采用红外光谱技术对未知气体组分进行监测, 需要对气体组分进行定性识别分析。 基于多元线性回归模型的LASSO变量选择技术广泛应用于数据分析领域。 将LASSO方法引入到红外光谱分析领域, 提出一种LASSO变量选择技术结合循环线性最小二乘(LCLS)分析的定性识别方法, 并开展了相关的实验对其进行验证。 实验采集CO, C2H4, NH3, C3H8, C4H10和C6H14六种单组分傅里叶变换红外(FTIR)光谱吸光度谱以及一组C2H4和NH3混合组分的吸光度谱, 结合实验室自建光谱数据库, 先采用LASSO方法对采集的光谱进行初步定性分析, 然后使用LCLS方法剔除干扰组分。 实验结果表明, LASSO结合LCLS的方法能有效识别出光谱中的目标组分, 即使是在干扰严重的光谱波段也可以剔除掉大部分的干扰组分。

In the case of infrared spectroscopy for the monitoring of unknown gas components, it is necessary to qualitatively identify the gas components. LASSO variable selection technique based on multiple linear regression model is widely used in data analysis. In this paper, the LASSO method is introduced into the field of infrared spectroscopy, and a qualitative identification method based on LASSO variable selection technique combined with cyclic least squares (LCLS) analysis is proposed and verified by relevant experiments. The absorbance spectra of six components of CO, C2H4, NH3, C3H8, C4H10, C6H14, and the absorbance spectra of a mixture of C2H4 and NH3 were measured. The LASSO method was used to analyze the collected spectra with the spectra library build by our lab, and then the LCLS method was used to eliminate the interfering components. The experimental results show that LASSO combined with LCLS can effectively identify the target components in the spectrum, even in the spectral band with serious interference, most of the interference components can also be removed.