激光诱导击穿光谱结合多变量回归法对石灰岩中主量元素的定量分析

石灰岩由于其优异的矿物学特性,在近年来已广泛用于冶金、制造、化学、建筑等领域。其中包含的碳酸盐与非碳酸盐成分如CaO、SiO₂、Fe₂O₃和MgO等对其工业应用起着重要作用。因此,为了实现对石灰岩中以上成分的准确定量,从而最大程度开发其商业价值,基于便携式激光诱导击穿光谱仪(LIBS-Tracer),分别结合单变量分析、偏最小二乘回归(PLSR)和主成分回归(PCR)对石灰岩样品中Ca、Si、Fe、Mg 4种主量元素进行定量分析。以交叉验证的结果作为多元回归模型参数寻优的标准,并以预测决定系数、预测均方根误差和测试集的相对标准偏差作为指标分别评估了上述3种回归模型的定量精度和稳定性。结果表明,多变量回归方法显著改善了传统单变量分析的定量效果。其中主成分回归表现最佳,4种目标元素分别达到了0.999 8、0.999 6、0.999 6和0.999 0的预测决定系数和0.066 6%、0.089 3%、0.014 8%和0.038 9%的预测均方根误差,测试样品中4种目标元素分别达到了1.00%、5.04%、5.03%和13.18%的相对标准偏差。研究表明,多变量回归模型不仅可以修正传统单变量分析由于基质效应、谱线干扰等影响所造成的定量精度偏差,还可以校正由于环境、硬件系统、样品等因素所导致的检测不稳定性。此外,主成分回归也可成为该便携式LIBS对于石灰岩样品中主量元素定量的可靠分析方法。

Quantification of major elements in limestones using laser-induced breakdown spectroscopy combined with multivariate regression approaches

Limestones had been widely used in metallurgy,manufacturing,chemistry,architecture,etc.in recent years due to its excellent mineralogical properties.The carbonate and non-carbonate components,such as CaO,SiO₂,Fe₂O₃ and MgO,played important roles in the industrial application of limestones.Therefore,in order to achieve accurate quantification of the components above in limestones and maximize their commercial value,the quantitative analysis of four major elements including Ca,Si,Fe and Mg was investigated based on a portable laser-induced breakdown spectrometer (LIBS) combined with univariate model,partial least square regression and principal component regression.The results of cross validation were used as the criteria for parameter optimization of multivariate regression models.The predictive determination coefficient,the root mean square error of prediction and the relative standard deviation of test set were applied to evaluate the quantitative accuracy and stability of these three regression models.The results showed that the multivariate regression methods improved the quantitative performance of traditional univariate analysis significantly.Among them,the principal component regression performed best.The four elements achieved the predictive determination coefficient of 0.999 8,0.999 6,0.999 6 and 0.999 0,and the root mean square errors of prediction of 0.066 6%,0.089 3%,0.014 8% and 0.038 9%,respectively,and the relative standard deviations of 1.00%,5.04%,5.03% and 13.18%,respectively.It was indicated that the multivariate regression model could not only correct the quantitative accuracy deviation of traditional univariate analysis caused by matrix effects and spectral line interference,but also adjust the detection instability caused by environment,hardware system and samples.Furthermore,principal component regression could be a reliable method for the portable LIBS quantitative analysis of major elements in limestone samples.