| Strong-flavor liquor,as one of the twelve major Chinese flavors,has the characteristics of rich cellar aroma,full taste,sweet taste and long clean tail flavor,which is favored by the majority of consumers.The quality control of strong-flavor base liquor is the key link to guarantee the final quality of liquor.It is difficult to improve the quality of base liquor by classifying the quality of base liquor and extracting each fraction accurately in the process of Baijiu picking.In this thesis,the volatile flavor components in distillates were detected by gas chromatography,and the variation rule of flavor components in distillates was summarized by heat map analysis.The base Baijiu quality grade determination model was established by combining support vector mechanism,and the quantitative detection model of 8 compounds in base Baijiu was constructed by Fourier transform mid-infrared combined with partial least square method.The main results are as follows:(1)Gas chromatography was used to analyze the volatile flavor components in distillates of fermented grains and rette-fermented grains of strong flavor liquor.After standardized treatment,the contents of different components were analyzed by clustering heat map.The variation of alcohol accuracy of secondary butanol,n-propanol,isobutanol and 2-methylbutanol in fermented grains showed a similar trend of slow decrease.Six esters,ethyl acetate,ethyl butyrate,isoamyl acetate,ethyl valerate,ethyl caproate and ethyl heptanate,showed a trend of gradual decline with the extension of fraction time.Ethyl lactate,n-hexyl alcohol,furfural,ethyl nonanoate,2,3-butanediol showed a change law of first low and then high.Six kinds of acids,caproic acid,heptanoic acid,acetic acid,butyric acid,valerate acid and isovalerate acid,showed a gradual rule of high content in two ends and low content in the middle.The thermogram analysis of retort content in retort 3 showed that 34 retort components could be divided into three parts,and there was obvious negative correlation between compound content in retort 1 and compound content in retort 2.There was obvious positive correlation between the compounds in the two parts.By summarizing the variation rule of volatile aroma substances in different distillate base Baijius,it can provide reference for base Baijiu evaluation and quality picking.(2)Gas chromatography combined with support vector machine was used to establish base Baijiu quality grade evaluation model.The proportion of test set and chromatographic data were optimized through single factor test and response surface optimization test,and then support vector machine type and kernel function were optimized in turn.The proportion of test set was 75%.The start time of analysis was 5min,the end time of analysis was 19 min,and the interval of data points was 1 min.The maximum efficiency of the quality grade model established by c multi-class kernel support vector machine was 94%,which can provide a scientific method for the quality grade classification of base Baijiu.In this study,a multi-step prediction experiment was designed to analyze the specific situation of classification of different levels of base Baijiu.The results showed that the accuracy of classification of third grade base Baijiu was 100%,that of second grade and first grade base Baijiu was 97.2%,and that of super grade and first grade base Baijiu was 95.1%.It was also found that the cumulative contribution rate of four compounds,ethyl caproate,ethanol,ethyl acetate and ethyl lactate,to the classification model was 79.1%,and they were the key flavor components affecting the judgment of quality grade,which provided an effective evaluation method for the classification of base Baijiu quality.(3)Fourier transform mid-infrared spectroscopy combined with partial least square method(PLS)was used to establish a rapid detection model of eight compounds in base Baijiu: ethanol,ethyl acetate,ethyl butyrate,ethyl caproate,ethyl lactate,acetic acid,caproic acid and butyrate.The results showed that,except butyric acid,the R2 of each component detection model was ≥0.90,and the repeatability error of the eight compound models was less than 10%,which had a good linear correlation with chemical values.Moreover,the wider the range of compound content,the more uniform the content distribution,the better the fitting degree of the model,indicating that the detection results were accurate and the system was stable.It can meet the requirements of intelligent Baijiu picking online detection.In addition,the global correction method of temperature compensation used in this study solves the problem of large deviation of detection system caused by the change of ambient temperature,and provides a feasible scheme for realizing automatic online Baijiu-picking. |
PDF Full Text Request