人工革基布纤维含量近红外定量分析模型的建立  

Establishment of near infrared quantitative analysis model of imitation leather base fabric fiber content

作　　者：程鑫桥[1] 张瑜[1] 吴利锋 CHENG Xinqiao;ZHANG Yu;WU Lifeng(Shenzhen Academy of Metrology and Quality Inspection,Shenzhen 518109,China)

机构地区：[1]深圳市计量质量检测研究院

出　　处：《中国皮革》2019年第6期46-49,52,共5页China Leather

摘　　要：研究了人工革基布纤维含量近红外光谱定量分析过程中异常值的剔除与模型建立。通过主成分分析法、马氏距离法、学生化残差与杠杆值法、残差法剔除异常值,建立了人工革基布纤维含量定量分析模型。试验表明:在剔除异常值后,建立的人工革基布纤维含量定量分析模型,实测值和预测值呈良好的线性关系,校正均方根误差(RMSEC)为1.44,相关系数(Rc^2)为0.9971,预测均方根误差(RMSEP)为1.79,相关系数(Rv^2)为0.9972,残差范围为-2.9-2.8,最优主成分数(n)为8,定量分析模型的准确度高、可靠性强,可以满足快速、准确检测的实际要求。The outliers eliminate and model establishment for near infrared spectroscopy analysis of fiber contents of imitation leather base fabric were studied. The quantitative analysis model of the fiber contents of imitation leather base fabric was founded by eliminating the outliers with principal component, mahalanobis distance, leverage and studentized residual, residua. The results demonstrate that the fitting curve of actual value and calculated value of quantitative analysis model in imitation leather base fabric have good linearity after eliminate outliers, the root mean square error of calibration ( RMSEC) is 1. 44 and linear correlation coefficient ( Rc^2) is 0. 997 1, the root mean square error of prediction( RMSEP) is 1. 79 and linear correlation coefficient( Rv^2) is 0. 997 2, the residual range is -2. 9 ~2. 8, and the optimal number of principal components( n) is 8. The accuracy and reliability of quantitative analysis model applied to blended fibers is relatively high which can meet the requirements of fast and accurate testing.

关 键 词：人工革 基布 纤维含量 近红外光谱法 异常值 定量分析 

分 类 号：TS56[轻工技术与工程—皮革化学与工程]