Study Of Principal Qualities Determination And Calibration Transfer Of Chinese Rice Wine By Near Infrared Spectroscopy

Near-infrared spectroscopy with its rapid, non-destructive characteristics,has been used to the detection of wine, beer and liquor. Rice wine is a Chinesetraditional wine. However, its conventional detection method is time-consuming andlabor–intensive. It will be of great importance for rice wine quality rapid detectionof NIR study to reduce testing costs and improve the wine quality. Calibrationtransfer technology is a key to solve the model sharing issues between NIRspectrometer. So studying the transfer of rice wine quality NIR model betweendifferent instruments has better theoretical significance and application value.This article applied near-infrared spectroscopy combined with chemometricmethods to research on modeling and calibration transfer of rice wine quality basedon NIR. The following is the main content and conclusions of this article:1）Nexus Fourier Spectrometer (FT-N) was applied to analyze the impact ofdifferent wavelength ranges and preprocessing to the PLS model results of rice winequality. The results showed that alcohol model was built better by using the spectralrange of1300-1800nm and2150-2350nm, with the preprocessing method of the firstderivative and mean center; Acidity model was built better by using the full-bandspectral, with mean centering pretreatment; Sugar content model was built better byusing the spectral range of1300-1800nm without any pretreatment.2）The S/B, DS and PDS algorithm were applied to study the calibration transferof rice wine quality based on NIR between the FT-N (main instrument) and MPAFourier Spectrometer (slave instrument). The results indicated that S/B method onlycan play a role in the transfer of alcohol model. However, the results transferred byDS and PDS algorithm were greatly improved, close to the prediction result of themaster instrument. When transferred, the determination coefficient R2and RMSEP ofthe alcohol, total acid and sugar content model were0.972and0.300,0.935and0.257,0.985and0.218, respectively.3）Calibration and performance testing of rice wine filter NIR spectrometermade by our own was completed. The results showed that the test on the same sample, prototype’s absorbance repeatability deviation was less than0.5%; Soprototype can be used for the rapid detection of actual rice wine quality. Calibrationof alcohol was ideal, the correlation coefficient R and RMSEP was0.993and0.269,Calibration of acidity and total sugar, correlation coefficient R and RMSEP was0.959and0.238,0.937and2.279, respectively. The accuracy of the total sugar modelis not ideal.4）The S/B and DS algorithm were used to explore the transfer of the alcoholNIR model from the standard instrument (FT-N) to the non-standard equipment(developed prototype). The results show that the calibration transfer betweenstandard instrument and non-standard instrument is feasible; When transferred by theDS algorithm, the correlation coefficient R was increased from0.666to0.874,RMSEP reduced to0.898. This result is satisfactory.5）Finally, the impact of temperature changes on the transfer results for thefructose content NIR model was explored. It Preliminary concluded that temperaturechanges can affect the transfer result greatly. The transfer result of the RMSEPbetween the spectra having same temperature was less than between the spectrahaving different temperature, and the results was better at lower temperature. Thus, itis need to consider the influence of temperature change during calibration transfer.The results provide a basis for the next step applying the prototype to actual ricewine quality testing, and using the calibration transfer algorithm to solve the modelsharing issues.