Identification Of Bast Fiber Based On Near Infrared Spectroscopy

Bast is a kind of natural fiber with abundant sources and is one of the important textile raw materials.With the development of degumming,printing and finishing technology,bast fiber is more and more favored by garment fabric designers,and the value of products is gradually increasing.Hemp fibers,flax,ramie and jute are commonly used.Their apparent morphology and chemical properties are similar,but the value of different fibers varies greatly.At present,the identification of hemp fiber at home and abroad is classified into hemp category,after degumming treatment of hemp,flax,ramie and other hemp subdivision identification is not clear,how to accurately identify different hemp has become an important research topic.The commonly used hand visual method,staining method,microscope observation and infrared spectroscopy have some problems,such as complicated operation,limited scope of application,and low identification accuracy of these methods.In this study,the qualitative identification model of bast fiber and the quantitative identification of hemp content in hemp/flax mixed samples will be established by means of near infrared spectroscopy combined with stoichiometry.Specific research and conclusions are as follows:（1）Explore the influence of samples,instruments and other factors on near infrared spectrum,and select appropriate parameters as the unified specifications for subsequent collection of near infrared spectrum of bast fiber.The results show that:the sample size is guaranteed by using a 40-mesh sampling screen.Under the condition of constant temperature and humidity of the test environment,the fluctuation between the 30 near-infrared spectra of the same sample is small,the stability is good,and the sample production time is short.The spectral curve is smoother,has good spectral stability and is representative,and there is less spectral independent information,when the scanning interval is 12500-4000cm^-1,the scanning times are 32.（2）A partial least square identification model was established by near infrared spectroscopy of hemp,flax,ramie and jute after removing abnormal data and pretreatment.The results show that:the training set determination coefficient（Rc²）of PLS model is 0.9566,which is close to 1,and the training set root mean square error（RMSEC）is 0.2765,which is close to 0,indicating that the model has good stability.The determination coefficient of prediction set（Rp²）is 0.9478,the root means square error of prediction set（RMSEP）is 0.3033,and the relative analysis error（RPD）is3.1361,indicating that the prediction results are accurate,the fluctuation is small,and the model has high reliability.The identification accuracy of hemp and flax was above95%by PLS model,and the accuracy of ramie and jute was 100%.The qualitative identification of hemp,flax,ramie and jute was realized.（3）With hemp and flax as the research object,11 mixed samples with hemp content of 0%to 100%increased by 10%each time were prepared,and near infrared spectra were collected.PLS model and MLR model were established after band selection and pretreatment,and the models were verified.The results showed that:in the PLS model,Rc²=0.9811,RMSEC=0.0436,Rp²=0.9829,RMSEP=0.0414,RPD=5.4306.The absolute error between the average predicted content and the actual value of cannabis content was less than 3%,and the predicted residual was less than6%.In the MLR model,Rc²=0.9792,RMSEC=0.0436,Rp²=0.9845,RMSEP=0.0397,RPD=5.7017.The absolute error between the mean and actual predicted content of cannabis was within 3%,and the predicted residual was within 8%.Compared with the two modeling methods,the PLS model showed less fluctuation in the prediction results,better prediction accuracy,and more reliable model,which realized the quantitative identification of single fiber in hemp/flax mixed fiber.