太赫兹光谱法检测面粉中过氧化苯甲酰含量

发布时间：2018-11-22 15:20

【摘要】：检测40个不同过氧化苯甲酰含量的面粉样品的太赫兹时域谱,获得其在0.2~2.0 THz频段内的频域谱、折射谱及吸收谱。时域谱表明,与参考信号相比,添加过氧化苯甲酰的面粉样品,其太赫兹波幅值出现了不同程度的降低,面粉样品的频域信号出现了不同程度的延迟现象。频域谱表明面粉样品中过氧化苯甲酰的含量越高,其频域谱的幅值越大,对太赫兹波的吸收强度越小,过氧化苯甲酰在一定程度上降低了面粉样品对太赫兹波的吸收。折射谱表明随着频率的增加,折射率逐渐降低,随着面粉中过氧化苯甲酰的含量增加,面粉样品的折射率随之增大。吸收谱表明面粉样品中的过氧化苯甲酰在1.6 THz处存在一个较为明显的特征吸收峰。因此,太赫兹光谱可用于面粉中过氧化苯甲酰的特征识别。利用40个面粉样品的吸收谱数据作为输入变量,分别采用偏最小二乘(partial least squares,PLS)、支持向量回归(support vector regression,SVR)及最小二乘支持向量回归(least squares support vector regression,LS-SVR)建立校正模型,其中PLS和SVR模型得到的决定系数(R2)都低于0.9,相对标准偏差(relative standard deviation,RSD)都大于10%,而LS-SVR模型的R2为0.986,RSD为4.30%,结果表明太赫兹光谱分析技术可以用于检测面粉中过氧化苯甲酰含量。
[Abstract]:The terahertz time-domain spectra of 40 flour samples with different benzoyl peroxide contents were measured, and their frequency domain spectra, refraction spectra and absorption spectra were obtained in the 0.2 ~ (2) THz band. The time domain spectra show that the terahertz amplitude of flour samples added benzoyl peroxide decreases in varying degrees compared with the reference signals, and the frequency domain signals of flour samples show different degrees of delay. The frequency domain spectrum showed that the higher the content of benzoyl peroxide in flour sample, the greater the amplitude of frequency domain spectrum and the smaller the absorption intensity of terahertz wave. To some extent, benzoyl peroxide reduced the absorption of terahertz wave in flour sample. The refraction spectra show that the refractive index decreases with the increase of frequency and increases with the increase of benzoyl peroxide content in flour. The absorption spectra showed that there was a characteristic absorption peak of benzoyl peroxide at 1. 6 THz. Therefore, terahertz spectrum can be used to identify the characteristics of benzoyl peroxide in flour. Using absorption spectrum data of 40 flour samples as input variables, partial least squares (partial least squares,PLS), support vector regression (support vector regression,SVR) and least squares support vector regression (least squares support vector regression,) were used respectively. LS-SVR). The coefficient of determination (R2) obtained by PLS and SVR models was lower than 0.9, and the relative standard deviation (relative standard deviation,RSD) was higher than 10, while that of LS-SVR model was 0.986nb (R ~ 2 = 4.30). The results showed that terahertz spectroscopy could be used for the determination of benzoyl peroxide in flour.
【作者单位】：东北农业大学食品学院;黑龙江北大荒丰威食品有限公司;哈尔滨商业大学计算机与信息工程学院;
【基金】：“十三五”国家重点研发计划重点专项(2016YD0401402) 国家自然科学基金面上项目(31571880)
【分类号】：O657.3;TS211.7

【相似文献】