近红外光谱分析技术在纺织品检测中的应用

发布时间：2018-08-02 12:56

【摘要】：本文从场收集的892个纺品样品对象,包括涤麻棉丝羊毛涤棉和涤毛等,开展了短波中波红外光谱分析技术在纺品检测中的应用研究分析了同Savitzky-Golay SG滑方式对模型效果的影响,提出了wald-wolfowitz法对校集和预测集样品的光谱和理化值分进行检验,利于确保建模过程中样品的表性,建立了基于中波红外光谱的纺品分鉴别和含检测模型提出了稀疏分分析SPCA方法,可提光谱中的效信息,克服样品颜色对模型的影响,建立了基于短波红外光谱的纺品分鉴别模型要研究内容和结果1分别使用SupNIR-1520和SupNIR-1100两种专用型红外光谱仪对涤麻棉丝羊毛涤棉和涤毛等892个样品进行了光谱采集,并对涤棉和涤毛等446个样品中的分含进行了理化分析2研究了同纺品的光谱特性结果表明,同种类纺品的红外光谱具定的的差异,利用红外光谱技术对纺品进行分鉴别和含检测提供了依据3利用偏最小二乘判别PLSDA法建立了基于中波红外光谱涤麻棉丝羊毛涤棉和涤毛等样品的鉴别模型结果表明SG分段滑结合二阶导数预处理方法所建模型效果最优,所建模型中涤麻丝羊毛涤毛样品的识别率均100%,棉涤棉样品的识别率分别96.55%和96.88%4利用偏最小二乘PLS法建立了基于中波红外光谱的涤棉混纺样品中棉含涤毛混纺样品中毛含的检测模型结果表明综合使用SG分段滑阶导数SNV和DT预处理方法所建模型效果最优利用wald-wolfowitz法对RS KS和SPXY种样品集划分结果进行显著性检验,结果表明SPXY方法选的样品集更具表性通过配对T检验表明,棉毛含红外光谱检测结果传统分析方法间存在显著差异,预测集的决定系数和均方根预测标准差分别0.99 2.22%和0.98 3.47%5建立了短波红外光谱的纺品分鉴别模型利用SPCA方法效提了光谱中的用信息,克服了纺品颜色对模型的扰,提高了模型精度,将所建模型识别率从94.55%提高到100%结果表明,短波红外光谱可现纺品的分鉴别。
[Abstract]:In this paper, 892 textile samples collected from the field, including polyester, hemp, wool, polyester, cotton and polyester wool, polyester wool and polyester wool, were used to study the application of short-wave mid-wave infrared spectroscopy in the detection of textile products. The effect of the same Savitzky-Golay SG slip mode on the model effect was analyzed. In this paper, wald-wolfowitz method is proposed to test the spectral and physicochemical values of calibrated and predicted samples, which is helpful to ensure the surface of samples in the modeling process. A method of sparse fractionation analysis (SPCA) based on medium wave infrared spectroscopy is proposed for the identification and detection of textile products. The effect information in the spectrum can be extracted to overcome the influence of the sample color on the model. The research contents and results of the model based on short wave infrared spectroscopy were as follows: 1 Spectral collection of 892 samples such as polyester wool and polyester wool were carried out by SupNIR-1520 and SupNIR-1100 infrared spectrometer respectively. Physical and chemical analysis was carried out on 446 samples, such as polyester, cotton and polyester wool. 2 the spectral characteristics of the same spinning products were studied. The results showed that the infrared spectra of the same kinds of textile products had definite differences. Based on the partial least square discriminant PLSDA method, the identification model of polyester wool, polyester wool and polyester wool based on mid-wave infrared spectrum was established. The effect of the model established by the method of subsection sliding and second derivative preprocessing is optimal. In the model, the recognition rates of polyester, linen, wool and polyester wool samples were 100 and 96.55% and 96.884 respectively. The partial least square PLS method was used to establish the polyester / cotton blended samples based on mid-wave infrared spectroscopy. The results of the middle gross content detection model show that the model established by using the SG piecewise sliding derivative SNV and DT pretreatment method is optimal and the wald-wolfowitz method is used to test the significance of RS KS and SPXY species sample set partition results. The results showed that the sample set selected by SPXY method was more apparent. The results of T pair test showed that there was a significant difference between the traditional analysis methods. The determination coefficient and root-mean-square prediction standard deviation of prediction set were 0.99 2.22% and 0.98 3.47%, respectively. In this paper, the separation model of short wave infrared spectrum was established and the useful information in the spectrum was effectively extracted by SPCA method, which overcame the disturbance of textile color to the model. The accuracy of the model is improved, and the recognition rate of the model is increased from 94.55% to 100%. The results show that the short-wave infrared spectrum can be used to distinguish the textile products.
【学位授予单位】：中国计量大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TS107;O657.33

【相似文献】