| The second-order calibration algorithm in chemometrics can quantify the components of interest in the presence of unknown interference.The second-order calibration is widely used in the fields of environment,medicine and food,and improves the efficiency and accuracy of its detection in traditional detection methods.The second-order calibration algorithms are optimized for different problems and many new algorithms are derived.For example,when the time shift and baseline drift in chromatographic data is heavy,work can use the algorithm which can tolerate nonlinear problems;when the spectrum of analytes is similar,work can use the algorithm which can handle the serious collinearity.Additionally,pattern recognition is also widely application in chemometrics.It can discriminate samples to achieve the effect of classification and identification.This paper applies the second-order calibration and pattern recognition to the Screening and Quantification of food and Chinese herb,and conducts the following research:In Chapter 2:In this chapter,the author uses alternating trilinear decomposition(ATLD)and multivariate curve resolution-alternating least squares(MCR-ALS)assisted high-performance liquid chromatography with diode array detection(HPLC-DAD)strategy for screening and quantifying twelve azo dyes illegally added into different food products.Under isocratic elution conditions,all analytes were successfully eluted within 6.5 min.With the aid of the prominent“second-order advantage”of the chemometric algorithms,the overlapping peaks and unknow interferences were solved.The spiked average recoveries obtained by ATLD and MCR-ALS were 82.8-111.5%and 78.7-122.6%,respectively.The limits of detection(LOD)obtained by ATLD and MCR-ALS were 0.01-2.56 mg kg-1and 0.01-2.95 mg kg-1,respectively.The results indicate that two algorithms have practical value for the simultaneous and rapid screening and determination of azo dyes in food products.Compared with other HPLC methods,the proposed method could flexibly handle different interferences,and has the advantages of rapid,efficient and environmentally friendly.In Chapter 3:In this chapter,this work uses alternating normatization-weighted error(ANWE)and unfold partial least squares/residual bilinearization(U-PLS/RBL)two algorithms assisted excitation-emission matrix fluorescence(EEM)to quantify two alkaloid components in Macleaya cordata.In the case of severe spectral overlap and other component interference,this proposed method uses mathematical separation to enhance the physical or chemical separation and can achieve rapid and accurate determination of analytes.The experiment also predicted the concentration of sanguinarine(SA)and chelerythrine(CHE)in the actual sample,the concentration of SA and CHE were 4.88±0.03 mg g-1and 1.56±0.03 mg g-1obtained by ANWE,respectively,and 4.99±0.01 mg g-1and 1.71±0.02 mg g-1obtained by UPLS/RBL,respectively.In addition,according to the experimental results such as spiked average recovery and analytical figures of merit,it shows that the two algorithms have good performance in quantification of SA and CHE.By comparing with other detection methods,this method is simple,rapid and environmentally friendly.In Chapter 4:In this chapter,the author uses EEM combined with chemical pattern recognition methods to discriminate the green and ripe forsythia.The experiment uses principal component analysis(PCA)and ANWE to extract features of the original data measured by EEM,where PCA selects the first 18 principal components and ANWE selects 5 factors.After that,the feature data obtained by the two methods were subjected to hierarchical clustering analysis(HCA)to perform clustering on 45 forsythia samples.The clustering results indicate that ANWE-HCA has better results,which due to the chemical meaning data obtained by ANWE.In addition,the experiment also uses partial least squares discriminant analysis(PLS-DA)to identify the green and ripe forsythia.And select 36 samples and 9 samples as the training set and the validation set from 45 samples,respectively.When the number of latent variables(LVs)is 19,the correct classification rate(CCR)obtained by using the leave-one method is the best.Therefore,the model was established when the number of LVs is 19.Then the model was used to predict the cross-validation set,training set and test set,and the CCRs were 75.0%,100.0%and 77.8%,respectively.In addition,the sensitivity and specificity are calculated,and the experimental results show that PLS-DA can discriminate between green and ripe forsythia. |
PDF Full Text Request