Research On Basic Theory And Innovative Application Of Chemical Multi-way Calibration And High-dimensional Pattern Recognition

With the development of modern analytical instruments,the data types obtained by analytical chemists are not only limited to traditional scalar and vector categories.More and more instrument response signals are high-dimensional chemical data,which consists of a large number of data points and contains abundant chemical information.These high-order data bring huge opportunities for the study of complex systems,however,how to fully analyze and use these high-dimensional data is also a considerable challenge.Fortunately,chemometrics can analyze complex chemical measurement data,and then fully extract the useful information,which allows analytical chemists to calmly meet this challenge.In the field of chemometrics,chemical multi-way calibration and pattern recognition are both research focuses and hotspots.By investigating their research status and development trends,the research works in this thesis focused on the novel quantitative analysis applications of high-order chromatographic instruments coupled with multi-way calibration,the efficient multi-way algorithms,and the development and application of chemical high-dimensional pattern recognition methods.The main contents of this thesis are as follows:Part Ⅰ:Chemical multi-way calibration assisted high performance liquid chromatography-diode array detector（HPLC-DAD）for the accurate quantitative analysis of complex systems（Chapter 2-Chapter 3）In Chapter 2,a simple,efficient,green and versatile analytical method,namely the ATLD-assisted HPLC-DAD method,was proposed for the fast and simultaneous determination of 12 main polyphenols in apple peel and pulp.The proposed method used“mathematical separation”to enhance the“physical and chemical separation”of chromatography,and it can achieve the quantification of target analytes in the presence of chromatographic peak overlap,unknown interference and baseline drift.The entire elution time took only 10 minutes,much shorter than that of traditional HPLC methods.In addition,the performance of the method was evaluated by statistical parameters such as recovery rate,sensitivity,selectivity,limit of detection（LOD）and limit of quantitation（LOQ）,and the quantitative results obtained by this method were compared with those obtained by traditional HPLC methods.The results show that the method is accurate,efficient,economical and relatively environmentally friendly,and it is expected to be extended for the efficient quantitative analysis of multiple analytes of interest in various complex systems.In Chapter 3,the ATLD-assisted HPLC-DAD method was used for the identification and quantitative analysis of 11 common NSAIDs in Chinese patent medicines and health products.The proposed method can decompose the three-way data array,extract pure signals of each component,and then obtain reliable qualitative and quantitative results even in the presence of serious overlapping peaks,unknown interference and baseline drift in chromatographic analysis,which benefits from the“second-order advantage”of this method.All 11 analytes were completely eluted in 14.5minutes without complicated sample preprocessing and chromatographic condition optimization.A positive sample was identified by this method,and the content of illegally added diclofenac（DCF）in the positive sample was as high as 18.38 mg g^-1.Compared with the traditional HPLC method,the chromatographic analysis time used in this method is greatly shortened,at the same time it can save a lot of organic solvents and reduce cost,which is in line with the goal of green analytical chemistry.The proposed method can be used as an alternative and attractive method for rapid identification and determination of illegally added substances in complex systems.Part Ⅱ:Development of novel chemical multi-way calibration algorithms（Chapter4 and Chapter 5）In Chapter 4,we proposed a simple second-order calibraion algorithm for directly handing second-order liquid chromatographic data with retention time shift.It can achieve qualitative and quantitative analysis of target analytes in the presence of overlapping peaks and unknown interference,which means that it also has the“second-order advantage.”Because the algorithm combines the principles and characteristics of the alternating trilinear decomposition（ATLD）algorithm and multivariate curve resolution（MCR）,it is called the alternating trilinear decomposition assisted multivariate curve resolution（ATLD-MCR）algorithm.ATLD-MCR was implemented by using the pre-decomposition results of ATLD as the initial values,MCR strategy for each sample slice matrix and the alternating least squares optimization strategy to obtain the final solution.Three simulated data sets,one semi-simulated LC-MS data set and one real HPLC-DAD data set were analyzed by the proposed algorithm.In addition,the qualitative and quantitative results obtained by ATLD-MCR were compared with those obtained by three classic second-order calibration algorithms.ATLD-MCR has obtained satisfactory qualitative and quantitative results for all data sets,which fully proves that it can correctly model second-order chromatographic data with retention time shift and severe signal overlap,and is expected to get more applications in chromatographic analysis.In Chapter 5,a novel,flexible and efficient quadrilinear decomposition algorithm（FACM）was proposed for four-way calibration.The FACM skillfully integrates the alternating quadrilinear decomposition（AQLD）algorithm with the four-way parallel factor analysis（FPARAFAC）algorithm and gives full play to their advantages by using them in different stages of the iterative process.By analyzing two simulated data sets,the performance of FACM and five existing four-way calibration algorithms（AQLD,SAQLD,RSWAQLD,APQLD,and FPARAFAC）were compared in detail.The results show that FACM has many attractive properties,such as very fast convergence,insensitive to initial values and excess number of components,and suitable for high noise level as well as severely collinear data.Its performance is better than a single algorithm.In addition,FACM was used to process two real four-way data arrays.The results show that FACM can accurately and efficiently extract qualitative and quantitative information of biomarkers and achieve satisfactory results in the presence of unknown interference and various backgrounds.Moreover,the structure of FACM is very flexible,therefore it can also be extended to combine other iterative algorithms besides AQLD and FPARAFAC,providing a feasible idea for the exploration and development of new algorithmsPart Ⅲ:Excitation-emission matrix fluorescence combined with chemical high-dimensional pattern recognition for authenticity identification of oil（Chapter 6）In Chapter 6,excitation-emission matrix fluorescence spectroscopy combined with a variety of chemometric methods were proposed for rapid identification and quantification of cheaper vegetable oil adulteration in camellia oil.Firstly,PARAFAC was used to characterize the spectra of different oil samples and obtain chemically meaningful information.Then,four chemometric high-dimensional pattern recognition methods（PARAFAC-LDA,PARAFAC-PLS-DA,（2D）²LDA,and N-PLS-DA）were used for the classification of camellia oil and other cheap vegetable oils（Model 1）and the classification of camellia oil and adulterated camellia oil（Model 2 and Model 3）,.Two-directional two-dimensional linear discriminant analysis（（2D）²LDA）was used in chemical data for the first time and showed huge potential.Furthermore,the N-PLS regression was used to predict the adulteration level in camellia oil.All the results confirm the reliability of the proposed methods.