Year prediction of a mild aroma Chinese liquors based on fluorescence spectra and simulated annealing algorithm

Abstract

Abstract Based on the three-dimensional fluorescence spectra of original degree Chinese liquors with different years, which belongs to a well-known series in China, this article establishes the year forecast model of Chinese liquors. Firstly, the analysis of correlation between three-dimensional fluorescence spectra and liquor’s year shows that the year information is mainly distributed in the area of excitation wavelength of 200–230 nm and 250–320 nm and the emission wavelength of 400–500 nm. Secondly, by analyzing the multiple correlations of two-dimensional fluorescence spectra with the excitation wavelength of 220 nm, it is found that the multiple correlation of original spectra is stronger than that under wavelet decomposition. The detail signals have weak multiple correlation, while the approximate signal has strong multiple correlation. Finally, the spectral forecast model related to the year and the statistical distribution of year information are studied based on the nonlinear partial least square regression method. The results show that a year forecast model with high accuracy can be obtained by the simulated annealing algorithm, up to 0.96 year. The original fluorescence spectra contain abundant year information. The year information of wavelet decomposition spectra has layered distribution, and the year information is mainly concentrated in the third and the fourth layer. Furthermore, by the comparison of fluorescence spectral models in above three typical excitation wavelength interval, the year forecast model with high precision can be established in the excitation wavelength of 200–230 nm. The partial least square regression algorithm can effectively eliminates the multiple correlations among the fluorescence spectra. The research results provide not only a convenient optical method for the year calibration of the liquor, but also an important reference for the fluorescence spectra research of multi-component gradient system.