Because of complex reactions in Basic Oxygen Furnace (BOF) for steelmaking, the main end-point control methods of steelmaking have insurmountable difficulties. Aiming at these problems, a support vector machine (SVM) method for forecasting the BOF steelmaking end-point is presented based on flame radiation information. The basis is that the furnace flame is the performance of the carbon oxygen reaction, because the carbon oxygen reaction is the major reaction in the steelmaking furnace. The system can acquire spectrum and image data quickly in the steelmaking adverse environment. The structure of SVM and the multilayer feed-ward neural network are similar, but SVM model could overcome the inherent defects of the latter. The model is trained and forecasted by using SVM and some appropriate variables of light and image characteristic information. The model training process follows the structure risk minimum (SRM) criterion and the design parameter can be adjusted automatically according to the sampled data in the training process. Experimental results indicate that the prediction precision of the SVM model and the executive time both meet the requirements of end-point judgment online.

Date Published: 20 May 2009
PDF: 4 pages
Proc. SPIE 7283, 4th International Symposium on Advanced Optical Manufacturing and Testing Technologies: Optical Test and Measurement Technology and Equipment, 728327 (20 May 2009); doi: 10.1117/12.828702

Published in SPIE Proceedings Vol. 7283:
4th International Symposium on Advanced Optical Manufacturing and Testing Technologies: Optical Test and Measurement Technology and Equipment
Yudong Zhang; James C. Wyant; Robert A. Smythe; Hexin Wang, Editor(s)