摘要
提出了一种最小二乘支持向量机的连铸板坯表面温度预测新模型。以中间罐温度、拉速、二冷水量等主要工艺因素为输入,连铸坯表面温度为输出,通过最小二乘支持向量机模型拟合输入与输出之间的复杂非线性函数关系。以现场采集的连铸生产工艺数据为样本对模型进行学习训练,用训练好的模型预测在一定工艺条件下板坯的表面温度。实践表明该方法具有建模速度快、预测精度高、操作简便等优点,不仅克服了常规的BP预测模型的不足,而且性能优于标准支持向量机预测模型。
Based on least square support vector machine（LS-SVM） a novel slab surface temperature prediction model was proposed. With a few main processing parameters such as the tundish temperature, drawing speed and volume of the secondary cooling water as inputs, the slab surface temperature as output, the complicated non-linear functional relation between input and output is derived by means of the LS-SVM based model. The new model was repeatedly practised for training purpose using the on-site collected continuous casting process parameters as the basic data and thereafter the well trained model was used to predict the surface temperature of the as cast slab under a specific processing condition. Practical results show that the LS-SVM model is advantageous of fast model construction, high prediction accuracy and ease of operation in comparison to the conventional BP prediction model. Moreover, the LS-SVM model is more superior over the standard SVM model in the performance of prediction.
出处
《炼钢》
CAS
北大核心
2009年第1期55-59,共5页
Steelmaking
关键词
连铸板坯
表面温度
预测模型
最小二乘支持向量机
continuous cast slab
surface temperature
prediction model
least square support vector machine
