摘要
采用"基于角度度量的多变量回归方法"对维纶和腈纶混纺纤维各组分含量进行检测,并与直接用偏最小二乘法(PLS)对混和纤维的预测结果作对比。实验结果显示,PLS法对维纶和腈纶预测值与实际值的线性相关系数r均为0.945 7,标准偏差为6.090 6,均方根误差为6.994 8。角度度量法对维纶预测值与实际值的线性相关系数r为0.999 0,标准偏差为0.892 9,均方根误差为2.189 6;对腈纶预测值与实际值的线性相关系数r为0.992 8,标准偏差为2.189 6,均方根误差为3.949 3。实验证明,角度度量法比PLS法更能准确表达定量关系,角度度量法可以显著降低分析操作对环境的要求,满足了近红外光谱在混纺纤维定量分析上的要求。
＂Multivariate regression method based on angle measurement＂ was used to detect vinylon and acrylic blended fiber content of the components, and contrasted it with the prediction which used the method of partial least squares regression（PLS）, Experimental results showed that through the method of PLS, the linear relationship r between vinylon, acrylic fibers prediction values and actual values were both 0.945 7, the standard deviation was 6.090 6, root mean square error was 6.994 8. Meanwhile, by using the method of angle measurement, the linear correlation coefficient r between prediction values and actual values were 0.999 0, 0.992 8, the standard deviations were 0.892 9, 2.189 6, root mean square errors were 2.189 6, 3.949 3 for vinylon and acrylic fibers, respectively. The experiment proved that angle measurement can exactly express quantitative relation than PLS method, and can significantly reduce requirements of environmental analysis, and meet requirements of the near-infrared spectroscopy in the blended fibers quantitative analysis.
出处
《化学分析计量》
CAS
2013年第6期46-48,共3页
Chemical Analysis And Meterage
关键词
化学计量学
近红外光谱
纤维含量
角度度量
chemometrics
near infrared spectroscopy
fiber content
angle measurement
