摘要
以纯纺织物(聚酯、羊毛、棉、锦纶)和混纺织物(聚酯/棉、聚酯/羊毛、聚酯/锦纶)共计7类1394个样本为研究对象,利用近红外光谱仪采集样本的原始NIRS谱图,并用近红外分析测量软件RIMP结合偏最小二乘法(PLS)在1 000~1 800 nm、1000-2 200 nm和1 000-2 500 nm(全波段)3个波段范围内,分别对聚酯/棉、聚酯/羊毛和聚酯/锦纶织物建立定量分析模型。当波段选择1 000~2 200 nm、预处理方法选择S-G平滑、S-G导数、多元散射校正(MSC)和均值中心化时,聚酯/棉和聚酯/羊毛混纺织物的定量分析模型预测效果最佳;而选择S-G平滑、S-G导数、正交信号校正(OSC)和均值中心化时,聚酯/锦纶模型的预测效果最佳。利用共计638个未参与建模的验证样本对三类模型进行外部检验,平均绝对误差均小于3%,预测准确率均达95%以上。因此,所建模型可以较好地对常见纤维制品进行含量预测。
A total 7 categories of 1394 samples of pure(polyester,wool,cotton and nylon)and blended textiles(polyester/cotton,polyester/wool and polyester/nylon)were used to establish models for quantitative prediction of textile content in terms of near-infrared spectra(NIRS).The NIRS were collected in the range of 1 000-1 800 nm,1000-2200 nm and 1000-2 500 nm(full band)respectively and the data were analyzed using software RIMP combined with partial least squares(PLS)method.In the band of 1000-2 200 nm the best prediction results were obtained for polyester/cotton and polyester/wool blend fabrics with the pretreatment method of S-G smoothing,S-G derivative,multiple scattering correction(MSC)and mean centralization.For polyester/nylon,the best prediction results were obtained with the pretreatment methods of S-G smoothing,S-G derivative,orthogonal signal correction(OSC)and mean centralization.Finally,a total of 638 samples which were not involved for establishing the models were used to test the validation of the models externally.The average absolute error was less than 3%,and the prediction accuracy was more than 95%.Therefore,the developed models are able to predict the content of common fabrics with excellent accuracy.
作者
魏子涵
李文霞
王华平
张朔
WEI Zi-han;LI Wen-xia;WANG Hua-ping;ZHANG Shuo(School of Materials Science and Engineering,Beijing Institute of Fashion Technology,Beijing 100029,China;School of Materials Science and Engineering,Donghua University,Shanghai 201620,China;Zhejiang Luyu Environmental Protection Ltd.Corporation,Shaoxing 312000,China)
出处
《北京服装学院学报（自然科学版）》
CAS
北大核心
2019年第2期30-37,55,共9页
Journal of Beijing Institute of Fashion Technology：Natural Science Edition
基金
国家重点研发计划项目(2016YFB0302900)
绿宇基金项目(HXKY05160384)
关键词
近红外
常见纤维织物
预处理方法
定量分析模型
analysis models
near-infrared spectra
common fabrics
content
quantitative
prediction
