Prediction on Molten Steel End Temperature during Tapping in BOF based on LS-SVM and PSO
A new molten steel end temperature prediction model is built employing LS-SVM. To seek the optimal parameters of regularization parameter γ and kernel parameter σ in LS-SVM, an PSO algorithm is also proposed. To test the proposed predictor, the prediction model is applied on practical data from Fujian Sangang steelmaking collected in 100t BOF, and the validation is carried on the performance of the prediction. The model overcomes the blindness and the burden in time consuming of cross validation method, and at the same time, inherits the strong learning ability from small sample and the characteristics of simple calculation of LS-SVM. In the LS–SVM optimized by PSO test cases, the Maximum Absolute Error (MAE) and the Root Mean Squares Error (RMSE) are the lowest, and the Pearson Relative Coefficient (PRC) is the highest. The results suggest that the LS-SVM optimized by PSO model can be extended to end-point judgment applications in achieving greater forecasting accuracy and quality.
Least Squares Support Vector Machine Particle Swarm Optimization End Temperature Basic Oxygen Furnace.
Yang Wei Meng Hongji Huang Yajuan Xie Zhi
Northeastern University, No.11, Lane 3, Wenhua Road, Heping District, Shenyang, 110004 China Shenyan Northeastern University, No.11, Lane 3, Wenhua Road, Heping District, Shenyang, 110004 China
国际会议
上海
英文
233-236
2011-10-27（万方平台首次上网日期，不代表论文的发表时间）