The Effect Of Baseline On Qualitative Results Of Complex Multicomponent Systems

In this thesis,target factor analysis(TFA)and moving window target factor analysis(MWTFA)are used to process open path Fourier transform infrared(OP/FTIR)spectra to detect complex components in atmosphere.Because baseline removal process affects target information and TFA has strong anti-interference ability,generally TFA can identify the target molecule without baseline correction.However,when baseline drift is serious,TFA and MWTFA give false negative results.The effects of standard normal variate transform(SNV),linear fitting,multiplicative scatter correction(MSC),adaptive iteratively reweighted penalized least squares(airPLS)and wavelet transform(WT)on TFA and MWTFA are compared.Results show that wavelet transform can effectively remove baselines,thus decrease false negative results of TFA and MWTFA.The research mainly includes the following three parts:1.OP/FTIR spectra are measured continuously near a farm to obtain three groups of data matrices with different baselines.NH3 is taken as the research object in low wavenumber region,which is 750～1250 cm-1.Orthogonal projection approach least squares(OPALS)is used to determine the number of principal factors in the spectral matrix,then TFA extracts NH3 information to confirm the presence of NH3 in air.Partial least squares(PLS)is used to predict concentrations of NH3 in the spectra,and the variance is calculated.Results show that when the concentration variance of NH3 is small and baseline drift is serious,false negative results of TFA are observed.Various baseline correction methods on TFA are compared in detail.It is found that wavelet transform could significantly remove baseline information,and make spectral information of NH3 appear in the eigenvector of PCA in advance,which reduce false negative results of TFA.2.Six data matrices are constructed using OP/FTIR spectra measured near the farm.TFA is used to quantify molecules in high wavenumber region,such as CH4 and CO.Wavenumber regions for CH4 and CO are 2850～3200 cm-1 and 2030～2240 cm-1,respectively.SNV,MSC,linear fitting,airPLS and wavelet transform are used to preprocess the spectral matrices when TFA yields false negative results.Results show that baseline information is largely removed by wavelet transform,and the reliability of TFA is improved significantly.3.By setting the appropriate window width,MWTFA is used to qualitatively analyze OP/FTIR spectra of NH3,the time region with NH3 and the time region without NH3 can be correctly distinguished.When baseline affects the qualitative ability of MWTFA,SNV,MSC,linear fitting,airPLS and wavelet transform are used to correct baseline.Results show that wavelet transform can effectively improve the qualitative ability of MWTFA.In addition,MWTFA can be used in experimental spectra of CH4 and CO accurately.