基于自适应模糊神经网络的铜闪速熔炼冰铜温度模型研究  

Research on the Matte Temperature Model of Copper Flash Smelting Process Based on ANFIS

作　　者：曾青云[1] 周立[1] 汪金良[1] 刘桦[1] 

机构地区：[1]江西理工大学,赣州341000

出　　处：《有色金属（冶炼部分）》2007年第2期2-4,9,共4页Nonferrous Metals（Extractive Metallurgy)

基　　金：国家自然科学基金资助项目(50364004);江西省自然科学基金资助项目(0250026)

摘　　要：基于自适应模糊神经网络系统(ANFIS)及利用某闪速炼铜厂生产实践的稳定数据,建立了网络结构为3输入、单输出、隶属度函数个数为(7,5,7)的闪速炼铜过程的冰铜温度模型。结果显示其训练数据平均绝对误差为5.0℃,相对误差为0.41%;仿真检验数据平均绝对误差为6.7℃,相对误差为0.55%,表明所建立的模型预测值与生产操作数据基本吻合,该模型对铜熔炼过程的最优化具有参考价值,可以代替现有的静态配料模型用于工业在线计算机控制。The matte temperature model of copper flash smelting process which has the net-structure of 3 put-in, 1 put-out data, and the membership functions are （7, 5, 7） was developed based on Adaptive Network Fuzzy Inference System （ANFIS） and by using the practical data from one copper smelter. The results indicate that the average absolute error of train samples is 5.0℃, the relative error percentage is 0. 14%, the simulation results show that the average absolute error is 6.7℃, and the relative error percentage is 0.55%. It means that the simulative results accord with the practical data. Thus, the model has good reference value on process optimized control of copper smelting. It also can be used in industrial online control to replace the model of static mixture.

关 键 词：铜 闪速熔炼 模糊控制 神经网络 仿真 

分 类 号：TF811[冶金工程—有色金属冶金] O242.1[理学—计算数学]