img

QQ群聊

img

官方微信

高级检索

黄金科学技术 ›› 2024, Vol. 32 ›› Issue (1): 170-178.doi: 10.11872/j.issn.1005-2518.2024.01.069

• 采选技术与矿山管理 • 上一篇    下一篇

基于双隐含层BP神经网络的某金矿回收率预测研究

张帅(),赵鑫,彭祥玉,王宇斌(),桂婉婷,田家怡   

  1. 西安建筑科技大学资源工程学院,陕西 西安 710055
  • 收稿日期:2023-05-05 修回日期:2023-09-04 出版日期:2024-02-29 发布日期:2024-03-22
  • 通讯作者: 王宇斌 E-mail:13474448550@163.com;wywywyb@xauat.edu.cn
  • 作者简介:张帅(1999-),男,陕西榆林人,硕士研究生,从事矿物材料及资源综合利用研究工作。13474448550@163.com
  • 基金资助:
    陕西省自然科学基金项目“双重难选碳质金矿中的石墨吸附机理研究”(2019JQ-545)

Prediction of the Recovery Rate of a Gold Mine Based on Double Hidden Layer BP Neural Network

Shuai ZHANG(),Xin ZHAO,Xiangyu PENG,Yubin WANG(),Wanting GUI,Jiayi TIAN   

  1. School of Resources Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,Shaanxi,China
  • Received:2023-05-05 Revised:2023-09-04 Online:2024-02-29 Published:2024-03-22
  • Contact: Yubin WANG E-mail:13474448550@163.com;wywywyb@xauat.edu.cn

摘要:

为掌握某金矿选矿工艺影响因素对金实际回收率的作用规律并预测金的回收率,采用正交试验方法开展了金矿浮选试验,通过Pearson系数分析金回收率对不同工艺因素的敏感性,并利用双隐含层BP神经网络对金回收率进行预测。结果表明:生产过程中金回收率对不同因素的敏感性由大到小依次为2#油用量、Na2S用量、丁基黄药用量、CuSO4用量和磨矿细度。在此基础上,选用2#油用量、Na2S用量和丁基黄药用量3个主要影响因素,使用不同隐含层激活函数的BP神经网络对金回收率进行预测。预测结果表明:当使用“logsig”作为激活函数时,其拟合度与精度较高,拟合优度R2为0.9792,相对平均误差仅为0.666%,说明该模型能够较好地预测金回收率。研究结果对贵金属矿山生产中金属回收率的预测有一定的参考意义。

关键词: BP神经网络, Pearson系数, 激活函数, 影响因素, 金矿, 回收率

Abstract:

In order to grasp the action law of process factors affecting the actual recovery rate of a gold ore and predict the gold recovery rate,the flotation test was carried out by the method of orthogonal experiment.The sensitivity of process factors was analyzed by Pearson coefficient,and the gold recovery rate was predicted by using double hidden layer BP neural network.The results show that the sensitivity of the gold recovery rate to different factors in the production process is in descending order:2# oil dosage,sodium sulfide dosage,butyl xanthate dosage,copper sulfate dosage and grinding fineness.The reasons for the significant differences in the effects of 2# oil dosage,sodium sulfide dosage and butyl xanthate dosage on gold recovery rate were also elucidated.On this basis,used three main influencing factors such as 2# oil dosage,the study selected different input layer to the first implicit layer functions,such as tansig,purelin and logsig,and the rest of the activation functions remained unchanged.The first hidden layer to the second hidden layer function was logsig,and the second hidden layer to the output layer function was purelin.When research used logsig as the activation function,the fitted degree and accuracy are high,the goodness of fit R2 is 0.9792,and the relative average error is only 0.666%.The model can be used to predict the recovery rate of gold.The research has certain reference significance for the prediction of metal recovery rate in the production of precious metal mines.

Key words: BP neural network, Pearson coefficient, activation function, influencing factors, gold mine, recovery rate

中图分类号: 

  • TD953

表1

金矿石多元素分析结果"

成分含量/%成分含量/%
Au*2.13Al2O312.66
Ag*1.31TFe6.40
S2.28K2O2.78
MgO3.34Pb0.06
SiO255.78Zn0.11
CaO2.83

表2

金物相分析结果"

物相类别含量/(g·t-1占比/%
合计0.94100.00
裸露金及半裸露金1.3364.42
硫化物包裹金0.3516.07
赤褐铁矿包裹金0.156.79
碳酸盐包裹金0.135.95
硅酸盐包裹金0.156.76

图1

金矿浮选试验流程图"

表3

正交试验因素和水平"

水平因素A因素B因素C因素D因素E
1551510600
26522128015
375281410030
485351612045

表4

相关系数与相关性之间的关系"

相关系数|ρ|相关程度
0.8~1.0极强相关
0.6~0.8强相关
0.4~0.6中等程度相关
0.2~0.4弱相关
0.0~0.2极弱相关或无相关

表5

正交试验结果"

试验

编号

水平回收率/%
ABCDE
11111165.46
21222275.92
31333384.15
41444483.52
52123470.42
62214377.98
72341282.71
82432180.13
93134275.69
103243182.54
113312480.02
123421376.69
134142364.32
144231477.45
154324182.12
164413284.30

表6

金回收率与主要影响因素的相关系数"

因素相关系数
磨矿细度(-74 μm含量)/%0.0052
2#油用量/(g·t-10.7436
丁基黄药用量/(g·t-10.1302
Na2S用量/(g·t-10.3319
CuSO4用量/(g·t-1-0.0553

图2

主要因素对金回收率的影响"

图3

BP神经网络模型示意图"

图4

3种激活函数的训练图和测试图"

表7

3种激活函数训练集和测试集的拟合优度"

激活函数训练集R2测试集R2
tansig0.97340.9842
purelin0.96510.9401
logsig0.97930.9862

图5

3种激活函数预测结果对比"

图6

logsig为激活函数的BP神经网络训练过程曲线"

表8

3种激活函数误差对比"

激活函数平均相对误差/%拟合优度R2
tansig0.7080.9700
purelin1.3660.9654
logsig0.6660.9792
Afzali E, Muthukumarana S,2023.Gradient-free kernel conditional stein discrepancy goodness of fit testing[J].Machine Learning with Applications,12(2):100463.
Cheng Juanjuan,2022.An empirical study on the relationship between research and teaching in universities:An analysis based on Pearson’s correlation coefficient[J].China University Science and Technology,(10):46-52.
Comley B A, Harris P J, Bradshaw D J,et al,2002.Frother characterization using dynamic surface tension measurements[J].International Journal of Mineral Processing,64:81-100.
Fan Songhao, Su Panyun, Hou Xiuhong,2022.Characteristics of gold resources and metallogenic regularity in China[J].China Metal Bulletin,(9):47-49.
Feng Yan, Liu Jian,2023.Prediction of mine friction resistance during tramcar running based on BP neural network[J].Journal of Safety Science and Technology,19(1):54-59.
Hamid K, Sam A,2011.Flotation frothers:Review of their classifications,properties and preparation[J].The Open Mineral Processing Journal,4(1):25-44.
Guo Rui,2020.Research on the Prediction of Reagent Addition in Concentrator Based on RSM and BP neural network[D].Kunming:Kunming University of Science and Technology.
Li Bin, Zhang Yifan, Yan Shiye,et al,2022.Research on photovoltaic power generation prediction method based on improved extreme learning machine(ELM)[J].Journal of Engineering for Thermal Energy and Power,37(10):207-214.
Li Liang, Wang Yubin, Lin Xingtong,et al,2022.Optimization of influence conditions on strength of gypsum-based composite cementitious materials using BP neural network[J].Nonferrous Metals(Mineral Processing Section),74(4):19-25.
Li Shuqin, Wang Yubin, Ma Xiaoxiao,et al,2022.Effect of magnetization parameters on removal efficiency of zinc in fly ash by flotation and its model[J].Nonferrous Metals(Mineral Processing Section),(4):27-32.
Li Z X, Yang Y, Li L W,et al,2023.A weighted Pearson correlation coefficient based multi-fault comprehensive diagnosis for battery circuits[J].Journal of Energy Storage,60:106584.
Mondal B, Meetei M S, Das J,et al,2015.Quantitative recognition of flammable and toxic gases with artificial neural network using metal oxide gas sensors in embedded platform[J].Engineering Science and Technology,an International Journal,18(2):229-234.
Nie Shanyu, He Guichun, Shi Yan,et al,2023.Research progress of flotation index prediction modeling based on data driven[J].The Chinses Journal of Nonferrous Metals,33(7):2330-2338.
Ren Chuancheng, Xia Wencheng, Wang Wenbo,et al,2023.Prediction model for iron concentrate grade based on unbiased grey GM(1,1)[J].Nonferrous Metals(Mineral Processing Section),(1):41-45,56.
Wang Bin, Li Jingchao, Wang Chengxi,et al,2020.An overview of characteristics and prospecting of gold ore deposits in China[J].Geological Journal of China Universities,26(2):121-131.
Wang Kai, Chen Yun, Tang Jianlin,2023.Research on the application of BP neural network in performance detection of long-span cable-stayed bridges[J].Journal of Zhejiang University of Technology,51(2):171-179.
Wang Mingli, Xu Baojin, Zhu Jiaqian,et al,2020.Flotation experiment of tailings from a gold mine in Jiangxi Province[J].Metal Mine,49(4):212-216.
Wang Xiaochuan, Shi Feng, Yu Lei,et al,2013.43 Case Studies of MATLAB Neural Network[M].Beijing:Beihang University Press.
Wang Xiu, Wang Jianping, Chen Hong,et al,2015.Situation analysis and sustainable development strategy of gold resources in China[J].Mining Research and Development,35(10):99-103.
Wang Yandong,2020.Analysis and suggestions of gold resources prospecting situation in China from 2009 to 2019[J].China Mining Magazine,29(11):7-13.
Xiang Qun,2019.Preliminary study on gold resources and geological exploration situation in China[J].Journal of the Science and Technology,(3):105.
Xie Fengyun, Dong Jiankun, Wang Erhua,et al,2021.Research on gearbox fault diagnosis based on double hidden layer RWPSO-BP neural network[J].Modern Manufacturing Engineering,(6):155-160.
Xu Xiaoyang,2013.Review of research on leaching process of carbonaceous refractory gold ore[J].Gold Science and Te-chnology,21(1):82-88.
Xun Jingwen, Wang Yubin, Lei Dashi,et al,2020.Research on orthogonal test of flotation of a gold ore in Gansu[J].Precious Metals,41(4):56-60.
Yan Zan, Wang Wendan, Wang Lu,et al,2018.Experimental study on beneficiation of one gold mine in Gansu Province[J].Gold Science and Technology,26(1):74-80.
Yu Shengli, Wang Yuhua, Zhang Ying,et al,2013.Beneficiation experimental study on a low-grade refractory gold ore[J].Nonferrous Metals(Mineral Processing Section),(2):17-21,25.
Zhao Xianghong, Bao Jingyang, Ouyang Yongzhong,et al,2019.Detecting outlier of multibeam sounding with BP neural network[J].Geomatics and Information Science of Wuhan University,44(4):518-524.
Zhou Guanglang, Zhou Dongyun,2023.Experimental study on gold recovery from polysulfide fine-grained disseminated gold mines[J].Precious Metals,44(1):47-53.
程娟娟,2022.高校科研与教学关系实证研究——基于皮尔逊相关系数的分析[J].中国高校科技,(10):46-52.
樊松浩,苏攀云,侯秀宏,2022.中国金矿资源特征及成矿规律概要[J].中国金属通报,(9):47-49.
冯燕,刘剑,2023.基于BP神经网络的矿车运行时矿井摩擦阻力的预测[J].中国安全生产科学技术,19(1):54-59.
郭锐,2020.基于RSM和BP神经网络预测选矿厂药剂添加量研究[D].昆明:昆明理工大学.
李斌,张一凡,颜世烨,等,2022.基于改进极限学习机ELM的光伏发电预测方法研究[J].热能动力工程,37(10):207-214.
李亮,王宇斌,林星彤,等,2022.利用BP神经网络优化石膏基复合胶凝材料强度的影响条件[J].有色金属(矿山部分),74(4):19-25.
李淑芹,王宇斌,马晓晓,等,2022.水体磁化参数对飞灰中锌的浮选去除效果的影响规律及其模型[J].有色金属(选矿部分),(4):27-32.
聂善煜,何桂春,石岩,等,2023.基于数据驱动的浮选指标预测建模研究进展[J].中国有色金属学报,33(7):2330-2338.
任传成,夏文成,王文博,等,2023.基于无偏灰色GM(1,1)的铁精矿品位预测模型[J].有色金属(选矿部分),(1):41-45,56.
王斌,李景朝,王成锡,等,2020.中国金矿资源特征及勘查方向概述[J].高校地质学报,26(2):121-131.
王凯,陈韵,汤建林,2023.BP神经网络在大跨斜拉桥性能检测中的应用研究[J].浙江工业大学学报,51(2):171-179.
王明莉,徐宝金,朱加乾,等,2020.江西某金矿尾矿再选试验研究[J].金属矿山,49(4):212-216.
王小川,史峰,郁磊,等,2013.MATLAB神经网络43个案例分析[M].北京:北京航空航天大学出版社.
王修,王建平,陈洪,等,2015.我国金矿资源形势分析及可持续发展对策[J].矿业研究与开发,35(10):99-103.
王燕东,2020.2009—2019年我国金矿资源勘查形势分析与对策[J].中国矿业,29(11):7-13.
相群,2019.我国金矿资源与地质勘查形势的初步研究[J].科技风,(3):105.
谢锋云,董建坤,王二化,等,2021.基于双隐含层RWPSO-BP神经网络的齿轮箱故障诊断研究[J].现代制造工程,(6):155-160.
许晓阳,2013.碳质难处理金矿浸出工艺研究进展[J].黄金科学技术,21(1):82-88.
荀婧雯,王宇斌,雷大士,等,2020.甘肃某金矿浮选正交试验研究[J].贵金属,41(4):56-60.
阎赞,王闻单,王露,等,2018.甘肃某金矿选别试验研究[J].黄金科学技术,26(1):74-80.
余胜利,王毓华,张英,等,2013.某难选低品位金矿的选矿试验研究[J].有色金属(选矿部分),(2):17-21,25.
赵祥鸿,暴景阳,欧阳永忠,等,2019.利用BP神经网络剔除多波束测深数据粗差[J].武汉大学学报(信息科学版),44(4):518-524.
周光浪,周东云,2023.多硫化物微细粒浸染型金矿回收金试验研究[J].贵金属,44(1):47-53.
[1] 王开彬, 刘钦, 王洪涛. 压力型锚索锚固段荷载传递特征及影响因素研究[J]. 黄金科学技术, 2024, 32(1): 123-131.
[2] 王兴春, 邱海城, 李建平, 智庆全, 李华, 武军杰, 邓晓红, 吴琼. 辽东半岛五龙金矿外围电性特征及找矿意义[J]. 黄金科学技术, 2024, 32(1): 1-12.
[3] 史磊, 王西荣, 宁霄峰, 鹿峰宾, 许延波, 李亚楠. 山东南吕—欣木金矿床金的赋存状态及富集机制[J]. 黄金科学技术, 2024, 32(1): 41-54.
[4] 周昌微, 谢贤平, 都喜东. 基于曲线拟合和神经网络的独头巷道CO浓度预测研究[J]. 黄金科学技术, 2024, 32(1): 75-81.
[5] 杨玮, 邓博, 龙涛, 邓莎, 薛梦鸽, 方楠. 基于效果—效率的金矿绿色矿山建设综合评价研究[J]. 黄金科学技术, 2023, 31(6): 919-929.
[6] 孙浩杰,杨俊彦,李雪林,徐忠敏,谷建国,游世辉. 黄金矿山格子型球磨机节能降耗应用研究[J]. 黄金科学技术, 2023, 31(6): 1044-1050.
[7] 许云美, 袁利伟, 龙皓楠. 干堆尾矿库稳定性影响因素的敏感性分析[J]. 黄金科学技术, 2023, 31(6): 1014-1022.
[8] 宋高瑞, 翟新伟, 王二腾, 武磊, 陈万峰, 郑菲菲, 王海东, 王金荣. 甘肃花牛山金矿床成矿流体性质及矿床成因[J]. 黄金科学技术, 2023, 31(6): 873-887.
[9] 陈国民, 杨洪英, 陈彦臻, 张广积. 金矿预氧化处理过程中砷的转化[J]. 黄金科学技术, 2023, 31(5): 865-872.
[10] 张国栋, 刘佳, 马凤山, 李光, 郭捷. 三山岛金矿海底开采井下沉降特点及影响因素浅析[J]. 黄金科学技术, 2023, 31(5): 785-793.
[11] 单文法, 毛先成, 刘占坤, 邓浩, 陈进, 张维, 王海正, 杨鑫. 胶东大尹格庄金矿床成矿过程数值模拟及其找矿意义[J]. 黄金科学技术, 2023, 31(5): 707-720.
[12] 许方颖, 邹艳红, 易卓炜, 杨福强, 毛先成. 基于非均衡数据的ADASYN-CatBoost测井岩性智能识别——以胶西北招贤金矿床为例[J]. 黄金科学技术, 2023, 31(5): 721-735.
[13] 廖秋敏,罗连英. 稀土产品全球贸易网络格局演化及其影响因素[J]. 黄金科学技术, 2023, 31(5): 823-834.
[14] 樊忠平, 张望, 王卫. 陕西省山阳—商南金矿成矿规律及找矿预测研究[J]. 黄金科学技术, 2023, 31(4): 560-579.
[15] 彭科淇, 周瑞仙. 贵州某卡林型金矿精锐微泡浮选生产实践[J]. 黄金科学技术, 2023, 31(4): 689-697.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!