基于LASSO方法的傅里叶变换红外光谱快速定性识别方法

发布时间：2018-06-12 02:08

  本文选题：LASSO + FTIR　； 参考：《光谱学与光谱分析》2017年10期

【摘要】：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,C_2H_4,NH_3,C_3H_8,C_4H_(10)和C_6H_(14)六种单组分傅里叶变换红外(FTIR)光谱吸光度谱以及一组C_2H_4和NH_3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。
[Abstract]:Infrared spectroscopy is used to monitor unknown gas components, which need to be identified qualitatively. LASSO variable selection technology based on multivariate linear regression model is widely used in data analysis field. In this paper, the LASSO method is introduced into the field of infrared spectrum analysis, and a qualitative identification method based on LASSO variable selection technique and cyclic linear least squares LCLS analysis is proposed, and the relevant experiments are carried out to verify it. Experimental collection of the absorbance spectra of the six kinds of one-component Fourier transform infrared (FTIRIRs) and a group of C _ 2H _ 4 and NH _ 3 mixed components, and a laboratory self-built spectral database, a preliminary qualitative analysis of the collected spectra was carried out by means of the LASSO method, after collecting the absorption spectra of the six kinds of one-component Fourier transform infrared (FTIRIRs) and a mixture of C _ 2H _ 4 and NH _ 3, combined with the self-built spectral database in the laboratory. Then the interference components are eliminated by LCLS method. The experimental results show that the method of LASSO combined with LCLS can effectively identify the target components in the spectrum, and most of the interference components can be eliminated even in the spectral bands with serious interference.
【作者单位】： 中国科学院安徽光学精密机械研究所 中国科学院环境光学与技术重点实验室;中国科学技术大学;
【基金】：国家重点研发计划项目(2016YFC0803001-08) 国家重大科学仪器设备开发专项(2013YQ22064302) 中国科学院前沿科学重点研究项目(QYZDY-SSW-DQC016) 国家自然科学基金项目(41405029)资助
【分类号】：O433.4
，

本文编号：2007775