基于FA-PSO-RBF神经网络的富氧底吹铜锍品位预测模型  

Grade Prediction Model of Oxygen-enriched Bottom Blowing Copper Matte Based on FA-PSO-RBF Neural Network

作　　者：黄旷 张晓龙[1] 胡建杭[2] 徐建新[2] 宋进 武龙飞 刘杰 HUANG Kuang;ZHANG Xiao-long;HU Jian-hang;XU Jian-xin;SONG Jin;WU Long-fei;LIU Jie(School of Mechanical and Electrical Engineering,Kunming University of Science and Technology,Kunming 650093,China;School of Metallurgical and Energy Engineering,Kunming University of Science and Technology,Kunming 650093,China)

机构地区：[1]昆明理工大学机电工程学院,昆明650093 [2]昆明理工大学冶金与能源工程学院,昆明650093

出　　处：《有色金属（冶炼部分）》2023年第6期61-68,102,共8页Nonferrous Metals（Extractive Metallurgy)

基　　金：国家自然科学基金联合基金资助项目(U2102213);云南省科技厅重大专项(202202AG050002)。

摘　　要：铜锍品位是富氧底吹铜熔炼过程中的一个关键工艺参数,针对铜锍品位实时检测困难、检测结果滞后时间长、指导生产工艺参数优化滞后等问题,基于生产数据深入挖掘及处理,提出了一种基于FA-PSO-RBF神经网络的铜锍品位预测模型。首先为了降低模型的预测误差,利用FA分析方法对原始生产数据进行降维处理,确定主要因子数量为6个,并计算因子得分,然后针对RBF神经网络模型对关键参数依赖性较大的不足,利用改进PSO算法对网络结构中的关键参数进行寻优,最后,以因子得分为输入,铜锍品位值为输出,通过实际生产数据验证模型的准确性,并与RBF、标准PSO-RBF预测模型进行对比,结果表明,本文构建的铜锍品位预测模型预测精度更高,与标准PSO-RBF预测模型相比,RMSE和MAE的值分别降低了17.2%和21.2%,该预测模型对富氧底吹铜熔炼生产过程参数优化控制提供了一种方法借鉴。Copper matte grade is a key process parameter in the oxygen rich bottom blown copper smelting process.Aiming at the difficulties of real-time detection of copper matte grade,the long lag time of detection results,and the delay in guiding the optimization of production process parameters,a copper matte grade prediction model based on FA-PSO-RBF neural network was proposed based on in-depth mining and processing of production data.First,in order to reduce the prediction error of the model,the FA analysis method is used to reduce the dimension of the original production data,determine the number of main factors as 6,and calculate the factor score.Then,in view of the deficiency of the RBF neural network model that relies heavily on key parameters,the improved PSO algorithm is used to optimize the key parameters in the network structure.Finally,the factor score is used as the input,and the copper matte grade value is used as the output,The accuracy of the model is verified by actual production data,and compared with RBF and standard PSO-RBF prediction models.The results show that the prediction accuracy of the copper matte grade prediction model constructed in this paper is higher,and the values of RMSE and MAE are reduced by 17.2%and 21.2%respectively compared with the standard PSO-RBF prediction model.This prediction model provides a method reference for the optimal control of parameters in the oxygen rich bottom blown copper smelting production process.

关 键 词：FA分析 改进PSO算法 RBF 铜锍品位预测 

分 类 号：TF811[冶金工程—有色金属冶金]