论文部分内容阅读
为了提高激光诱导击穿光谱(LIBS)技术检测自然土壤中Pb的检测精度,提出采用间隔偏最小二乘法(IPLS)定量预测模型。对土壤在400~417nm波段的特征光谱进行平滑预处理后,建立偏最小二乘法(PLS)定量模型,得到训练集的相关系数为0.974 2,且斜率为0.983。建立IPLS模型时,把所选波段均分成了25个子区间,得到第八个子区间包含了Pb的特征光谱405.78nm,且交叉验证均方根误差最小,选择该区间建立模型得到训练集相关系数为0.985 3,斜率为1.121。预测集中,土壤样品Pb的真实浓度与预测浓度之间的相对误差在13%以内,平均相对误差为7.00%。研究表明IPLS法应用于LIBS定量检测土壤中的Pb是可行的,且该定量模型预测效果优于PLS法。
In order to improve the detection accuracy of Pb by laser induced breakdown spectroscopy (LIBS) in natural soils, a quantitative prediction model based on interval partial least squares (IPLS) was proposed. After the soil spectra were preprocessed in the 400 ~ 417nm band, a partial least squares (PLS) quantitative model was established. The correlation coefficient of the training set was 0.974 2 and the slope was 0.983. When the IPLS model is established, the selected bands are all divided into 25 sub-intervals. The eighth sub-interval contains the characteristic spectrum of Pb (405.78 nm), and the cross validation RMS is the smallest. The correlation coefficient of the training set 0.985 3, the slope is 1.121. In the prediction set, the relative error between the true and predicted concentrations of Pb in soil samples was within 13% with an average relative error of 7.00%. The results show that it is feasible to apply IPLS method to quantitatively detect Pb in soils by LIBS, and the quantitative model predicts better than PLS.