The Application Of Near-infrared Technology In The Fiber Identification And Composition Prediction Of Polyester-Wool Textile Fabrics

To improve the sorting speed of waste textile fibers,to promote the recycling of waste polyester/wool blend fabrics,410 polyester/wool blend fabric were studied by a portable near infrared spectroscopy,using partical least squares(PLS)as a correction method.Combined with qualitative identification coefficient set by the content of polyester and wool,a semi-quantitative and qualitative analytical calibration models of polyester/wool blend fabrics was established.In the modeling process,the model was optimized by the selection of the best modeling band,different pretreatment methods and the main factor of the modeling.Based on the main absorption peak of polyester and wool,1300nm~2100nm(correlation coefficient was0.6)was selected as the best waveband by bidirectional expansion.Pretreatment methods were the combination of MSC,DT,S-G smooth,S-G first derivative and mean centering,RC and RP were 0.991 and 0.983 respectively,SEC and SEP were 3.107 and 3.122,and the external validation accuracy reached to 97.01%,indicating that the qualitative analysis models can predict the polyester/wool blend fabrics well.The forecast accuracy of the model with all samples was low,whose contents range from0% to 100%.The sample set was divided into four small sample set by the wool content of the sample,which were 0% to 30%,30% to 50%,50% to 70% and 70% to 100% respectively.After four-part segment establish quantitative analysis model,the model were significantly optimization.Model effect of 0% to 30% and 70% to 100% were good,its prediction accuracy of validation sample were 99%,To verify the practicability of the model,186 polyester/wool samples of bounds were predicted,results showed that external verification accuracy of the segmentation model was 91.40%,compared with the all sample model increased by 17%.Analyzing the variance of the NIR method results and the national standard method results,when the confidence interval of 95%,significant level of 5%,the result of F test was 0.82,less than the threshold of 1.00,indicating that the results of the two methods did not exist significant differences.