不同成分比例废旧涤纶/粘胶混纺织物近红外定性分析

NIR Characterization of Waste Polyester/Viscose Blended Fabrics with Different Composition Ratios

为了找到更适合废旧纺织品分拣的方法,分别用在线近红外光谱仪和手持式近红外光谱仪对467个不同成分比例的废旧涤纶/粘胶混纺织物样本进行了实验。其中184个样本作为训练集用于建模,其余283个样本用来对模型进行验证。对在线近红外光谱仪采集的光谱进行了一阶求导、平滑处理和趋势校正处理,利用主成分分析(PCA)和软独立建模聚类分析算法(SIMCA)对废旧织物中的纤维进行分类识别,最终模型的外部验证准确率为100%,未知样本的识别率为98.37%;对手持式近红外光谱仪采集的光谱利用PCA和最小二乘法(OLS)建立待测组分的分段回归模型,并导入手持式光谱仪,废旧织物的识别率为97.88%。废旧涤纶/粘胶混纺织物样本随着粘胶纤维含量的增加,近红外光谱对样本的定性准确率降低,说明粘胶纤维含量与样本的定性准确率成反比。同时涤纶匹配度的平均值与各组分含量最大误差小于4.5%,该数据可用于辅助下一阶段定量分析的测定。

To find a more suitable method for waste textiles sorting,467 polyester/viscose blend fabrics samples with different contents were tested via an online NIR spectrometer and a handheld NIR spectrometer,respectively.Among them,184 samples were used as a training set for modeling and the remaining 283 samples were used for model validation.The spectra collected by the online NIR spectrometer were subjected to firstorder derivatives with smoothing and trend correction processes;the fiber content of waste fabrics is classified and identified by principal component analysis(PCA)and soft independent modeling of class analogy(SIMCA).The final model had an external validation accuracy of 100% and an identification rate of 98.37% for unknown samples.The spectra collected by the handheld NIR spectrometer were analyzed by PCA and ordinary least squares(OLS)to build segmental regression models of the components for unmeasured fabrics.And then,importing the model into the handheld spectrometer can generate 97.88% of the recognition rate of the waste fabrics.The qualitative accuracy of NIR spectra for the waste polyester/viscose blended textiles samples decreased as the content of viscose fiber increased,indicating that the viscose fiber content was inversely related to the qualitative accuracy of the samples.The maximum error of the average value of polyester fiber matching degree and the content of each component are all below 4.5%,which provides a reference for the quantitative analysis in the next stage.