激光诱导击穿光谱结合随机森林的稀土矿石中钪元素定量分析
Sc element is an indispensable strategic resource that is widely used in the manufacture of solid oxide fuel cells, ceramic materials, catalysts, and lightweight high-temperature alloys. Sc mainly exists in minerals, such as rare-earth ores, bauxites, uranium ores, and manganese iron ores. However, its scarcity (the Sc content in the crust accounting for approximately 0.0005%) and the high cost of extraction lead to the current shortage of Sc in the rare-earth market. Therefore, quantitatively analyzing the Sc content in rare-earth ores holds great significance for exploring and mining these ores. Laser-induced breakdown spectroscopy (LIBS) is an analytical detection technology based on the atomic emission spectrum of plasma generated by laser ablation of samples. The characteristic emission spectrum produced by the laser ablation of the sample surface determines the elemental composition and content of the samples. Compared to other technologies, LIBS offers numerous advantages such as no sample preparation, simultaneous multi-element analysis, on-site rapid detection, remote detection, real-time online detection, and microdamage detection. It is extensively used in various fields such as metallurgical analysis, geological exploration, environmental monitoring, and space exploration. However, the spectrum’s stability is affected by the sample’s matrix effect, the unevenness of the sample surface, and changes in the detection environment during the LIBS detection process. In recent years, multivariate correction methods, such as random forest (RF), partial least squares (PLS), artificial neural network (ANN), and support vector machine (SVM) successfully address and resolve these issues. These methods enhance the accuracy and precision of LIBS spectral analysis, yielding impressive results. RF is an integrated learning method that is based on a regression tree, stands out for its resistance to overfitting, high accuracy, and straightforward optimization of model parameters. It is in widespread use in LIBS spectral analysis. This study aims to explore the feasibility of combining LIBS with RF for the quantitative analysis of Sc in rare-earth ores.
In this study, a method for quantitative analysis of Sc in rare-earth ores based on laser-induced breakdown spectroscopy (LIBS) combined with a random forest (RF) algorithm had been proposed. Six rare-earth ore standard samples were prepared, and the reference values for the Sc content in rare-earth ore samples are shown in Table 1. Each powder sample was compressed into thin slices using a tablet press at 20 MPa for 5 min before the LIBS spectra were collected. Next, an LIBS device was set up to collect the spectra. To enhance the stability of the LIBS spectra of rare-earth ore samples, 16 sampling points were randomly chosen for LIBS spectra collection for each compressed piece. The LIBS spectra were obtained by accumulating five laser pulses and averaging them to minimize the effect of laser pulse fluctuations. A total of 96 spectral data points were collected from the six rare-earth ore samples (16 spectra for each sample). The influence of different spectral preprocessing methods on the performance of the RF model was then investigated, and variable importance measurement (VIM) was applied for feature screening and parameter optimization for the RF calibration model. To further validate the prediction performance of the RF model, it was compared with other models. Finally, a VIM-RF calibration model was established based on the optimized spectral preprocessing and VIM threshold conditions.
First, the influence of different spectral preprocessing methods on the performance of the RF model is investigated. As Table 2 demonstrates, the RF correction model processed by the WT (where the wavelet basis function is Coif5, and the decomposition level is 1) realizes better cross-validation and internal validation results within the correction set. Compared to the original spectral model, the cross-validation results of the WT-RF correction model show an increase in
This study establishes a method for detecting Sc in rare-earth ores based on LIBS combined with VIM-RF. First, the influence of different spectral preprocessing methods on the performance of the RF model is investigated. Then, variable importance measurement (VIM) is applied for feature screening and parameter optimization for the RF calibration model. To further verify the prediction performance of the RF model, it is compared with other models. Finally, a VIM-RF calibration model is established based on the optimized spectral preprocessing (WT) and VIM threshold (0.016) conditions. The results show that the VIM-RF model exhibits excellent prediction performance (
1 引言
稀土元素具有独特的磁性、发光和催化等性能,在**、冶金、农业、玻璃、石油化工、陶瓷等领域得到了广泛应用,特别是在电子产品、汽车工业、航空航天等高新产业有着不可替代的作用,被誉为“工业维生素”[1]。在稀土元素中,钪(Sc)凭借其良好的物理化学属性常被用来制造半导体器件、计算机磁芯存储器、导弹、轻质高温合金和固体氧化燃料等,是至关重要的稀缺战略资源[2]。自然界中的Sc主要存在于稀土矿、铝土矿、铀矿和锰铁矿等矿物中[3]。然而,其稀缺性(Sc在地壳中的质量分数约占0.0005%)以及提取Sc的高昂成本导致目前其在稀土市场上的供应比较紧张[4]。因此,稀土矿中Sc元素的定量分析对于稀土产业勘探、采矿、工业原料提取和产品质量检查等具有重要意义。
激光诱导击穿光谱(LIBS)是一种通过激光烧蚀样品表面产生等离子体进而对等离子体的原子发射光谱进行检测的技术,通过分析激光烧蚀样品表面产生的特征发射光谱可以确定样品的元素组成以及待分析元素的含量。与其他技术相比,LIBS具有现场快速检测、远程检测、实时在线检测和微损检测等特点,而且不需要制备样品,可以同时进行多元素分析,已被广泛应用于冶金分析、地质勘探、环境监测、太空探索[5-8]等领域。特别是在地质勘探领域,LIBS可以获得矿物中元素的成分信息并支持矿物的分类与鉴别[9],已被成功应用于铝土矿、铁矿石、锰矿石、铬矿石等矿石的定性及定量分析[9-10]。近年来,LIBS在稀土元素分析中的应用越来越多,如:刘小亮等[11]将LIBS应用于稀土矿区土壤中Sm含量的测定,该技术对Sm元素两条分析线的预测相对误差分别为0.97%和0.26%。Devangad等[12]将LIBS技术应用于磷酸盐玻璃中稀土元素Sm、Tm和Yb的定量分析,三种元素的相关系数R2分别为0.9980、0.9930、0.9780。上述研究表明,LIBS是一种可行的稀土元素分析方法。
在LIBS检测过程中,光谱的稳定性会受到样品基体效应、样品表面不均匀性及检测环境变化等的影响。因此,传统的单变量分析方法在LIBS数据处理时无法获得理想的分析结果。近年来,多元校正方法,如随机森林(RF)、偏最小二乘(PLS)、人工神经网络(ANN)、支持向量机(SVM)等,已被应用于提高LIBS光谱分析的精确度及准确度,并取得了较好的分析结果[13-17]。RF是一种基于回归树的集成学习方法,具有不易过拟合、准确率高、模型参数优化简单等优势,已被广泛应用于LIBS光谱分析。Li等[18]将LIBS与RF算法结合并应用于含油污泥样本污染风险因子的预测,预测集的R2和均方根误差(RMSEP)分别为0.9681和0.6009。李铸等[19]将LIBS与RF算法结合并应用于高碳铬轴承钢(GCr15钢)硬度的定性分析与识别,对于未知硬度的样品,RF模型的平均预测准确率可达96%。LIBS光谱数据通常包含着大量的变量数据,其中包括待分析元素的特征信息以及大量的冗余变量。因此,在构建基于LIBS数据的分析模型前需要进行变量筛选。变量重要性测量(VIM)是一种基于RF算法的变量选择方法,已被成功应用于LIBS光谱数据的变量筛选上。它计算所有变量对模型预测性能的贡献度,即变量的重要性,然后通过不断优化变量的重要性阈值来筛选、保留对模型预测性能贡献度高的变量,从而提高模型的预测性能并缩短建模时间。丁宇等[20]将VIM应用于铝合金中Mg元素的定量分析,结果表明,基于VIM筛选所构建的RF校正模型的预测性能优于基于全谱构建的RF校正模型的预测性能,其预测集
本文拟探究LIBS技术结合RF算法用于稀土矿石中Sc元素定量分析的可行性。首先,采集稀土矿石标准样品的LIBS光谱,并基于美国国家标准与技术研究院(NIST)的原子光谱数据库对稀土矿石样本中Sc元素的LIBS发射谱线进行识别;其次,考察不同光谱预处理方法(标准正态变换、归一化、一阶导数、二阶导数和小波变换)对RF校正模型预测性能的影响;然后,利用VIM进行RF校正模型输入变量的选择与优化,并基于留一法交叉验证对VIM的阈值进行优化。为了进一步验证该模型的预测性能,将其与标准曲线法、偏最小二乘(PLS)以及基于波段选择的RF模型进行了比较。最后,在最优化的光谱预处理和VIM阈值等条件下,建立针对稀土矿石中Sc定量分析的RF校正模型。
2 实验过程
2.1 样品制备
6个稀土矿石标准样品购于武汉睿辰标物科技有限公司,稀土矿石样品中Sc含量的参考值见
表 1. 稀土矿石样品中Sc元素含量的参考值
Table 1. Reference value of Sc content in rare earth ore samples
|
2.2 LIBS光谱采集
实验中稀土矿石样品LIBS光谱的采集装置主要由Nd∶YAG激光器(Nano,Litron,UK,波长为1064 nm,脉冲能量为80 mJ,激光频率为5 Hz)、样品台、光路系统、光纤、光谱仪(Aryelle 200,LTB,Germany,光谱范围为190~840 nm,分辨率为0.015 nm,门宽为10 μs)、增强型电荷耦合检测器(iStar 340t,Andor Technology,UK)及计算机系统等构成。在检测过程中,将稀土样品放置于X-Y-Z手动可调样品台上,在实验室环境下进行LIBS光谱的采集,延迟时间为1 μs。
为了提高LIBS光谱的稳定性,降低样品不均匀性对稀土矿石样品LIBS光谱准确度和精度的影响,每个压片随机选择16个采样点进行LIBS光谱的采集(点位尽可能在样品表面上均匀分布),每个采样点通过累计进行5次激光脉冲作用后求5张光谱平均值的方式获得LIBS光谱,以降低激光脉冲波动的影响。6个稀土矿石样品共采集96条光谱数据(每个样品采集16条光谱)。构建RF模型时,选取#3、#4、#5及#6样品作为校正集,用于构建RF校正模型;选取#1及#2样品作为预测集,用于验证RF模型的预测性能。
2.3 随机森林
RF是Breiman[21]在2001年提出的一种将多个决策树集成来构建预测模型的统计学习理论,具有较好的模型稳定性,常被用于处理分类和回归问题。RF在模型构建过程中有两个重要的参数,即决策树节点数(mtry)和决策树数(ntree)。本文的研究思路及RF建模流程如
3 分析与讨论
3.1 稀土矿石样品的LIBS光谱分析
在RF校正模型构建之前,需要对Sc的LIBS光谱特征谱线进行识别。本研究根据NIST的原子光谱数据库对稀土矿石样品LIBS光谱中Sc元素的特征谱线进行识别。
图 2. #4稀土矿石样品的LIBS光谱图。(a)200~320 nm;(b)320~380 nm;(c)380~420 nm;(d)420~840 nm
Fig. 2. LIBS spectrum of #4 rare-earth ore sample. (a) 200‒320 nm; (b) 320‒380 nm; (c) 380‒420 nm; (d) 420‒840 nm
3.2 预处理方法的选择与优化
LIBS光谱常用的预处理方法主要包括标准正态变换(SNV)、归一化(Nor)、一阶导数(D1st)、二阶导数(D2nd)和小波变换(WT)等。其中:SNV常用于校正由样品颗粒大小不同及颗粒分布不均匀等导致的散射所引起的光谱误差;Nor常用于校正由光谱数据尺度差异过大带来的误差;D1st与D2nd常用于降低仪器背景或基线漂移对信号的影响;WT可同时对信号的时间和频率进行分析,常用于光谱数据的平滑滤噪、去除背景等。从
本研究将探讨上述5种预处理方法对稀土矿石样品LIBS光谱的影响。首先分别采用上述5种方法对稀土矿石的LIBS光谱进行预处理;然后,将经过预处理优化的LIBS光谱作为输入变量,构建RF校正模型,并通过对比不同RF校正模型的性能来寻找最优的预处理方法。在光谱预处理中,Nor将LIBS光谱的强度最大值作为参考进行归一化;D1st与D2nd中可选择的平滑点数为3、5、7、9、11;WT中可选择的小波基函数为Coif1、Coif2、Coif3、Coif4、Coif5、Db2、Db3、Db4、Db5,分解层数为1、2、3、4、5。本研究采用留一法交叉验证对上述预处理方法的参数进行优化。
图 3. 不同参数条件下D1st-RF与D2nd-RF校正模型的预测性能。(a)D1st-RF交叉验证结果;(b)D1st-RF内部验证结果;(c)D2nd-RF交叉验证结果;(d)D2nd-RF内部验证结果
Fig. 3. Predictive performance of first derivative-random forest (D1st-RF) and second derivative-random forest (D2nd-RF) calibration models under different parameters. (a) D1st-RF cross validation; (b) D1st-RF internal validation; (c) D2nd-RF cross validation; (d) D2nd-RF internal validation
图 4. 不同参数条件下WT-RF校正模型的预测性能。(a)(b)WT-RF交叉验证结果;(c)(d)WT-RF内部验证结果
Fig. 4. Predictive performance of wavelet transform-random forest (WT-RF) calibration model under different parameters. (a)(b) WT-RF cross validation; (c)(d) WT-RF internal validation
表 2. 基于不同预处理方法的RF校正模型的预测性能
Table 2. Predictive performance of RF calibration model based on different pretreatment methods
|
3.3 输入变量的选择与优化
对稀土矿石样品的LIBS原始光谱进行预处理并不会减少数据的变量数,变量数依然为43399,这些变量不仅包括Sc含量定量分析所需的有效信息,还包括大量的无效或噪声信息。这些无效信息不仅会影响模型的预测精度,还会增加建模时间。因此,在模型构建前,需要对输入变量进行筛选,以确保模型的预测性能。本研究考察了VIM对基于稀土矿石样品LIBS光谱的RF校正模型预测性能的影响。
从
表 3. 基于VIM的WT-RF校正模型的预测性能对比
Table 3. Comparison of the predictive performance of WT-RF calibration models based on VIM
|
3.4 不同校正模型预测性能的比较
为了进一步验证VIM-RF校正模型对稀土矿中Sc元素含量的预测性能,将VIM-RF校正模型与标准曲线法、PLS、RF校正模型的预测性能进行比较。同时,为了验证变量选择方法对Sc元素含量的预测性能,考察了基于波段的变量选择方法对RF校正模型预测性能的影响。在标准曲线法构建过程中,分析线通常优先选择具有相对强度较大、跃迁概率大的特征谱线,而且应避免选择距离较近、易出现叠加的特征谱线。笔者根据NIST原子光谱数据库选择Sc I 327.362 nm作为分析线。在PLS校正模型构建过程中,将WT(小波基函数为Coif5,分解层数为1)预处理后的光谱作为PLS校正模型的输入变量,并采用留一法交叉验证对模型参数潜变量进行优化,最终选择最优潜变量为6。Sc元素的特征峰主要处于200~650 nm范围内,因此在基于波段选择的RF校正模型的构建过程中,在200~650 nm范围内对输入波段进行筛选,将WT(小波基函数为Coif5,分解层数为1)预处理后的光谱作为输入变量并采用留一法交叉验证对其波段进行优化。对于Sc元素来说,其对应的最优波段为300~400 nm。
图 6. 不同校正模型预测性能的比较。(a)留一法交叉验证;(b)预测集
Fig. 6. Comparison of predictive performance of the different calibration models. (a) Leave-one-out cross validation; (b) prediction set
图 7. VIM-RF校正模型对稀土矿中Sc含量的预测结果
Fig. 7. Predictive content of Sc in the rare-earth ores with VIM-RF calibration model
4 结论
笔者建立了一种基于LIBS结合VIM-RF的稀土矿石中Sc元素含量的检测方法。首先,探究不同光谱预处理方法对RF模型性能的影响,以获得一种最优的预处理方法。当选择WT(小波基函数为Coif5,分解层数为1)作为预处理方法时,RF模型的交叉验证结果相较于原始光谱构建的RF模型有较好的提升,
[1] Luo X P, Zhang Y B, Zhou H P, et al. Review on the development and utilization of ionic rare earth ore[J]. Minerals, 2022, 12(5): 554.
[2] 刘峰, 胡文鑫. 钪在铝、镁合金中的应用研究进展[J]. 稀土, 2019, 40(2): 130-141.
Liu F, Hu W X. Research progress in application of Sc-containing aluminum and magnesium alloys[J]. Chinese Rare Earths, 2019, 40(2): 130-141.
[3] 范亚洲, 周伟, 王子玺, 等. 稀散元素Sc的矿床类型及找矿前景[J]. 西北地质, 2014, 47(1): 234-243.
Fan Y Z, Zhou W, Wang Z X, et al. The types of rare and disperse scandium deposits and prospecting potential[J]. Northwestern Geology, 2014, 47(1): 234-243.
[4] Dutta T, Kim K H, Uchimiya M, et al. Global demand for rare earth resources and strategies for green mining[J]. Environmental Research, 2016, 150: 182-190.
[5] Dong F Z, Chen X L, Wang Q, et al. Recent progress on the application of LIBS for metallurgical online analysis in China[J]. Frontiers of Physics, 2012, 7(6): 679-689.
[6] Harmon R S, Lawley C J M, Watts J, et al. Laser-induced breakdown spectroscopy—an emerging analytical tool for mineral exploration[J]. Minerals, 2019, 9(12): 718.
[7] Zhang Y, Zhang T L, Li H. Application of laser-induced breakdown spectroscopy (LIBS) in environmental monitoring[J]. Spectrochimica Acta Part B: Atomic Spectroscopy, 2021, 181: 106218.
[8] Gasda P J, Acosta-Maeda T E, Lucey P G, et al. Next generation laser-based standoff spectroscopy techniques for Mars exploration[J]. Applied Spectroscopy, 2015, 69(2): 173-192.
[9] 邱苏玲, 李安, 王宪双, 等. 基于激光诱导击穿光谱的矿石中铁含量的高准确度定量分析[J]. 中国激光, 2021, 48(16): 1611002.
[10] Carvalho A A C, Alves V C, Silvestre D M, et al. Comparison of fused glass beads and pressed powder pellets for the quantitative measurement of Al, Fe, Si and Ti in bauxite by laser-induced breakdown spectroscopy[J]. Geostandards and Geoanalytical Research, 2017, 41(4): 585-592.
[11] 刘小亮, 孙少华, 孟祥厅, 等. 激光诱导击穿光谱法测定稀土矿区土壤中钐含量[J]. 中国光学, 2022, 15(4): 712-721.
[12] Devangad P, Unnikrishnan V K, Nayak R, et al. Performance evaluation of laser induced breakdown spectroscopy (LIBS) for quantitative analysis of rare earth elements in phosphate glasses[J]. Optical Materials, 2016, 52: 32-37.
[13] 李正风, 徐广晋, 王家俊, 等. 模型诊断用于近红外光谱建模校正集中奇异样本的识别[J]. 分析化学, 2016, 44(2): 305-309.
Li Z F, Xu G J, Wang J J, et al. Outlier detection for multivariate calibration in near infrared spectroscopic analysis by model diagnostics[J]. Chinese Journal of Analytical Chemistry, 2016, 44(2): 305-309.
[14] 袁自然, 魏立飞, 张杨熙, 等. 优化CARS结合PSO-SVM算法农田土壤重金属砷含量高光谱反演分析[J]. 光谱学与光谱分析, 2020, 40(2): 567-573.
[15] Borin A, Ferrão M F, Mello C, et al. Least-squares support vector machines and near infrared spectroscopy for quantification of common adulterants in powdered milk[J]. Analytica Chimica Acta, 2006, 579(1): 25-32.
[16] Liu K, Tian D, Xu H Y, et al. Quantitative analysis of toxic elements in polypropylene (PP) via laser-induced breakdown spectroscopy (LIBS) coupled with random forest regression based on variable importance (VI-RFR)[J]. Analytical Methods, 2019, 11(37): 4769-4774.
[17] 孟德硕, 赵南京, 马明俊, 等. LIBS结合ANN对不同类型土壤中的Cu的定量检测[J]. 光电子·激光, 2015, 26(10): 1984-1989.
Meng D S, Zhao N J, Ma M J, et al. Quantitative detection of Cu in different types of soils using laser induced breakdown spectroscopy combined with artificial neural network[J]. Journal of Optoelectronics·Laser, 2015, 26(10): 1984-1989.
[18] Li M G, Fu H, Du Y, et al. Laser induced breakdown spectroscopy combined with hybrid variable selection for the prediction of the environmental risk Nemerow index of heavy metals in oily sludge[J]. Journal of Analytical Atomic Spectrometry, 2022, 37(5): 1099-1108.
[19] 李铸, 张庆永, 孔令华, 等. 基于激光诱导击穿光谱与随机森林识别GCr15钢的硬度[J]. 中国激光, 2022, 49(9): 0911002.
[20] 丁宇, 杨淋玉, 陈靖, 等. 基于激光诱导击穿光谱法的铝合金中Mg元素定量分析[J]. 激光与光电子学进展, 2022, 59(13): 1314006.
[21] Breiman L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32.
Article Outline
周嘉俊, 李茂刚, 张天龙, 汤宏胜, 李华. 激光诱导击穿光谱结合随机森林的稀土矿石中钪元素定量分析[J]. 中国激光, 2024, 51(2): 0211001. Jiajun Zhou, Maogang Li, Tianlong Zhang, Hongsheng Tang, Hua Li. Quantitative Analysis of Sc in Rare‐Earth Ores via Laser‐Induced Breakdown Spectroscopy Combined with Random Forest[J]. Chinese Journal of Lasers, 2024, 51(2): 0211001.