應用類別不平衡資料探勘於粗軋機打滑之研究

本研究主要的目的為應用類別不平衡資料探勘技術來探討鋼鐵廠粗軋機打滑現象之成因與預測。打滑為粗軋機在軋延過程中，軋輥與軋物相互不平衡而產生滑動的現象，可能造成軋機設備損壞或軋物品質不良等問題，進而影響到生產成本與時程。因此，打滑問題已成為鋼鐵廠高度重視且急需解決的問題之一。本研究以中鋼公司熱軋場之粗軋機為例，收集大量打滑及正常相關的製程及量測參數資料做為本研究分析的基礎。由於所收集的資料集合中，打滑的資料數遠少於正常的資料數，傳統的分類技術並不適用。因此，我們採用類別不平衡分類技術來進行分析，包含少數增加法、多數減少法及成本敏感支援向量機。實驗結果顯示，上述類別不平衡技術除了可以有效地找出導致打滑的重要成因之外，並可以大幅提高打滑與正常資料分類的準確度。此外，藉由所建立的分類模型可以對製程及量測資料進行打滑預測。根據分析的結果，粗軋機打滑現象之成因與速度、厚度、軋延力跟平均扭矩的下軸有相關，未來可藉由這結果來建立打滑預測之模型，以達到減少打滑現象。

關鍵字

粗軋機；打滑；資料探勘；決策樹；支援向量機；類別不平衡

並列摘要

The main purpose of this study is applying the class-imbalanced data mining techniques to analyze the contributing factors and predict the occurrence of the slip phenomenon of rough mills in an iron and steel works. The slip phenomenon is caused by the imbalanced mutual influence between the roller and the rolled item in the rolling process. This phenomenon may cause the problems of damage on the rolling equipment and poor quality of rolled items. Therefore, the slip analysis has become a highly respected issue in an iron and steel works. This study takes the rough mill in China Steel Corporation for instance and make the analysis based on a huge collection of process and measured data related to the slip and normal phenomenons. Since the number of slip-related data is much less than the normal data in the collected dataset, traditional classification techniques are unsuitable for the slip analysis. Therefore, we adopt three class-imbalanced classification techniques, including under-sampling, over-sampling, and cost-sensitive support vector machine. Experimental results have revealed the adopted class-imbalanced classification techniques can effectively discover the important contributing factors of slip phenomenon and produce high rates of precision and recall. Moreover, the construed and trained classifier can be applied to predict the occurrence of the slip. In addition, the analytic results have shown the slip phenomenon of rough mills is correlated to the speed, thickness, rolling force, and the averaged torque of bottom work roll.

並列關鍵字

rough mill ； slip ； data mining ； decision tree ； support vector machine ； class-imbalanced

參考文獻

11.吳欣儒(2011)。資料探勘技術於病人疼痛自控裝置之應用與分析。國立交通大學資訊科學系碩士論文。

1.Barandela, R., Sanchez, J.S., Garcia, V. and Rangel, E. (2003). Strategies for learning in class imbalance problems. Pattern Recognition,36, 849-851.

3.Tan,C.,Gilbert,D.,Deville,Y.(2013).Multi-Class Protein Fold Classification Using a New Ensemble Machine Learning Approach, Genome Informatics, 14, 206-217.

4.Sun, Y. ,Kamel, M. S., Wang, Y. (2006).Boosting for Learning Multiple Classes with Imbalanced Class Distribution, Proc. Int'l Conf. Data Mining, 592-602.

7.Lewis, D. ,Catlett, J.(1994). Heterogeneous Uncertainty Sampling for Supervised Learning, Proceedings of the 11th International Conference on Machine Learning, 144-156.

國際替代計量

應用類別不平衡資料探勘於粗軋機打滑之研究

全文下載

主題瀏覽