激光诱导击穿光谱技术在提高矿冶分析准确度的研究进展

Research progress on laser-induced breakdown spectroscopy for improving the accuracy of mining and metallurgical analysis

激光诱导击穿光谱法(LIBS)是一种基于原子发射光谱的多元素分析方法,具有快速、准确、无需复杂的样品制备和远程分析的优点.然而,由于矿石、冶金样品化学成分的复杂性和多样性,干扰信号多,以及激光光谱的谱线维度较高和自吸收效应严重,LIBS技术在矿冶领域定性、定量分析的准确性受到了一定影响.本文综述了LIBS在矿冶领域3种信号增强方法,分别是双脉冲、纳米粒子增强和空间约束,以及综述了降噪、归一化和自吸收校正3种光谱预处理方法.此外,为提高定性、定量模型的泛化能力和分析的准确性,人们在模型算法和参数优化做了大量的工作.简要概述了主成分分析、偏最小二乘判别分析、支持向量机、随机森林和人工神经网络5种LIBS定性分析建模方法在矿石、冶金样品中的应用,以及概述了多元线性回归、偏最小二乘法、支持向量机、人工神经网络和自由定标法5种定量分析建模方法在矿石、冶金样品中的应用成果,并对LIBS技术未来在矿冶分析领域的发展进行了展望.

Laser-induced breakdown spectroscopy(LIBS)is a type of atomic emission spectroscopy for multi-element analysis.This analysis is rapid and accurate,has a simple sample preparation,and realizes remote analysis.However,the accuracy of the qualitative and quantitative LIBS analysis methods in the field of mining and metallurgy has suffered from the complexity and diversity of the chemical composition of ore and metallurgical samples,interference signals,high dimension of the laser spectrum line,and severe self-absorption effect.To enhance the accuracy of LIBS analysis in the mining and metallurgy field,researchers have conducted numerous research on signal enhancement,spectral pretreatment,and modeling methods.In this review,three signal enhancement methods of LIBS in mining and metallurgy are evaluated:double pulse,nanoparticle enhancement,and space constraint.To avoid noise interference,overfitting,and“self-erosion,”three spectral preprocessing methods,including noise reduction,normalization,and self-absorption correction,are also reviewed.Moreover,to improve the generalization ability and analysis accuracy of the qualitative and quantitative modeling methods,extensive research has been conducted on model algorithms and parameter optimization.This paper briefly outlines the application of five typical LIBS qualitative analysis modeling methods in ore and metallurgical samples:principal component analysis method,partial least squares discriminant analysis method,support vector machine,random forest,and artificial neural network,and application results of five quantitative analysis modeling methods in ore and metallurgical samples:multiple linear regression method,partial least square method,support vector machine,artificial neural network,and free calibration method.Currently,light element ores,such as phosphate and lithium ores,rare earth and scattered elements,and the combined use of instruments are rarely investigated using LIBS;thus,future developments in LIBS technology for mineral and metallurgical analysis should mainly focus on the following aspects:(1)Research on LIBS online monitoring technology and suitable instrumentation because the application of online real-time and in situ monitoring analysis in the mining and metallurgical process has not been fully achieved.(2)Application of this method for the rapid analysis of light elements and complex ore and metallurgical samples,especially for online analysis,under special environmental conditions.(3)Improvement in the accuracy of the LIBS analysis and its application range in combination with other analytical techniques,such as Raman spectroscopy and infrared spectroscopy.