledgements
The authors would like to gratefully acknowledge the financial support provided by the Austrian Federal Ministry of Labor and Economy, the National Foundation for Research, Technology and Development, and the Christian Doppler Research Association.
5 Danksagung
Die Autoren bedanken sich beim österreichischen Bundesministerium für Wirtschaft und Arbeit, der Nationalstiftung für Forschung, Technologie und Entwicklung und der Christian Doppler-Forschungsgesellschaft für die finanzielle Unterstützung.
References / Literatur
[1] Ramesh Babu, S.; Musi, R.; Thiele, K.; Michelic, S. K.: Classification of Nonmetallic Inclusions in Steel by Data-Driven Machine Learning Methods, Steel Res. Int. 94 (2023) 1, 2200617. DOI: 10.1002/srin.20220061710.1002/srin.202200617Search in Google Scholar
[2] Ramesh Babu, S.; Michelic, S. K.: Analysis of Non-Metallic Inclusions by Means of Chemical and Electrolytic Extraction – A Review, Materials 15 (2022) 9. DOI: 10.3390/ma1509336710.3390/ma15093367Search in Google Scholar PubMed PubMed Central
[3] Zhuang, C.; Liu, J.; Mi, Z.; Jiang, H.; Di Tang; Wang, G.: Non-Metallic Inclusions in TWIP Steel, Steel Res. Int. 85 (2014) 10, 1432–1439. DOI: 10.1002/srin.20130035410.1002/srin.201300354Search in Google Scholar
[4] Michelic, S. K.; Bernhard, C.: Michelic, S. K.; Bernhard, C.: Steel Res. Int. 93 (2022) 7, 2200086., Steel Res. Int. 93 (2022) 7, 2200086. DOI: 10.1002/srin.20220008610.1002/srin.202200086Search in Google Scholar
[5] Mapelli, C.: Non-metallic inclusions and clean steel la metallurgia italiana 100 (2008) 6, 43–52.Search in Google Scholar
[6] Kaushik, P.; Pielet, H.; Yin, H.: nclusion characterisation–tool for measurement of steel cleanliness and process control: Part 1, Ironmaking and Steelmaking 36 (2009) 8, 561–571. DOI: 10.1179/030192309X1249291093813110.1179/030192309X12492910938131Search in Google Scholar
[7] Kaushik, P.; Pielet, H.; Yin, H.: Inclusion characterisation-tool for measurement of steel cleanliness and process control: Part 2, Ironmaking and Steel-making 36 (2009) 8, 572–582. DOI: 10.1179/030192309X1249291093817710.1179/030192309X12492910938177Search in Google Scholar
[8] Kaushik, P.; Lowry, M.; Yin, H.; Pielet, H.: Inclusion characterisation for clean steelmaking and quality control, Ironmaking & Steelmaking 39 (2013) 4, 284–300. DOI: 10.1179/1743281211Y.000000006910.1179/1743281211Y.0000000069Search in Google Scholar
[9] Abdulsalam, M.; Gao, N.; Webler, B. A.; Holm, E. A.: Prediction of Inclusion Types From BSE Images: RF vs. CNN, Front. Mater. 8 (2021), 754089. DOI: 10.3389/fmats.2021.75408910.3389/fmats.2021.754089Search in Google Scholar
[10] Abdulsalam, M.; Zhang, T.; Tan, J.; Webler, B. A.: Automated Classification and Analysis of Nonmetallic Inclusion Data Sets, Metall. and Materi. Trans. B 49 (2018) 4, 1568–1579. DOI: 10.1007/s11663-018-1276-x10.1007/s11663-018-1276-xSearch in Google Scholar
[11] Burkov, A.: The hundred-page machine learning book, 2019.Search in Google Scholar
[12] Torlay, L.; Perrone-Bertolotti, M.; Thomas, E.; Baciu, M.: Machine learning-XGBoost analysis of language networks to classify patients with epilepsy, Brain Inf. 4 (2017) 3, 159–169. DOI: 10.1007/s40708-017-0065-710.1007/s40708-017-0065-7Search in Google Scholar PubMed PubMed Central
[13] Chen, T.; Guestrin, C. (Eds.): XGBoost: A Scalable Tree Boosting System in: Proc. of the 22nd acm sigkdd Int. Conf. Knowledge Discovery and Data mining, ACM, 2016. DOI: 10.1145/2939672.293978510.1145/2939672.2939785Search in Google Scholar
[14] Zhang, D.; Qian, L.; Mao, B.; Huang, C.; Huang, B.; Si, Y.: A data-driven design for fault detection of wind turbines using random forests and XGboost, IEEE Access 6 (2018), 21020–21031. DOI: 10.1109/ACCESS.2018.281867810.1109/ACCESS.2018.2818678Search in Google Scholar
[15] Cheng, F.; Yang, C.; Zhou, C.; Lan, L.; Zhu, H.; Li, Y.: Simultaneous determination of metal ions in zinc sulfate solution using UV-Vis spectrometry and SPSE-XGBoost method, Sensors (Basel, Switzerland) 20 (2020) 17. DOI: 10.3390/s2017493610.3390/s20174936Search in Google Scholar PubMed PubMed Central
[16] Pedregosa, F.; Varoquaux, G.; Gramfort, A.; Michel, V.; Thirion, B.; Grisel, O.; Blondel, M.; Prettenhofer, P.; Weiss, R.; Dubourg, V.; Vanderplas, J.; Passos, A.; Cournapeau, D.; Brucher, M.; Perrot, M.; Duchesnay, E.: Scikit-learn: Machine Learning in Python, Journal of Machine Learning Research 12 (2011), 2825–2830.Search in Google Scholar
[17] James, G.; Witten, D.; Hastie, T.; Tibshirani, R.: An Introduction to Statistical Learning, 2. Aufl., Springer US, New York, NY, 2021. DOI: 10.1007/978-1-4614-7138-710.1007/978-1-4614-7138-7Search in Google Scholar
[18] Caelen, O.: A Bayesian interpretation of the confusion matrix, Ann Math Artif Intell 81 (2017) 3–4, 429–450. DOI: 10.1007/s10472-017-9564-810.1007/s10472-017-9564-8Search in Google Scholar
[19] Powers, D. M. W.: Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation, arXiv preprint arXiv:2010.16061, 2011. DOI: 10.48550/arXiv.2010.1606110.48550/arXiv.2010.16061Search in Google Scholar
[20] Davis, J.; Goadrich, M.: The relationship between Precision-Recall and ROC curves, in: Proc 23rd Int. Conf. Machine Learning. – ICML ‘06. ACM Press. New York, New York, USA, 2006. doi.org/10.1145/1143844.1143874doi.org/10.1145/1143844.1143874Search in Google Scholar
[21] Erfani, S. M.; Rajasegarar, S.; Karunasekera, S.; Leckie, C.: High-Dimensional and Large-Scale Anomaly Detection using a Linear One-Class SVM with Deep Learning , Pattern Recognit 58 (2016), 121–134. DOI: 10.1016/j.patcog.2016.03.02810.1016/j.patcog.2016.03.028Search in Google Scholar
[22] M. Erfani, S.; Baktashmotlagh, M.; Rajasegarar, S.; Karunasekera, S.; Leckie, C.: R1SVM: A Randomised Nonlinear Approach to Large-Scale Anomaly Detection, AAAI 29 (2015) 1. DOI: 10.1609/aaai.v29i1.920810.1609/aaai.v29i1.9208Search in Google Scholar
[23] Liu, D.; Qian, H.; Dai, G.; Zhang, Z.: An iterative SVM approach to feature selection and classification in high-dimensional datasets, Pattern Recognition 46 (2013) 9, 2531–2537. DOI: 10.1016/j.patcog.2013.02.00710.1016/j.patcog.2013.02.007Search in Google Scholar
[24] Jakkula, V.: Tutorial on SVM, Pullman, Washington, USA, 2006.Search in Google Scholar
[25] Vempati, S.; Vedaldi, A.; Zisserman, A.; Jawahar, C. V.: Generalized RBF feature maps for Efficient Detection, in: Proc. British Machine Vision Conference 2010. British Machine Vision Association, 2010.10.5244/C.24.2Search in Google Scholar
[26] Fan, J.; Wang, X.; Wu, L.; Zhou, H.; Zhang, F.; Yu, X.; Lu, X.; Xiang, Y.: Comparison of Support Vector Machine and Extreme Gradient Boosting for predicting daily global solar radiation using temperature and precipitation in humid subtropical climates: A case study in China, Energy Conversion and Management 164 (2018), 102–111 DOI: 10.1016/j.enconman.2018.02.08710.1016/j.enconman.2018.02.087Search in Google Scholar
© 2023 Walter de Gruyter GmbH, Berlin/Boston
Articles in the same Issue
- Inhalt
- Editorial
- Editorial
- Dezincification in cast and heat-treated alpha-beta brass samples
- Robotization of conventional electrolytic process in metallography
- Comparison between image based and tabular data-based inclusion class categorization
- Development of an automated 3D metallography system and some first application examples in microstructural analysis
- Picture of the Month
- Picture of the Month
- News
- News
- Meeting Diary
- Meeting Diary
Articles in the same Issue
- Inhalt
- Editorial
- Editorial
- Dezincification in cast and heat-treated alpha-beta brass samples
- Robotization of conventional electrolytic process in metallography
- Comparison between image based and tabular data-based inclusion class categorization
- Development of an automated 3D metallography system and some first application examples in microstructural analysis
- Picture of the Month
- Picture of the Month
- News
- News
- Meeting Diary
- Meeting Diary
