神经网络在矿物加工中预测应用的研究现状及展望  

Research Status and Prospect of Neural Network Application in Mineral Processing Prediction

作　　者：郭锐 刘丹[1,2] 杜钰 余龙舟 GUO Rui;LIU Dan;DU Yu;YU Longzhou(Faculty of Land Resource Engineering, Kunming University of Science and Technology, Kunming 650093, China;State Key Laboratory of Complex Nonferrous Metal Resources Clean Utilization, Kunming University of Science and Technology,Kunming 650093, China;Yunnan Amde Electrical Engineering Company, Kunming 650033, China)

机构地区：[1]昆明理工大学国土资源工程学院,昆明650093 [2]昆明理工大学,省部共建复杂有色金属资源清洁利用国家重点实验室,昆明650093 [3]云南阿姆德电气工程有限公司,昆明650033

出　　处：《硅酸盐通报》2020年第4期1178-1185,共8页Bulletin of the Chinese Ceramic Society

摘　　要：神经网络作为一个近十几年来快速发展的崭新领域,具有适应性强、容错能力强、鲁棒性强等优势。但是在矿物加工中存在传统人工离线分析化验指标滞后2~3 h,磨矿粒度缺乏在线监测设备无法监测,人工调节浮选加药量人为影响因素大等问题。应用神经网络的优势特征克服了过去被认为传统难以解决的问题,神经网络通过非线性变化对原始采集数据在隐含层传递函数计算下,经输入层传输传递到输出层,然后神经网络训练自我学习出对应关系之间的客观规律,最终输入类似样本数据在对应规律下输出预测数据。在目前矿物加工领域采用神经网络方法解决了精矿品位回收率预测,浮选过程加药量预测等问题的研究。在将来神经网络的应用中,在矿物加工领域能够为选矿厂带来智能化数字化信息化,提高选矿厂控制系统的效率和准确度。As a new field of rapid development in recent decades,neural network has strong adaptability,fault tolerance and robustness.However,there are some problems in mineral processing,such as the lag of traditional off-line analysis assay indexes for 2-3 h,the lack of on-line monitoring equipment for grinding particle size and the influence factors of artificially adjusting the dosage of flotation.Application of the neural network advantages to overcome the difficult to solve the problem,which is considered by neural network nonlinear changes on the original data in the hidden layer under the action of transmission from the input layer to output layer,then neural network training self learning out corresponding relationship between the objective laws,the final input similar to the sample data in the corresponding output under the law.At present,in the field of mineral processing,neural network method has been used to solve the problems such as the prediction of concentrate grade recovery and the prediction of dosage in flotation process.In the future the application of neural network,mineral processing field can bring intelligent digital information to the mineral processing plant,improve the efficiency and accuracy of the control system of the mineral processing plant.

关 键 词：神经网络 选矿自动化 指标预测 建模 

分 类 号：P578[天文地球—矿物学]