Research on Prediction Method of Alloy Element Yield in Smelting Stage of Iron and Steel Product Based on Improved Support Vector Regression

Abstract:

In the actual production of iron and steel enterprises, alloy element yield is difficult to predict because it changes with different materials, processes, etc. Then planning weights of raw materials can't be calculated accurately so as to influence raw material cost planning control. Taking raw material attributes, process parameters, and etc. of smelting stage as the influence factors, the prediction model of alloy element yield is built. In order to increase the model’s prediction accuracy, parameter optimization method for support vector regression (SVR) based on ant colony algorithm (ACO) is designed, which optimizes punish parameter, nuclear parameter and sensitive coefficient. The performance of the SVR algorithm with optimized parameters is compared with the grid search algorithm to verify that the former’s performance and efficiency are better. The prediction method of alloy element yield based on the above improved support vector regression is built, whose regression and generalization performance are better compare with BP neural network, so that the relationship between influence factors and the alloy element yield is established. It can predict alloy element yield accurately according to the actual process and provide methods for realizing lean production in iron and steel enterprises.

• Periodical: Advanced Materials Research (Volumes 562-564)
• Pages: 302-307
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.562-564.302
• Online since: August 2012
• Authors: Qiu Nan Meng, Guang Qin Guo, Wen Jing Shi, Zhi Ying Nan
• Keywords: Alloy Element Yield, Ant Colony Algorithm (ACA), Iron, Parameter Optimization, Steel, Support Vector Regression (SVR)

References

[1] Peng Yu, Dongping Zhan, Zhouhua Jiang. Development of A Terminal Composition Prediction Model for Steel refine with Ladle Furnace [J]. Journal of Materials and Metallurgy. 2006, 5(1): 20-22.

[2] Hongyuan Ma. Model of Alloy Element Yield Based on Multiple Regression Analysis [J]. Iron & Steel Technology. 2000, 4: 13-16, 36.

[3] Xianwen Gao, Aoan Zhang, Qinglai Wei. Neural Network Based Prediction of Endpoint in Ladle Refining Process [J], Journal of Northeastern University. 2005, 26(8): 726-728.

[4] Ping Yuan, Fuli Wang, Zhizhong Mao. Alloying Component Controlling Model for EAF Steelmaking [J]. Industrial Heating. 2006, 35(2): 1-4.

[5] Amir Safari. Non-parametric Estimation of A Multiscale[J]. Routledge. 2009, 9(l): 105-121.

[6] Dmitry Kropotov, Nikita Ptashko. On Kernel Selection in Relevance Vector Machines Using Stability Principle [C]. The 18th International Conference on Pattern Recognition. 2006, (4): 233–236.

DOI: 10.1109/icpr.2006.900

[7] Sathiya S, Keerthi CJ. Asymptotic Behavior of Support Vector Machines with Gaussian kernel [J]. Neural Computation, 2003, 15(7): 1667-1689.

DOI: 10.1162/089976603321891855

[8] Jianqiang Gao, Junyou Zhao, Xiaoying Fan. Two Improved D ata Processing Methods and Their Applications in SIS [J]. Automation of Electric Power Systems. 2009, 33(1): 90-92.