The Application Of Near-Infrared Technology In The Fiber Identification And Composition Prediction Of Polyester-cotton Textile Fabrics

A series of 354 polyester-cotton blended fabrics were studied through the near-infrared spectroscopy(NIRS)technology,a NIR qualitative analysis model was established by partial least squares(PLS)method combined with qualitative identification coefficient.There are two categories in the NIR spectrum of dying polyester-cotton blended fabrics:normal spectrum and slash spectrum.The slash spectrum lost its spectral characteristics,which effected by the samples' dyes,pigments and matting agents chemical additives.It was in low recognition rate when the model was established by the total samples,so the samples needed to divide into two sets:normal spectral sample set and slash spectral sample set,and two NIR qualitative analysis model was established respectively.The results showed that the model recognition rate were improved greatly when the samples were modeled respectively,the recognition rate were up to 99%when the two models verified by the internal validation.150 bounds samples were used for to verify the model,tested the samples by the normal spectrum of sample model at first,then use the slash spectrum model to test those couldn't be recognized by the normal spectrum sample model.At last the recognition rate was up to 95.3%.It showed that the model can be identified in the recovery site for the polyester-cotton blended fabrics.Due to the impact of the samples' stitch structure,color and other factors,the polyester content distribution of 255 cotton-polyester blend fabric samples were from 0 to 100%.A NIR quantitative analysis model was established for a series of 255 samples,and the recognition rate was not very high.According to the polyester content,the sample sets were divided into 0-30%,30%?60%,60%?80%and 80%?100%four groups,and the NIR quantitative analysis models were established for each sample set at the same time.The quantitative calibration mode was verified by the validation samples.When the absolute error of predicted results and the standard method results in(-3%,3%)that sample predictive results were correct,whereas the prediction results were not correct.The model prediction accuracy rate of 0?30%,30%?60%and 60%?80%sample set were reached 100%,and the prediction accuracy rate of 80%-100%sample set was up to 97.3%.The polyester content of cotton-polyester blend fabric can be predicted well by the NIR quantitative analysis model.422 cotton-polyester samples were predicted to verify the practicality of the model,the results showed that the external validation accuracy rate of NIR method was 93.4%.Do the analysis variance of Near-infrared method results and national standard method results,when the confidence interval was 95%and the significant level was 5%,the F test ratio was 1.17,less than 1.35(critical value),indicating that there was no significant difference between two method results.The average absolute error of prediction was 1.78%,which was less than 3%.It indicated that the accuracy of the model was well.