摘要
精矿品位是矿物浮选过程中的一项关键工艺指标,但在实际生产中该变量难以在线检测;且该过程非常复杂,具有高度的非线性及不确定性,常用的建模方法精度不够。针对上述问题,设计了一种顺序后向回归(SBS)与非线性自回归(NARX)神经网络结合的软测量算法。算法利用SBS对输入变量进行筛选以降低冗余变量对模型精度的影响,采用NARX神经网络对精矿品位进行预测。工业运行数据的仿真结果表明,该算法可以有效预测精矿品位,并且与其他算法相比,具有一定的优越性。
The concentrate grade is a key technologic parameter in the process of mineral flotation,but it’s difficult to be detected online in practical production.Moreover,the process is very complex and has a high degree of nonlinearity and uncertainty that make the commonly used modeling methods are not accurate enough.Regarding the issue above,a soft sensor algorithm combining sequential backward selection(SBS)and nonlinear auto-regressive models with exogenous inputs(NARX)neural network is designed.In this algorithm,the SBS is used to select the input variable to reduce the influence of redundant variables on the accuracy of the model,and the NARX neural network is used to predict the concentrate grade.Simulation results of industrial operation data show that the algorithm can effectively predict the concentrate grade,and has certain superiority compared with other algorithms.
作者
王洪勋
刘全
吴浩
吴修粮
孙凯
WANG Hongxun;LIU Quan;WU Hao;WU Xiuliang;SUN Kai(School of Electrical Engineering and Automation,Qilu University of Technology(Shandong Academy of Sciences),Jinan 250353,China;State Key Laboratory of Process Automation in Mining&Metallurgy,Beijing 100160,China;Beijing Key Laboratory of Process Automation in Mining&Metallurgy,Beijing 100160,China)
出处
《有色金属（选矿部分）》
CAS
北大核心
2020年第2期95-99,共5页
Nonferrous Metals（Mineral Processing Section）
基金
国家自然科学基金资助项目(61603203)
矿冶过程自动控制技术国家(北京市)重点实验室开放研究基金(BGRIMM-KZSKL-2018-01)
国家级大学生创新创业训练计划(201910431063)。
关键词
浮选过程
软测量
NARX神经网络
顺序后向回归
flotation process
soft sensor
NARX neural network
sequential backward selection
