基于DLSSVM的铜闪速熔炼过程工艺参数预测  

Craft parameters prediction in the process of copper flash smelting based on DLSSVM

作　　者：谢永芳[1] 夏巨龙[1] 刘建华[1] 桂卫华[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《中国科技论文》2012年第1期52-57,共6页China Sciencepaper

基　　金：国家自然科学基金资助项目(61104135;60904077);中央科研基本业务费中南大学国家杰青培育专项项目(2011JQ009)

摘　　要：针对铜闪速熔炼工业数据中含有噪声和随机干扰,难以精确地建立工艺参数预测模型的缺陷,提出小波去噪最小二乘支持向量机(DLSSVM)的工艺参数预测方法。首先采用小波变换对数据进行去噪处理,然后利用最小二乘支持向量机建立熔炼过程与工艺指标之间的非线性关系模型,并借助去噪后的样本数据进行模型参数辨识,实现铜闪速熔炼过程三大工艺指标的有效预测。某冶炼厂的实际生产数据仿真结果表明,冰铜品位、冰铜温度和渣中铁硅比的最大相对误差分别为0.72%、2.24%和0.57%,相对均方根误差分别为0.36%、0.91%和0.17%,满足实际工业生产对精度的要求,可有效指导熔炼过程操作参数的优化调整。Considering the induslrial data of copper flash smelting including noise and random disturbance, it is difficult to accurately build the prediction model of craft parameters. A prediction method of craft parameters based on wavelet denoising least squares support vector machine （DLSSVM） is proposed. Firstly, it removes the noise from data set using wavelet transform, then the nonlinear relation model is established with least square support vector machine between melting process and craft indicators, and with the denoising sample data to conduct model parameter identification, which realizes the effective prediction of the three craft indicators in the copper flash smelting process. Simulation results of actual production data in some smeltery show that the biggest relative error of the matte grade, matte temperature and mass rate of Fe and SiO2 are 0.72%, 2.24% and 0.57%, and the relative root mean square error are 0.36%, 0.91% and 0.17%. These results indicate that this precision can meet the needs of the practical industrial production, which can guide the optimal adjusting for operational parameters in the smelting process.

关 键 词：小波变换 最小二乘支持向量机 铜闪速熔炼过程 预测模型 

分 类 号：TF811[冶金工程—有色金属冶金]