Multi-indicator Simultaneous Analysis And Brand Identification Methods Of Wine Based On Vis-NIR Spectroscopy

Wine is an alcoholic beverage with mild alcohol content,diversified taste and high popularity among consumers.Vinification of wine includes microbial fermentation and the leaching and transformation of many organic substances.High-quality wine with delicate flavor can be produced through a unique and slow process.Alcohol,total sugar,total acid and total phenol are the main quality monitoring indicators for large-scale wine production.Traditional detection methods require multiple detection methods,equipment and reagents,which are tedious and time-consuming,and cannot meet the real-time detection requirements of the vinification process.Visible-near infrared?Vis-NIR?spectroscopy is an indirect analysis technique.It can use the spectra of known samples and the physical and chemical values of the indicators to calibrate,enabling simultaneous and rapid quantitative/qualitative analysis of multiple indicators of unknown samples.This paper systematically studies the simultaneous quantitative analysis methods of quality indicators?alcohol,total sugar,total acid and total phenol?based on wine spectroscopy,and conducts research and integration of new spectral pre-treatment and wavelength model optimization methods to improve the analysis accuracy of spectroscopy.On the other hand,the identification of premium wine brand can avoid adulteration and fraud,and protect the intellectual property rights of producers and the interests of consumers.Brand identification method of wine is difficult and complex because of high similarity.This paper also conducts method research on brand identification of wine based on qualitative discriminant analysis of Vis-NIR spectroscopy.The main contents and results are as follows:1.Simultaneous quantitative analysis methods of four quality indicators of wine based on NIR spectroscopy:1)The multi-partition modeling system was established based on the partial least squares?PLS?method.And parameters were optimized based on comprehensive prediction error?SEP⁺?to avoid data overfitting.The samples that were not involved in modeling were used for independent validation to make the results objective.2)The spectral pre-processing optimization platform based on Norris derivative filter was established.And the parameters?number of smoothing points s,derivative order d and number of differential gaps g?according to the effect of the Norris-PLS model were determined.The optimal parameters?d,s,g?for alcohol,total sugar,total acid and total phenol were?2,9,3?,?1,19,5?,?1,17,11?and?1,1,1?,respectively.3)The wavelength model optimization platform based on equidistant combination PLS?EC-PLS?was established.And the wavelength step-by-step phase-out PLS?WSP-PLS?method and exhaustive method were integrated for secondary optimization to further improve the spectral prediction ability.For the above four indicators,the number of wavelengths of the optimal model of the secondary optimization was 7,10,15 and 13,respectively.In validation,the predicted root-mean-square errors?SEP?between the predicted and actual values of the four indicators were 0.41 v/v,1.48 g/L,0.68 g/L and 0.181 g/L,respectively;the predicted correlation coefficient?R_P?were 0.947,0.992,0.981 and 0.948,respectively;the ratio of sample performance-to-deviation?RPD?were 3.2,6.8,5.1 and 2.9,respectively.The results indicated the high correlation and low error between the predicted and actual values.2.Qualitative discriminant analysis of visible-near infrared?Vis-NIR?spectroscopy for brand identification of wine:1)The model evaluation system of calibration-prediction-validation for discriminant analysis of negative?Chile Aoyo wine?and positive?other brands of wine?samples was estalished.The total recognition accuracy rate(RAR_Total)was used as the optimization index to select parameters,and the balance of discrimination effect for each attribute sample was taken into account.2)Discriminant analysis models were established in the regions of visible?400-780nm?,short-NIR?780-1100nm?,long-NIR?1100-2498nm?,NIR?780-2498nm?and whole scanning?400-2498nm?with the partial least squares discriminant analysis?PLS-DA?.In validation,the visible region model achieved the best prediction results.The recognition-accuracy rates of negative,positive and total achieved 100%,95.6%and 97.5%,respectively.The research shows that:1)NIR spectroscopy combined with a new chemometric method can be used for rapid and simultaneous quantitative analysis of alcohol,total sugar,total acid and total phenol in wine.2)Vis-NIR spectroscopy combined with PLS-DA method can be used for qualitative discriminant analysis of wine brands.The proposed wavelength model optimization method is novel and can provide valuable reference for designing wine spectrometer.This technology is fast and simple,and has application potential in real-time quality monitoring of the wine vinification process.