PSO-LSSVM对LIBS定量分析精度的提高

PSO-LSSVM Improves the Accuracy of LIBS Quantitative Analysis

针对土壤定量分析受基体效应影响大,LIBS定量分析精度不佳等问题,采用粒子群算法对LSSVM进行优化,提高模型的精确度。选取PbⅠ405.78 nm和CrⅠ425.44 nm作为分析谱线进行分析。采集十二个不同浓度样品的特征光谱,每个浓度样品在不同点采集20组数据,将其中17组数据设为训练集,3组数据设为预测集,用LSSVM和PSO-LSSVM两种方法建立定标模型。对比两种模型的拟合相关系数(R2)、训练集均方根误差(RMSEC)和预测集均方根误差(RMSEP)。由于自吸收效应的影响,随着浓度的增加,预测值逐渐低于实际值,LSSVM定标模型的拟合程度较低,无法达到实验要求,模型性能有待提高。利用粒子群算法对LSSVM的模型参数惩罚系数和核函数参数进行优化,得到最佳的参数组合,Pb元素为(8 096.8,138.865 7), Cr元素为(4 908.6,393.563 5),用最佳的参数组合构建LSSVM的定标模型。相比于LSSVM, PSO-LSSVM定标模型的精确度更高,Pb和Cr元素的R2提高到了0.982 8和0.985 0,拟合效果明显提升。Pb和Cr元素的训练集均方根误差由0.026 0 Wt%和0.027 2 Wt%下降到0.022 4 Wt%和0.019 1 Wt%,预测集均方根误差由0.101 8 Wt%和0.078 8 Wt%下降到0.045 8 Wt%和0.042 0 Wt%,模型的稳定性进一步提高。说明PSO-LSSVM算法能够更好地降低土壤基体效应和自吸收效应带来的影响,提高分析结果的精确度与稳定性。

Aiming at the problem that the quantitative analysis of soil is greatly affected by the matrix effect and the accuracy of the quantitative analysis of LIBS is not good. The particle swarm algorithm is used to optimize the LSSVM to improve the accuracy of the model. Pb Ⅰ 405.78 nm and Cr Ⅰ 425.44 nm was selected as the analysis lines for analysis. Collect the characteristic spectra of twelve samples with different concentrations. The LSSVM calibration model has a low degree of fitting and cannot meet the experimental requirements. The performance of the model needs to be improved. Use particle swarm optimization to optimize the model parameter penalty coefficient γ and kernel function parameter g of LSSVM to obtain the best combination of γ and g. The Pb element is(8 096.8, 138.865 7), and the Cr element is(4 908.6, 393.563 5). Compared with LSSVM, the accuracy of the PSO-LSSVM calibration model is higher. The R2 of Pb and Cr elements is increased to 0.982 8 and 0.985 0, and the fitting effect is significantly improved. The root means square error of the training set of Pb and Cr elements decreased from 0.026 0 Wt% and 0.027 2 Wt% to 0.022 4 Wt% and 0.019 1 Wt%, and the root means square error of the prediction set was reduced from 0.101 8 Wt% and 0.078 8 Wt% to 0.045 8 Wt% and 0.042 0 Wt%, the stability of the model is further improved. It shows that the PSO-LSSVM algorithm can better reduce the influence of the soil matrix effect and self-absorption effect, and improve the accuracy and stability of the analysis results.