Subspace-based Measure Of Coincidence Judgment And Perspective Of Blended Fiber Analysis

Qualitative and quantitative research of blended fiber has already been a hot topicin the textile industry. The conventional methods are microscope observation, thecombustion method, dissolution method, however, There are significant drawbacks withthis approach, such as low speed, the samples will be destroyed, and also can pollute theenvironment. Therefore, researchers have tried to find a different way to obtain the sameor even more precise results, and the spectral method can effectively overcome thedefects, thus increasing the industry's attention. At present, the spectroscopy qualitativeanalysis commonly used for textile fiber analysis depends on pattern recognition methods,The quantitative analysis of the textile fiber is required to establish a quantitative modelby the spectral intensity and fiber constituent content, but as the influence of theenvironment makes spectroscopic measurements of instability, it even appears peakoverlap and offset phenomenon, so that the analysis model of fiber spectrum is notaccurate, cannot up to the qualitative and quantitative requirements of textile fiber.There are however significant drawbacks with this approach, so I want to searchfor some special methods to get the same or more precise result. This paper adopts amethod that combines near infrared spectroscopy and chemometrics to analyze the textilefibers components, near-infrared spectrum data can be got by scanning textile fibers,qualitative analysis to blended fibers can be got by the Subspace overlap judgmentalgorithm, and quantitative analysis can be got by angle measurement algorithm, by these,we can achieve fast and accurate analysis of blended fibers. The main content of thisstudy include:?1? Introducing the Spectrum qualitative analysis method and quantitativeanalysis method. Mainly introducing the principles, methods and calculation steps ofSubspace overlap judgment algorithm and angle measurement algorithm.?2?Feasibility study of subspace overlap method and angle measurement methodfor the analysis of fiber. In space, the pure spectrum of each fiber is referred to as a vector,the ideal condition with a fiber vector angle is zero, and the different fiber vector angle isnot zero. By constructing fiber standard library, calculating the angle between the11fibers and standard A&B, the feasibility of subspace method to identify textile fibers isproved. Taking cotton/polyester for example, through calculating the angle betweencotton/polyester blended fibers and cotton&polyester NIR for semi-quantitative analysis,taking angle value as a component content metrics, the feasibility of angle measurementmethod to identify textile fibers is proved.?3? the qualitative analysis methods of blended fiber is established by thecoincidence between near infrared spectral analysis and subspace overlap judgment. near-infrared spectroscopy is used to scan pure fiber sample library and fiber blendedsamples being tested, which can build the pure fiber spectrum database of components tobe tested and the spectrum database of the sample be tested, and then utilize subspaceoverlap judgment methods to analyze fiber components of the inspected sample. theresult obtained by this methods is accurate and rapid, and will not damage the sample?4? By testing PVA/PAN's Binary blended fibers system and NIR multiplewavelength spectrum, the calibration set and prediction set are built, then using PLS andangle measurement algorithm to establish the calibration models, by these, We canpredict the purity of vinylon and acrylic in samples. Experimental results show that, byusing traditional method of PLS to model NIR, the linear relationship R between PVA'sprediction values and actual values and PAN's are both0.9457, the standard deviation?STDEV?is6.0906, root mean square error(RMSE) is6.9948. While using the methodof angle measurement, R is0.9928, the standard deviation is2.1896, RMSE is3.9493.The experiment proves that angle measurement can more exactly and rapidly expressquantitative relation than PLS method, so as to complete the determination ofeach content in the blended fiber?5?By testing Cotton/polyester/viscose's three-component mixing system andeach pure fiber's NIR multiple wavelength spectrum, the calibration set and predictionset are built, then using PLS and angle measurement algorithm to establish theircalibration models, by these, We can predict each purity of cotton/polyester/viscose insamples. It has been verified by analyzing that the relative error of PLS method is larger,even some sample's absolute error reaches more than7%, does not conform to therequirements, while angle measurement method has better prediction results, itsmaximum absolute error is less than3.3%. Using traditional PLS method to predictcotton, polyester, viscose, their correlation coefficient r are0.9253,0.9139,0.9223,while through angle measurement method which is not affected by unstable strength andother factors, their correlation coefficient r are0.9965,0.9739,0.9880, anglemeasurement method's STDEV and RMSE are significantly less than PLS'. Theexperiment proves that in the quantitation of three-component mixing system, anglemeasurement method is superior to PLS method, can complete the determination of eachcontent in the blended fiber well.