基于遗传算法和偏最小二乘法的土壤激光诱导击穿光谱定量分析研究

Quantitative Analysis of Soil by Laser-induced Breakdown Spectroscopy Using Genetic Algorithm-Partial Least Squares

在空气环境下,采用激光诱导击穿光谱(LIBS)技术对土壤成分进行检测,建立了基于遗传算法(GA)和偏最小二乘法(PLS)的定量分析模型。将配制的58个土壤样品分为定标集、监控集和预测集,对11种组分Mn,Cr,Cu,Pb,Ba,Al2O3,Ca O,Fe2O3,Mg O,Na2O和K2O的含量分别进行预测。结果表明,GA作为一种谱线选择的预处理方法,可以有效减少用于PLS建模的光谱谱线的数目,从而简化模型。对于土壤中的大部分组成成分,GA-PLS模型能够显著改善传统PLS模型的预测能力。以Mn元素为例,浓度预测均方根误差(RMSEP)从0.0215%降低至0.0167%,平均百分比误差(MPE)从8.10%降低至5.20%。本研究为进一步提高土壤的LIBS定量分析准确度提供了方法参考。

Laser-induced breakdown spectroscopy （LIBS） was used to detect the compositions of soil in the air, and the quantitative analysis model with genetic algorithm-partial least squares （GA-PLS） was established. A total of fifty-eight soil samples were split into calibration, monitoring and prediction sets. Eleven soil compositions including Mn, Cr, Cu, Pb, Ba, Al2O3, CaO, Fe2O3, MgO, Na2O, and K2O were quantitatively analyzed. The results demonstrated that, as a pretreatment method for optimizing the selection of spectral lines, GA could be effectively used to reduce the number of spectral lines for use in building PLS model, and hence simplify the quantitative analysis model. More importantly, for most of the soil compositions, GA-PLS could significantly improve the prediction ability compared with the conventional PLS model. Take Mn as an example, the root-mean-square error of prediction （RMSEP） was decreased from 0.0215% to 0.0167%, and the mean percent prediction error （MPE） was decreased from 8. 10% to 5.20%. The research provides an approach for further improving the accuracy of LIBS quantitative analysis in soil.