变量优选在纺织品棉含量近红外分析中的应用

为实现用较少的变量建立适当的模型,以准确预测未知棉涤样品的棉含量,用NIRFlex N-500近红外光谱仪采集297个棉涤样品的光谱,用蒙特卡罗无信息变量消除法(MC-UVE)对变量进行筛选,偏最小二乘法(PLS)建立棉含量的定标模型,根据各个模型所用的样品数、交叉验证均方根误差(RMSECV)、预测集均方根误差(RMSEP)和预测相关系数(r)评价定标模型的精度和稳定性。结果表明:通过上述数据预处理方法进行变量筛选后,用PLS建立的校正模型不仅使数据的运算量大幅度降低,还能很好地预测未知样品的棉含量,使得基于近红外光谱的棉涤样品中棉含量的定量分析方法进一步简化。

Application of variable preference method on near infrared analysis of cotton content in textiles

In order to set up appropriate model with less variables and accurately predict cotton content of unknown cotton polyester samples,the spectra of 297 cotton polyester samples were collected with NIRFlex N-500 near infrared spectrometer for variable screening by Monte Carlo uninformative Variables elimination method(MC-UVE) and calibration model establishment of cotton content with Partial least square method(PLS).The accuracy and stability of the calibration model were evaluated according to the number of sample used in each model,RMS error of cross validation(RMSECV),RMS error Predicted(RMSEP) and Predicted correlation coefficient(r).The results showed that,after variable screening through above data pretreatment method,the calibration model established with PLS not only greatly reduced the data computational complexity but also accurately predicted cotton content of the unknown samples,further simplifying the quantitative analysis of cotton content in cotton polyester samples based on near infrared spectra.