基于量子遗传算法和荧光光谱某清香型白酒年份预测研究

发布时间：2018-06-10 18:50

  本文选题：白酒 + 荧光光谱　； 参考：《光谱学与光谱分析》2017年05期

【摘要】：年份白酒现已成为企业开发重点,但年份标准有较大的随意性,建立年份标准已成为规范行业和市场的迫切需要。基于某品牌原浆白酒的三维荧光光谱,对白酒年份预测模型进行了研究。研究内容和创新工作如下:首先,研究了荧光光谱与白酒年份的相关性。研究发现:0.5年与其他年份白酒的三维荧光光谱之间的相关系数达0.811 4;原始光谱中年份信息主要分布在激发波长为200~230和250~320nm、发射波长为400~500nm的光谱区;导数光谱的年份信息分布区域广且离散性高。其次,研究了荧光光谱之间的相关性。研究表明:原始光谱具有严重的多重共线性,在400~600nm的区间内,相关系数接近1;求导能提高光谱分辨能力并降低多重共线性,二阶导数具有更好的抑制多重共线性的作用,相关系数大部分小于0.6。最后,基于量子遗传算法-小波神经网络研究了激发波长为300nm的白酒年份预测模型,并提出了光谱建模信息密度的概念。研究发现:原始光谱年份预测误差达5.4年,效果最差,其原因是原始光谱具有严重的多重共线性以及光谱与年份的相关性不显著;导数光谱具有更高的信息密度和更好的建模效果,二阶导数光谱预测集的相关系数达0.999 8,年份预测误差达0.79年。研究成果将为白酒年份标定提供一种便捷的光学手段,同时也为多组分渐变体系的荧光光谱研究提供重要的参考。
[Abstract]:Years of liquor has become the focus of enterprise development, but the year standard has a greater arbitrariness, the establishment of the year standard has become the urgent need to standardize the industry and market. Based on the three-dimensional fluorescence spectrum of a brand of original liquor, the prediction model of liquor year was studied. The research contents and innovations are as follows: firstly, the correlation between fluorescence spectrum and liquor year was studied. It was found that the correlation coefficient between the three dimensional fluorescence spectra of liquor in the year of 1: 0.5 and that in other years was 0.811, and the year information in the original spectrum was mainly distributed in the spectral region with excitation wavelengths of 200 ~ 230 and 250 ~ 320nm and emission wavelength of 400~500nm. The year information distribution of derivative spectrum is wide and discrete. Secondly, the correlation between fluorescence spectra was studied. The results show that the original spectrum has serious multiplex collinearity, the correlation coefficient is close to 1 in the range of 400~600nm, the derivative can improve the spectral resolution and reduce the multiplex collinearity, and the second derivative has a better effect on suppressing the multiplex collinearity. Most of the correlation coefficients are less than 0.6. Finally, based on quantum genetic algorithm (QGA)-wavelet neural network, the prediction model of liquor year with excitation wavelength of 300nm is studied, and the concept of spectral modeling information density is proposed. It is found that the prediction error of the original spectral year is 5.4 years, the effect is the worst, the reason is that the original spectrum has serious multiple collinearity and the correlation between the original spectrum and the year is not significant. The derivative spectrum has higher information density and better modeling effect. The correlation coefficient of the second derivative spectrum prediction set is 0.999, and the year prediction error is 0.79 years. The research results will provide a convenient optical means for the year calibration of liquor, and also provide an important reference for the study of fluorescence spectrum of multicomponent gradient system.
【作者单位】： 河海大学理学院;江南大学理学院;
【基金】：国家自然科学基金项目(61378037)资助
【分类号】：O657.3;TS262.3
，

本文编号：2004238