Quality and Safety Detection of Edible Vegetable Oil Based on Time-resolved Fluorescence

Author:Chen Zuo

Supervisor:chen bin

Database:Doctor

Degree Year:2019

Download:4

Pages:116

Size:5444K

Keyword:

China is a large producer and consumer of edible vegetable oil and the largest importer of edible vegetable oil raw materials in the world.Recently,quality and safety events of edible vegetable oil often occur,such as high-price oil adulterated with low-price oil,repeated use of frying oil,dealing with oxidized deteriorated oil behind time,which affect the health of the consumer directly.Hence,there is great social and economic significance to establish rapid and simple methods for the quality and safety detection of edible vegetable oil to guarantee the health of countrymen and to regulate market behavior.In this study,rapid detection methods for adulteration,polycyclic aromatic hydrocarbon(PAHs)pollution and oxidation of 4 kinds of edible vegetable oils in large domestic market supply,namely rapeseed oil,peanut oil,sunflower seed oil and camellia oil,were developed and their detection mechanisms were analyzed based on time-resolved fluorescence(TRFS)analytical technique,and it provides a new solution for the application of this technology in the quality and safety detection of edible vegetable oil.The main research contents and results are as follows1.Research on the TRFS spectral characteristics of edible vegetable oil complex system.In order to select chemometric methods suitable for TRFS analysis,the TRFS spectral characteristics of edible vegetable oil were studied and analyzed through the emission dimension and time dimension by analyzing the instantaneous fluorescence emission spectra and fluorescence decay spectra of 4 edible vegetable oils.The results showed that:Non-linearity and strong fingerprint characteristics appeared in the TRFS of edible vegetable oil complex system,and it is necessary to select chemometric methods suitable for non-linear and fingerprint processing to establish eff-icient methods for the quality and safety detection of edible vegetable oil based on TRFS2.Research on the TRFS dimensionality reduction method based on discriminant autoencoders.In order to effectively reduce the dimensionality and to extract feature information from TRFS of edible vegetable oil complex system,the application of discriminant autoencoders(DAE)non-linear dimension reduction method was carried out,and then a rapid method for the adulteration detection of edible vegetable oil based on TRFS was established by combining with artificial neural network(ANN).The results showed that:DAE is a new method for dimension reduction and effective information extraction of TRFS in complex systems for its excellent ability of non-linear data processing,and therefore the DAE-ANN model showed the best performance for the adulteration detection of edible vegetable oil.The prediction root mean square error(RMSEP)of DAE-ANN for the adulteration detection of edible vegetable oil is lower than 0.0139 and the predictive correlation coefficient(Rp2)is higher than 0.9202;the RMSEP of DAE-ANN for the adulteration detection of reused frying oil in gradient amount is lower than 0.0141 and the Rp2 is higher than 0.9221;the RMSEP of DAE-ANN for the adulteration detection of reused frying oil in gradient frying time is lower than 0.0142 and the Rp2 is higher than 0.9222.It is showed that nonlinear dimension reduction and effective information extraction for complex system of edible vegetable oil TRFS could be achieved by using DAE,which provides a method for dimension reduction and characteristic extraction of other complex systems TRFS.3.Research on the prior knowledge-based TRFS multi-PAHs simultaneous detection method.In order to realize the simultaneous and rapid detection of multi-target analytes based on TRFS,fluorescence decay functions at characteristic wavelengths of PAHs were used as prior knowledge to replace part of Sigmoid excitation functions in DAE to guide and to strengthen DAE for the characteristic information extraction of PAHs in the fluorescence background of edible vegetable oil.The characteristic information attribution of PAHs was also enhanced by DAE-PARAFAC secondary projection because of the unique solution of parallel factor analysis(PARAFAC),a simultaneous and rapid method(DAE-PARAFAC-RVM)for the PAHs detection was established by combining with the correlation vector machine(RVM),which achieved the simultaneous and rapid detection of Chrysene(Chr),benzo[a]anthracene(B[a]A),benzo[a]pyrene(B[a]P),benzo[b]fluoranthene(B[b]F)in edible vegetable oil.The results showed that:The simultaneous extraction ability of DAE for the characteristic information in 4 kinds of PAHs was effectively improved by the application of prior knowledge,the attribution of 4 PAHs characteristic information was effectively enhanced by DAE-PARAFAC,and the prior knowledge-based DAE-PARAFAC-RVM model provides a new method for multi-objective analyte detection The RMSEP of model for Chr detection is lower than 0.0186,the Rp2 is higher than 0.9033,and the recovery is 94.8%~106.8%;the RMSEP of model for B[a]A detection is lower than 0.0175,the Rp2 is higher than 0.9104,and the recovery is 97.9%~103.2%;the RMSEP of model for B[a]P detection is lower than 0.0167,the Rp2 is higher than 0.9089,and the recovery is 97.4%~111.5%;the RMSEP of model for B[a]F detection is lower than 0.0171,the Rp2 is higher than 0.9022,and the recovery is 97.4%~104.3%.The sensitivity(SEN)of model for PAHs detection is higher than 51.9 μg/kg,the selectivity(SEL)is higher than 0.74 μg/kg,and the detection limit(LOD)is 0.115μg/kg.It is showed that the prior knowledge-based DAE-PARAFAC-RVM is an effective method for the simultaneous and rapid detection of multi-target analytes,which could be a beneficial reference for other simultaneous and rapid detection of multi-target analytes based on TRFS4.Research on the detection method for edible vegetable oil oxidation based on TRFS.In order to study the rapid detection method and detection mechanism for the edible vegetable oil oxidation based on TRFS,a comparative study between discriminant autoencoders-artificial neural network(DAE-ANN)and discriminant autoencoders-multidimensional partial least squares(DAE-N-PLS)for the detection of peroxide value(PV)and acid value(AV)of 4 kinds of edible vegetable oils was conducted;an analytical method for studying the influence of microenvironment on TRFS detection by using chlorophyll a as fluorescent probe was explored,and the mechanism of rapid detection method for vegetable oil oxidation based on TRES was revealed.The results showed that:the RMSEP of DAE-ANN for PV prediction of edible vegetable oil is lower than 0.0187 and the Rp2 is higher than 0.9023;and the RMSEP for AV prediction of edible vegetable oil is lower than 0.0178 and the Rp2 is higher than 0.9111.The RMSEP of DAE-N-PLS for PV prediction of edible vegetable oil is lower than 0.0185 and the Rp2 is higher than 0.9011;and the RMSEP for AV prediction of edible vegetable oil is lower than 0.0182 and the Rp2 is higher than 0.9077.Research showed that molecular fluorescence lifetime decreases with the increase of microenvironment polarity in the early stage of oxidation;while molecular fluorescence lifetime increases with the increase of microenvironment viscosity in the late stage of oxidation.Research has proved that:DAE-ANN and DAE-N-PLS are effective and rapid methods for the oxidation detection of edible vegetable oil,and the detection mechanism of rapid method for oxidation detection of edible vegetable oil based on TRFS was clarified preliminarily,which could be a reference for establishing other rapid methods and explaining mechanisms of detections based on TRFS.