针对浮选过程变量滞后、耦合特征及建模样本数量少所导致精矿品位难以准确预测的问题,提出了一种基于改进麻雀搜索算法(Improved Sparrow Search Algorithm,ISSA)优化混核最小二乘支持向量机(Hybrid Kernel Least Squares Support Vector Machine,HKLSSVM)的浮选过程精矿品位预测方法.首先采集浮选现场载流X荧光品位分析仪数据作为建模变量并进行预处理,建立基于最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)的预测模型,以此构建新型混合核函数,将输入空间映射至高维特征空间,再引入改进麻雀搜索算法对模型参数进行优化,提出基于ISSA-HKLSSVM方法实现精矿品位预测,最后开发基于LabVIEW的浮选精矿品位预测系统对本文提出方法实际验证.实验结果表明,本文提出方法对于浮选过程小样本建模具有良好拟合能力,相比现有方法提高了预测准确率,可实现精矿品位的准确在线预测,为浮选过程的智能调控提供实时可靠的精矿品位反馈信息.
Abstract
A flotation process concentrate grade prediction method based on the Improved Sparrow Search Algorithm (ISSA) optimized Hybrid Kernel Least Squares Support Vector Machine (HKLSSVM), a flotation process concentrate grade prediction method is proposed to address the issues of delayed variables, coupling characteristics, and limited modeling sample size in the flotation process, which make it difficult to accurately predict the concentrate grade. Firstly, collect data from the flotation site current carrying X-ray fluorescence grade analyzer as modeling variables and preprocess them to establish a prediction model based on the Least Squares Vector Machine. On this basis, a new mixed kernel function is constructed to map the input space to the high-dimensional feature space. Then, an Improved Sparrow Search Algorithm is introduced to optimize the model parameters, and an ISSA-HKLSSVM method is proposed to achieve concentrate grade prediction. Finally, a flotation concentrate grade prediction system based on LabVIEW is developed to verify the proposed method in practice. The experimental results show that the proposed method has a better fitting ability for small sample modeling in the flotation process. It can improve prediction accuracy compared to existing methods, and can achieve accurate online prediction of concentrate grade, providing real-time and reliable concentrate grade feedback information for intelligent control of the flotation process.
针对浮选过程精矿品位检测方法,国内外学者进行了大量的研究,主要分为基于机理的建模方法和基于数据的建模方法[3].浮选机理复杂,其过程涉及到化学、物理学且伴随较多随机因素,为准确描述真实浮选过程,传统基于机理的建模方法会对其进行一定的假设和简化[4].目前浮选产品指标的检测主要采用基于数据的建模方法[5],该方法只需在获得输入和输出数据后,利用误差最小化原理和统计分析方法拟合已有数据,而无需了解复杂的浮选内部机制,因而逐渐成为主流的浮选精矿品位检测方法.Kaartincal等[6]采集给矿流量、气量pH值等为输入变量,利用偏最小二乘法(Partial Least Squares, PLS)建立铜浮选过程精矿品位预测模型,得到较为满意的预测效果.Shean等[7]对比研究线性回归、PLS、神经网络等方法,以给矿品位、给矿浓度等作为建模参数建立精矿品位预测模型,对比结果显示基于PLS方法建模效果更好,可为现场操作人员提供有效指导.Cook等[8]利用泡沫大小、个数、RGB值及其他工况参数作为输入变量,在多层感知机基础上开展浮选精矿等级预测.Shahbazi等[9]以给矿品位值、精矿阀门开度、尾矿阀门开度为输入变量,引入动态前馈神经网络,使得浮选过程的关键性能指标预测的精度有一定提升.Allahkarami等[10]以药剂量、泡沫层厚度、进气量等作为输入,采用多层前馈神经网络构建铜精矿及钼精矿品位预测模型,其实测及仿真结果表明多层前馈神经网络相较线性回归预测建模更准确.
为解决上述问题,考虑浮选过程工艺参数的耦合性、滞后性及精矿品位小样本建模预测的需求,本文采集浮选现场载流X荧光品位分析仪数据,并在最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)方法基础上构建精矿品位预测模型.为了解决单一核函数在保证良好泛化性和较高预测准确度方面存在的困难,通过对现有经典核函数特征的分析,构建一种新型混合核函数,用于解决浮选精矿品位预测过程中输入特征空间的映射问题.同时为增强模型的自适应性,设计改进麻雀搜索算法(Improved Sparrow Search Algorithm,ISSA)对混核最小二乘支持向量机(Hybrid Kernel Least Squares Support Vector Machine,HKLSSVM)预测模型的参数进行优化,建立基于ISSA-HKLSSVM的浮选过程精矿品位预测模型,并基于某选矿厂浮选车间的实际生产工况验证本文提出方法的准确性和有效性.
为比较本文提出的基于ISSA-HKLSSVM方法建立的精矿品位预测模型的性能,将LSSVM方法用于建模讨论,其参数与ISSA-HKLSSVM预测模块中的保持一致,目的是为了对比分析引入的ISSA及新的混合核函数对精矿品位预测性能的影响.考虑到浮选工业过程中经常采用最小二乘法(Ordinary Least Squares,OLS)、PLS、BP神经网络等方法实现矿石品位建模,因此这三种方法也被用来建模比较.实验使用64位Windows处理系统,利用Matlab软件进行数据分解和模型预测的计算.详细的实验对比结果如表3所示.
GUIW H, YANGC H, XUD G, et al. Research progress in mineral flotation process monitoring technology based on machine vision [J]. Journal of Automation, 2013, 39(11): 1879-1888. (in Chinese)
LIUC X, DINGJ L, JIANGB, et al. Adaptive online support vector regression prediction model for concentrate grade of the ore-dressing processes [J]. Control Theory & Applications, 2014, 31(3):386-391. (in Chinese)
CAOB F, XIEY F, YANGC H, et al. Integrated modeling for production index of bauxite flotation based on multi-source data [J]. Control Theory & Applications, 2014, 31(9): 1252-1261. (in Chinese)
[10]
KAARTINCALJ, HATONENJ, HYOTYNIEMIH, et a1. Machine-vision-based control of zinc flotation: a case study [J]. Control Engineering Practice, 2006, 14(12): 1455-1466.
[11]
SHEANB J, CILLIERSJ J. A review of froth flotation control[J]. International Journal of Mineral Processing, 2011, 100(3): 57-71.
[12]
COOKR, MONYAKEK C, HAYATM B, et al. Prediction of flotation efficiency of metal sulfides using an original hybrid machine learning model [J]. Engineering Reports, 2020, 2(6): 12167.
[13]
SHAHBAZIB, CHELGANIS C, MATINS S. Prediction of froth flotation responses based on various conditioning parameters by random forest method [J]. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 2017, 529: 936-941.
[14]
ALLAHKARAMIE, NURIO S, ABDOLLAHZADEHA, et al. Estimation of copper and molybdenum grades and recoveries in the industrial flotation plant using the artificial neural network [J]. International Journal of Nonferrous Metallurgy, 2016, 5(3): 23-32.
[15]
WANGZ J, ZHAOZ, DONGL, et al. Data-driven soft sensor modeling for algal blooms monitoring [J]. IEEE Sensors, 2015, 15(11): 4681-4690.
[16]
LIUC, HUANGY H. Short term photovoltaic power prediction based on BP neural network optimized by improved sparrow search algorithm [C]// 2022 4th International Academic Exchange Conference on Science and Technology Innovation (IAECST), Guangzhou: IEEE, 2022:314-317.
[17]
HAAVISTOO, KAARTINENJ, HYOTYNIEMIH. Optical spectrum based estimation of grades in mineral flotation [C]// 2006 IEEE International Conference on Industrial Technology, Mumbai: IEEE, 2006:2529-2534.
JIANGY C, YANGX Q, HEF, et al. Super-short-time wind power forecasting based on EEMD-IGSA-LSSVM [J]. Journal of Hunan University(Natural Sciences), 2016, 43(10): 70-78. (in Chinese).
WANGW G, WEIY B, TENGX D. Short-term wind power forecasting based on VMD-SSA-LSSVM [J]. Acta Energiae Solaris Sinica, 2023, 44(3): 204-211. (in Chinese)
[24]
GAOB W, SHENW, HAOG, et al. Research on multistrategy improved evolutionary sparrow search algorithm and its application [J]. IEEE Access, 2022, 10:62520-62534.
MAOQ H, ZHANGQ, MAOC C, et al. Mixing sine and cosine algorithm with Levy flying chaotic sparrow algorithm [J]. Journal of Shanxi University(Natural Science Edition), 2021, 44(6):1086-1091. (in Chinese)
JIANGK, JIANGZ H, XIEY F, et al. Online prediction method for silicon content of molten iron in blast furnace based on dynamic attention deep transfer network [J]. Acta Automatica Sinica, 2023, 49(5): 949-963. (in Chinese)