| Chinese liquor,as one of the six world-renowned distilled liquors,is the unique liquor produced in China,which gradually plays an important role in people's daily life and communication.The existing methods of evaluating the liquor include sensory evaluation and physico-chemical analysis,which can't provide real-time information and rapid detection because of some disadvantages,such as high subjectivity,cumbersome methods,bad timeliness and high cost.In recent years,spectroscopy has been widely used in the field of food identification for it's convenient operation,nondestructive testing,high sensitivity and good reproducibility.In this research,the spectroscopy(NIR,FT-IR,UV-Vis,Raman spectra)were used for classifing different grades of base liquor and quantitatively analyzing the main esters combined with chemometrics,which means to build an integrated system to detect the quality of strong-flavor base liquor rapidly.Specific content and conclusions are as follows:Firstly,strong-flavor base liquor samples were collected and the sensory evaluation were conducted on the samples.A total of 75 base liquor samples from different workshops and different teams were collected,and three grades were obtained after the preliminary classification by workers.Then,the strict rating were implemented by professionals,and five grades of liquor were collected eventually,including the good quality of the first class,the slightly worse quality liquor of the first class,the good quality of the second class,the slightly worse quality liquor of the second class and the third class.Secondly,different grades of strong-flavor base liquor were distinguished rapidly.In this research,the discriminant models were established and compared by four kinds of spectral information of the base liquor combined with principal component analysis(PCA),liner discriminant analysis(LDA)and back propagation artifical neural network(BPANN).The results indicated that the four spectral could distinguish the different grades of base liquor,and the model performed better.The FTIR combined with LDA and BPANN could distinguish the base liquor of five different grades,and the accuracy of the training set and test set were 100%.The results of other three models had slightly worse effects,but could basically achieve the distinction of different base liquor.Thirdly,in the paper,the content of ethyl hexanoate,ethyl lactate,ethyl butyrate and ethyl acetate were detected quantitatively by Gas Chromatography.Each kind of esters content was measured as references to establish synergy interval partial least squares(SiPLS)models combined with spectroscopy.It is reported that the esters content of base liquor can be rapidly determined by the four kinds of spectroscopy.The FTIR combined with SiPLS models for predicting the content of the ethyl hexanoate,ethyl lactate and ethyl acetate were realized with the best performance.The correlation coefficient of the training set of the ethyl hexanoate,ethyl lactate and ethyl acetate in the SiPLS models were 0.9866,0.9942,0.9908 respectively.And the correlation coefficient of the test set were 0.9847,0.9937,0.9852 respectively.The NIR combined with SiPLS model for predicting the content of ethyl hexanoate was realized with the best performance.The correlation coefficient of training set and test set were 0.9262,0.9707 respectively.All the models could meet the requirements of the content of esters in liquor production.This research accomplished that spectroscopy can be effectively achieved accurate determination for the strong-flavor base liquor of different grades combined with chemometrics.At the same time,the quantitative model for rapidly detecting esters has reached desired results,which can meet testing requirements during liquor production as well as provides a fast and accurate method for the determination of the grade of strong-flavor liquor base,and effectively enhance the level of intelligent production of liquor. |
PDF Full Text Request