基于人工智能神经网络的浮选回收率软测量模型研究  

Flotation Recovery Soft Sensor Model Based on Artificial Intelligence Neural Network

作　　者：邢亚英 付丽琴[1] 段晓亮 侯佳丽 闫硕 XING Yaying;FU Liqin;DUAN Xiaoliang;HOU Jiali;YAN Shuo(Beijing Institute of Economics and Management,Beijing 100102,China)

机构地区：[1]北京经济管理职业学院,北京100102

出　　处：《有色金属工程》2021年第9期95-99,共5页Nonferrous Metals Engineering

基　　金：北京经济管理职业学院校级重点课题(19ZHD02)。

摘　　要：在浮选流程中,回收率是一项重要的工艺指标,目前主要通过离线人工化验的方式获取,少量的企业通过在线仪表测量。人工化验具有2-3 h的时间延迟,而在线测量仪表价格高昂,导致工人不能及时有效地得到回收率实时数据,从而对工艺做出相应的调整。人工智能神经网络具有容错能力强、鲁棒性好等优点,在分析浮选工艺中多个影响因素的基础上,提出了基于人工智能神经网络浮选回收率的软测量模型。该模型可以有效克服浮选过程变量耦合性强、工艺复杂等因素。通过仿真验证,该模型的输出与实际样本数据误差很小,可有效对浮选回收率进行软测量。The recovery rate is an important production index in the flotation process.At present,it is mainly obtained through manual testing,and a small number of mining companies obtain it through online measurement with a fluorescence analyzer.Manual testing has a time delay of 2-3 hours,and online measuring instruments are expensive,resulting in workers cannot obtain real-time data on the recovery rate in a timely,and making corresponding adjustments to the process.Artificial intelligence neural network has the advantages of robustness and strong fault tolerance.Based on the analysis of relevant influencing factors in the flotation process,this paper proposes a soft sensor model based on artificial intelligence neural network.The soft sensor model of recovery rate can effectively overcome the incomplete factors such as the process complexity,the large noise,and the strong correlation.Simulation shows that the output of the model can approximate the actual sample with high accuracy,and can effectively identify the nonlinear relationship between the impact factor and the recovery rate.

关 键 词：人工智能 软测量 浮选回收率 

分 类 号：TD923[矿业工程—选矿] TD928.9