基于泡沫图像特征加权K近邻算法的锌矿浮选工况识别方法

浮选工况识别在泡沫浮选工程中起着至关重要的作用,仅依靠人工经验进行主观性识别,准确性和效率都低。为此提出了一种考虑泡沫图像特征间相互作用的加权K近邻(KNN)算法用于实现浮选工况类别的识别。在本研究中,首先,通过信息熵对泡沫图像特征与浮选工况类别之间的相关性进行量化,同时评估该特征与其他特征之间的冗余性。然后,计算该特征与浮选工况类别相关性和该特征与其他特征冗余性之间的差值,将这一差值作为特征的权重。其次,在KNN算法中针对欧式距离进行特征加权,以实现KNN算法的特征加权。然后,将特征选择过程嵌入到特征加权KNN分类算法的训练过程中,并选取分类准确率最高的特征子集作为最优特征子集。最后,基于最优特征子集完成浮选工况的识别。研究结果表明,本方法与其他基准分类算法相比,在分类准确度和时间上都达到了最佳效果,验证了本研究所提出的浮选工况识别方法的有效性。

结合双模多尺度CNN特征及自适应深度KELM的浮选工况识别

针对可见光图像特征驱动的浮选工况识别方法的不足,提出一种基于双模态图像多尺度CNN特征及自适应深度自编码核极限学习机(Kernel Extreme Learning Machine,KELM)的浮选工况识别方法。先对泡沫的可见光、红外图像进行非下采样剪切波多尺度分解,设计双通道CNN网络对双模态多尺度图像进行特征提取及融合,将多个双隐层自编码极限学习机串联成深度学习网络对CNN特征逐层抽象提取,然后通过核极限学习机映射到更高维空间进行决策,最后改进量子细菌觅食算法并应用于深度自编码KELM识别模型参数优化。实验结果表明采用双模多尺度CNN特征较单模多尺度、双模单尺度CNN特征的识别精度提高了2.65%,自适应深度自编码KELM模型具有较好的分类精度和泛化性能,各工况识别的平均准确率达到95.98%,识别精度和稳定性较现有方法有较大提升。

基于双模态卷积神经网络自适应迁移学习的浮选工况识别

为提高小规模训练集下CNN特征驱动的浮选工况识别效果,提出一种基于泡沫红外与可见光图像CNN特征提取及自适应迁移学习的工况识别方法.首先构建基于AlexNet的双模态CNN特征提取及识别模型,并通过RGB-D大规模数据集对模型的结构参数进行预训练;其次,用多个串联的双隐层自编码极限学习机代替预训练模型的全连接层,实现对双模态CNN特征的融合及逐层抽象提取,然后通过核极限学习机映射到更高维空间进行决策;最后构建浮选小规模数据集对迁移后的模型进行训练,并改进量子狼群算法用于模型参数优化.实验结果表明:自适应迁移学习能够明显提高小样本数据集下的识别准确度,采用双模态CNN迁移学习较单模态CNN迁移学习的工况识别精度提高了3.06%,各工况的平均识别准确率达到96.83%,识别精度和稳定性较现有方法有较大提升.

结合泡沫图像多特征及模糊数学的浮选工况评判

针对矿物浮选过程中视觉特征多样性、工况识别在线检测难、耗时长的问题,提出一种基于泡沫图像多特征和模糊数学的浮选工况评判方法。利用像素点覆盖、水平集等方法提取图像多个特征,采用改进的模糊数学分析法构成浮选工况的评价体系,把浮选状态分成良好、中等、差三种工况。实验结果表明,相比于人工神经网络和支持向量机方法,识别率均有所提高,能够代替人工做一个初步的判断和调整,并且较接近实时判断浮选状况的要求,可作为生产优化及其生产参数调节的重要参考。

动态纹理建模在硫浮选工况的识别分析

泡沫浮选广泛应用于选矿领域,它是一种能够有效提取矿粒的方法。但是,浮选过程存在着大量的影响因素和严重的非线性,这使得浮选过程的优化控制很难实现。因此,为了保证浮选处于最优工况,有必要依据浮选泡沫的表面特征来调整相应的操作变量。本文提出了基于动态纹理建模的方法应用于浮选工况的分类。采用ARMA模型进行动态纹理建模,通过样本学习得到模型参数A,C,Q。对不同类样本模型参数A,C计算其Martin距离,根据最小距离原则来进行分类识别。仿真结果表明:所提出的动态纹理模型能准确地描述动态泡沫,且能有效地检测浮选泡沫状态。

基于LBPV的浮选泡沫图像纹理特征提取

在浮选生产中,浮选泡沫表面纹理与浮选工况密切相关,直接反映泡沫层的矿化程度(品位高低)。为了给浮选操作提供指导,提出了一种基于LBPV(local binary pattern variance)的泡沫图像纹理特征提取方法。该方法通过融合泡沫图像局部空间结构和对比度来提取泡沫图像纹理特征,然后将LBPV纹理特征应用于浮选工况状态的聚类分析。结果表明,该方法提取的纹理特征能有效反映浮选工况,且能获得更优的浮选泡沫聚类质量。

球磨机橡胶衬板在我厂的应用

赣州有色金属冶炼厂选矿车间的磨机,长期以来使用高锰钢衬板,存在着价格高、寿命短、消耗大等问题。现改用橡胶衬板,具有重量轻、耐腐蚀、噪音小、使用寿命长、密封性能好等优点,取得了明显的经济效益。

缪沅振:仅凭肉眼即可分辨精矿品位的“拼命三郎”

简介缪沅振,男,彝族,1983年2月出生,中共党员,大学专科,高级工,现任玉溪矿业有限公司磨选工段工段长。先后荣获全国优秀农民工、云南省劳动模范、云南省浮选工技能比武大赛技术状元荣誉称号,云南省五一劳动奖章获得者。2016年至2019年选厂在保证铜回收率完成考核指标的基础上,铜精矿品位超考核指标3.58%,为铜冶炼厂提供了优质原料,同时,铁精矿数量突破50万吨,铁精矿品位较计划提高一个百分点,创造效益近1000万元。通过日常细心的观察记录和调整判断,缪沅振提出了浮选岗位“三辨”“四调”和“稳当先”的操作方法,稳定提升浮选经济技术指标。这套“缪沅振工作法”大大提高了工作效率。

Effects of sodium hexmetaphosphate on flotation separation of diaspore and kaolinite

In order to offer high grade concentrate to produce alumina by Bayer, the reverse-flotation technology of bauxite was investigated. The results show that sodium hexmetaphosphate has different depression on the flotation of diaspore and kaolinite. The recoveries of diaspore decrease markedly with the increase of the concentration of sodium hexmetaphosphate with dodecyl-amine as collector. Fourier transform infrared spectra indicate that the absorbance band 875cm -1 of the asymmetric stretching vibration of bridge oxygen P—O—P shifts to 880cm -1 , and the 1264cm -1 of the asymmetric stretching vibration of the bridge PO2 shifts to 1267cm -1 in the diaspore’s spectra. This demonstrates that sodium hexmetaphosphate interacts through bridging PO2 groups with the Al atoms exposed on the diaspore and kaolinite surfaces to form P—O—Al bond. Adsorption measurements also testify that sodium hexmetaphosphate adsorbs easily on the surface of diaspore. The adsorption density of PO-3 on the surface of diaspore is about 4.7×10 -6 mol/m2, while that on the kaolinite is only about 3.5×10 -7 mol/m2 when pH value is 6.

Influences of collector DLZ on chalcopyrite and pyrite flotation

The interaction mechanism of collector DLZ in the flotation process of chalcopyrite and pyrite was investigated through flotation experiments,zeta potential measurements and infrared spectrum analysis.Flotation test results indicate that DLZ is the selective collector of chalcopyrite.Especially,the recovery of chalcopyrite is higher than 90% in neutral and weak alkaline systems,while the recovery of pyrite is less than 10%.When using CaO as pH regulator,at pH=7-11,the floatability of pyrite is depressed and the recovery is less than 5%.Zeta potential analysis shows that the zeta potential of chalcopyrite decreases more obviously than that of pyrite after interaction with DLZ,confirming that collector DLZ shows selectivity to chalcopyrite and pyrite.And FTIR results reveal that the flotation selectivity of collector DLZ is due to chemical absorption onto chalcopyrite surface and only physical absorption onto pyrite surface.

Analysis of Texture of Froth Image in Coal Flotation

Froth image features of coal flotation have been extracted and studied by neighboring grey level dependence matrix, spatial grey level dependence matrix and grey level histogram. In this paper, a basic algorithm of unsupervised learning pattern classification is presented, and coal flotation froth images are classified by means of self organizing map (SOM). By extracting features from 51 flotation froth images with laboratory column, four types of froth images are classified. The correct rate of SOM cluster is satisfactory. And a good relationship of froth type with average ash content is also observed.

Process intensification of fine coal separation using two-stage flotation column

Flotation column is widely used as the separation equipment for fine mineral due to its high selectivity. However, this device may be unsuitable for the coarse particle flotation and has high handling ability. A two-stage flotation column with dimensions of 2 000 mm×1 000 mm×4 000 mm was designed to enhance the column flotation process. The energy input was modified by adjusting the flow rate and the head of circulating pump. The flotation column was designed with low energy input in the first stage(speed flotation stage) to recover easy-to-float materials quickly, and high energy input in the second stage(recovery stage) to recover difficult-to-float minerals compulsorily. Contrast experiments on the throughput and coarse coal recovery of high ash coal from the Kailuan Mine were conducted using conventional single-stage flotation column and the two-stage flotation column. The results show that the combustible matter recovery of the two-stage flotation column is 5.25% higher than that of the conventional single-stage flotation column. However, the ash contents of clean coal for both columns are similar. Less coarse coals with low ash are obtained using the two-stage flotation column than that using the single-stage column flotation with the same handling ability. The two-stage flotation column process can enhance coal flotation compared with the conventional single-stage column flotation.

Research on Factors Affecting Flotation and Desulfurization of Coal by Electrochemical Method

Influence of factors such as potential of electrolyte, electrolyzing time, concentration of electrolyte, and pH value on flotation and desulfurization of coal was researched. The result shows that the electrochemical reduction can enhance the desulfurization effect and increase the yield of clean coal under certain conditions. So it is an effective method.

Pico–nano bubble column flotation using static mixer-venturi tube for Pittsburgh No.8 coal seam

The flotation process is a particle-hydrophobic surface-based separation technique. To improve the essential flotation steps of collision and attachment probabilities, and reduce the step of detachment probabilities between air bubbles and hydrophobic particles, a selectively designed cavitation venturi tube combined with a static mixer can be used to generate very high numbers of pico and nano bubbles in a flotation column. Fully embraced by those high numbers of tiny bubbles, hydrophobic particles readily attract the tiny bubbles to their surfaces. The results of column flotation of Pittsburgh No. 8 seam coal are obtained in a 5.08 cm ID and 162 cm height flotation column equipped with a static mixer and cavitation venturi tube, using kerosene as collector and MIBC as frother. Design of the experimental procedure is combined with a statistical two-stepwise analysis to determine the optimal operating conditions for maximum recovery at a specified grade. The effect of independent variables on the responses has been explained. Combustible material recovery of 85–90% at clean coal product of 10–11% ash is obtained from feed of 29.6% ash, with a much-reduced amount of frother and collector than that used in conventional column flotation. The column flotation process utilizing pico and nano bubbles can also be extended to the lower limit and upper limit of particle size ranges, minus 75 lm and 300–600 lm, respectively, for better recovery.

Beneficiation of a low grade limestone sample

Pilot scale column flotation studies were conducted on a low grade siliceous limestone ore. Silica content was reduced to less than 13g in the concentrate so that it became satisfactory for use in the paper or rubber industries. The limestone sample was crystalline and constituted primarily of calcite that contained quartz, feldspar, pyroxene, and biotite as gangue minerals. Quartz is the major silicate gangue whereas feldspar, pyroxene, and biotite exist in minor to trace quantities. Traces of pyrite were also observed within the sample. A reverse flotation process was adopted where the silicate gangue minerals were floated using two different commercial cationic collectors： Chem-750 F or Floatamine-D. The studies clearly suggest it is possible to produce a limestone concentrate assaying around 96-97% CaCO3 containing less than 1 % Si02. The effect of feed flow rate, percent solids, froth depth, and wash water on the grade and recovery of the CaC03 concentrate is discussed.

Electrode process of diethyldithiocarbamate on surface of pyrrhotite

The electrode process of diethyldithiocarbamate on the surface of pyrrhotite was studied using systematic electrochemical analysis, including cyclic voltammetry, chronopotentiometry and galvanostatic. Experimental results show that tetraethylthioram disulphide（TETD） is electrodeposited on pyrrhotite electrode surface in the presence of 1.0×10^-4 mol/L diethyldithiocarbamate when the electrode potential is higher than 0.25 V. The electrochemical kinetics parameters of the electrode process of diethyldithiocarbamate on surface of pyrrhotite are calculated as follows： the exchange current density is 2.48μA/cm^2 , and the transmission coefficient is 0.46. The electrodeposition includes two steps electrochemical reaction. The first reaction is electrochemical adsorption of diethyldithiocarbamatc ion, then the adsorbed ion associates with a diethyldithiocarbamate ion from the solution and forms tetraethylthioram disulphide on the surface of pyrrhotite.

A novel approach to prevent bubble coalescence during measurement of bubble size in flotation

Effect of frothers in preventing bubble coalescence during flotation of minerals has long been investigated.To evaluate the performance of a frother,an apparatus to measure the bubble size is a basic necessity.McGill Bubble Size Analyzer(MBSA) or bubble viewer that has been developed and completed by McGill University's Mineral Processing Group during the last decade is a unique instrument to serve this purpose.Two parameters which are thought to influence the bubble size measurements by McGill bubble viewer include water quality and frother concentration in the chamber.Results show that there is no difference in Sauter mean(D32) when tap or de-ionized water was used instead of process water.However,the frother concentration,in this research DowFroth 250(DF250),inside the chamber exhibited a pronounced effect on bubble size.Frother concentration below a certain point can not prevent coalescence inside the chamber and therefore caution must be taken in plant applications.It was also noted that the frother concentration which has been so far practiced in plant measurements(CCC75-CCC95) is high enough to prevent coalescence with the bubble viewer.

Research on Effect of Magnetized Pulp on Coal Slime Flotation

Effect of magnetization on oxygen concentration, pH, surface zeta potential, and wet heat of flotation pulp were researched. The result shows that magnetization treatment can improve the floatablility of coal and increase the difference in wet heat among coal, refuse, and pyrite, which is favorable for slime flotation and for removing sulfur and ash from coal.

Flotation tree analysis for recovery of sillimanite from red sediments

In this paper an attempt is made to recover sillimanite by flotation tree analysis process and conventional flotation process from non magnetic fraction of red sediments.The experimental results of both the processes are presented.The data reveal that the deslimed sample contains 33.2%(by weight) total heavy minerals and out of which the sillimanite mineral content is 3.6%(by weight).It is observed that flotation tree analysis needs 10 cells to get five output products and where as conventional flotation process needs 15 cells to recover similar grade of five output products.Thus,flotation tree analysis is not only economic process but also efficient process(to say efficient process,the tree analysis product should be higher grade).

A dynamic size-based time series feature and application in identification of zinc flotation working conditions

Conventional feature description methods have large errors in froth features due to the fact that the image during the zinc flotation process of froth flotation is dynamic,and the existing image features rarely have time series information.Based on the conventional froth size distribution characteristics,this paper proposes a size trend core feature(STCF)considering the froth size distribution,i.e.,a feature centered on the time series of the froth size distribution.The core features of the trend are extracted,the inter-frame change factor and the inter-frame stability factor are given and two calculation methods of the feature factors are proposed.Meanwhile,the STCF feature algorithm was established based on the core features by adding the inter-frame change factor and the inter-frame stability factor.Finally,a flotation condition recognition model based on BP neural network was established.The experiments show that the recognition model has achieved excellent results,proving that the method proposed effectively overcomes the limitation of the lack of dynamic information in the existing traditional size distribution features and the introduction of the two factors can improve the classification accuracy to varying degrees.