近红外技术用于白砂糖色值检测模型的建立

Near-infrared Technology for the Establishment of Detection model for Color of White Granulated Sugar

采用近红外漫反射光谱技术和改进的偏最小二乘法(MPLS),对辽宁某糖厂的白砂糖建立了色值定量分析模型。采用丹麦FOSS 2500近红外光谱分析仪对白砂糖进行扫描获得光谱数据,运用主成分分析法(PCA)、标准正常化(SNV)和去偏差技术(Detrend)对光谱进行预处理,然后用MPLS建立关于白砂糖色值的数学模型,并进行验证。模型内部交互验证偏差(SEC)为4.1838,交互验证相关系数(1-VR)为0.9831,外部验证标准偏差(SEP)为5.97,相关系数(RSQ)为0.969。结合单样本t检验,模型对同一样本的3次平行预测没有显著差异。结果表明,定标模型的预测结果符合糖厂色值检测要求。

With near-infrared diffuse reflectance spectroscopy and modified partial least squares(MPLS)method,a quantitative analysis model was established for the color of white granulated sugar from a sugar factory in Liaoning.Spectral data of the white granulated sugar were obtained by scanning with the Danish FOSS 2500 near-infrared analyzer.After preprocessed by principal component analysis(PCA),standard normalized(SNV)plus detrend method,and then MPLS was used to establish a mathematical model about the color of white granulated sugar and verify it.The standard deviation of interaction validation(SEC)was 4.1838,with correlation coefficient(1-VR)of 0.9831,and the standard deviation of external validation(SEP)was 5.97,with correlation coefficient(RSQ)of 0.969.Combined with the single sample t-test,there was no significant difference in the three parallel predictions for the same sample.The results show that the prediction results of the calibration model are consistent with requirements of color value detection in sugar refinery.