基于广义最大熵回归的自适应艾萨炉铜熔炼工艺参数的软测量

Adaptive Soft Measurement of Process Parameters of ISA Furnace during Copper Melting Based on Generalized Maximum Entropy Regression

冰铜品位、冰铜温度、渣中铁硅比是艾萨炉铜熔炼过程中的三个主要工艺参数,针对这三大参数在线检测时存在成本高、滞后大、实现困难等问题,提出了一种基于广义最大熵回归的自适应艾萨炉铜熔炼过程三大参数软测量方法.首先基于核聚类的局部线性嵌入算法对熔炼过程的输入数据进行降维预处理,然后利用隐马尔科夫模型对工况进行检测,最后结合工况建立广义最大熵自适应模型.实验表明,提出的方法不仅能明显改善误差,而且测量稳定性得到提高,能为实际生产提供有益的指导.

Matte grade, matte temperature and mass rate of Fe and SiO2 are the three key process parameters during the process of copper ISA smelting. The measurement of these three parameters is not only hard to be detected on - line, but also has time - delaying and costs a lot. The adaptive soft measurement method of the three key process parameters is proposed based on generalized maximum entropy regression. Firstly, locally linear embedding algorithm based on kernel clustering is adopted to reduce the dimensionality of the input data of the smelting process as a pretreatment. Secondly, hidden Markov model is employed to detect the operating conditions. Finally, combined with the working conditions, an adaptive model based on generalized maximum entropy is established. The experiments show that the proposed method can not only significantly improve the error and measurement stability, but also be used to optimize parameters in the practical production.