Inspection Chinese Rice Wine By Electronic Tongue

As a kind of traditional and unique Chinese wine, Chinese rice wine is a treasure in China eventhe world. Nowadays, the wines which are healthy, nutritious, safe and sanitarian are more and morepopular, as the economy of our country and the daily life of our people being improved day by day.With a moderate alcoholic content, Chinese rice wine is mellow and nutritious, which is tally for thepeople's taste, and its yield and consumption are increasing year by year. But compared with otherkinds of wines, the actuality of Chinese rice wine is not optimal and it developed very slowly. In orderto ameliorate the deficiency of the rice wine, there is a great need to establish a detection methodwhich is simpler, faster, more scientific and objective. Electronic tongue is developing very fast as anew non-destructive technique of detection these days. Electronic tongue, as a system mimicsorganization of human taste sense, is an analytical method comprising the analysis of liquid systemsusing an array of sensors and complex mathematical signal treatment, which allows the classificationof liquid samples and quantification of present chemical species.This study is dedicated to employing the electronic tongue in detection of Chinese rice wine withdifferent brand and the ones with different marked ages(the same brand), and the data obtained byelectronic tongue are analyzed by different pattern recognition technique, such as PCA, DFA, PLS,BPNN, improved BPNN and GABP. The results of the different methods above are discussed to findout the most appropriate method; some chemistry indexes of Chinese rice wine such as alcoholiccontent, amino acid, volatilizezing ester, total sugar, total acid, solid contents and pH, are determinedto establish the relativity with the results of the electronic tongue respectively. Based on therelativities, the possibility of using electronic tongue for monitoring these values in Chinese rice winefermentation was checked.First, different brands Chinese rice wine in different marked age were detected by electronic tongue,when the sensors were conditioned and washed by solvent with different alcoholic content. Theconclusion was that when the solvent with 10%(V/V) alcoholic content was used, the data of everysample were focused, and the reproducibility of the experiment was higher.Nverhong and Guyuelongshan rice wine with three different marked age (1,3and5years) andKuijishan with the marked age of 1 year were detected by electronic tongue. The data obtained byelectronic tongue were analyzed by PCA, DFA, BPNN, improved BPNN, GABP and the results werecompared. The conclusion was that: they can be classed obviously by electronic tongue; when thedistances of different species and the standard error of every species were both considered, the resultof DFA is much better than PCA; based on the DFA prediction model, the marked age of Chinese ricewine can be predicted, and the correct prediction percentage is more than 90%, which means that themarked age can be predicted by electronic tongue signal. As to the correct prediction percentage,GABPNN is 98.2%, which is higher than BPNN 89.3% and improved BPNN 94.6%; the epochs of the program to converged to the same precision were GABPNN 403, BPNN 1502 and improvedBPNN 1008, and the time cost GABPNN 29.871s, BPNN 102.874s and improved BPNN 87.644s,and it can be concluded that: compared with the standard back propagation algorithm and itsimproved method, the result shows the GA-BP algorithm has good prediction precision, highconvergent speed and less running time, and it is a fast and credible method.Based on PLS, BP, GABP, the correlation between the electronic tongue signal and the chemistryindexes such as alcoholic content, amino acid, volatilizezing ester, total sugar, total acid, solidcontents, CaO and pH, were studied. It shows: as to alcoholic content, the results of the imitation ofstudy samples and the prediction of predicted samples are very satisfying, and the standard error ofimitation is PLS 0.301919, BP 0.119798, GABP 0.194097 respectively and the standard error ofprediction is PLS 0.367009, BP 0.699641, GABP 0.381734; as to amino acid, the standard error ofimitation is PLS 0.031543, BP 0.016466, GABP 0.01290947 and the standard error of prediction isPLS 0.048712, BP 0.018372, GABP 0.015347, which indicated that the result of GABP is the best,and PLS is the worst; as to total sugar and total acid, the results of imitation are satisfying, but theresults of prediction are not good; as to solid contents, CaO and pH, the results of imitation andprediction are unacceptable.