摘要
转炉终点温度是决定钢质量的关键因素,在炼钢的恶劣环境中难以检测。建立了基于核独立元回归(KICR)方法的终点温度的预测模型。将核独立元分析(KICA)与回归分析相结合,利用KICA方法提取输入数据矩阵的独立元(KIC)矩阵;分别以KIC、实值矩阵为自变量和因变量进行训练,求取最小二乘回归(LSR)系数建立预测模型。工业现场生产数据仿真结果表明:与PCR,PLSR和ICR等预测模型相比,基于KICR的转炉终点温度预测模型,预测精度高、跟踪性能较好,可为实际生产中的终点控制提供参考,提高生产效益。
Endpoint temperature is the key factor determining the steel quality and one of the most difficulty to control in the process of converter steel-making. A predicting model for endpoint temperature based on kernel independent component regression （KICR） in converter steel-making is established. This method combines KICA and regression analysis, which extracts KIC matrix firstly, and then establish the predicting model by the KIC and real matrix. Data simulation results show that compared with ICR, partial least squares regression （PLSR） and principal component regression（PCR） method, KICR method has a higher regression precision, and good tracking characteristics. This method can provide reference for end point control in actual production to improve production efficiency.
出处
《传感器与微系统》
CSCD
2017年第1期153-156,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61064004)
关键词
转炉炼钢
终点控制
核独立元分析
核独立元回归
回归预测
converter steel-making
end point control
kernel independent component analysis （KICA）
kernelindependent component regression（KICR）
regression prediction
