Study Of Terahertz Spectroscopy Detection Methods For Textiles And The Additives

The textile industry is one of the mainstay industries of our national economy and is of great significance to the economic development of our country.With the rapid development of the textile industry,the variety of textiles is becoming more and more abundant and the demand for testing is increasing.Existing testing methods for textiles and additives are complex,time-consuming and inefficient,and there is a need for a simple,fast and accurate testing method to make up for the shortcomings of existing testing methods.Terahertz time-domain spectroscopy is widely used in the detection of substances because of its advantages such as rapidity,safety and accuracy.The main research is as follows:(1)A support vector machine optimised classification model based on a data fusion sparrow search algorithm was developed for six textile materials.The absorption spectra of six textiles in the range of 0.3-1.6 THz were obtained and calculated using the Terahertz Time Domain Spectroscopy(THz-TDS)system.Sample derivative spectral data were obtained by finding the first-order derivative.To address the problem of complex textile compositions and low recognition accuracy,a low-level and mid-level data fusion method were used to process the spectral data and a support vector machine(SVM)model was used as the basis for the classification model.To address the problem of slow convergence and low accuracy of conventional search algorithms,the Sparrow Search Algorithm(SSA)was used to optimize the SVM model.The results indicate that the prediction set accuracy of the SSA-SVM model based on mid-level data fusion is 96.6667%,with high classification accuracy and recognition efficiency,providing an effective method for the qualitative detection of textile fibers.(2)Cotton and polyester two-component mixture were used as the study object,the IVISSA algorithm was used for feature extraction based on SVR to address the non-linear phenomenon of the absorption spectrum of the mixture and the difficulty in balancing effective information extraction and redundant information rejection.The terahertz timedomain spectroscopy(THz-TDS)system was used to collect spectra of cotton and polyester two-component blends at six ratios(0%-100% with a gradient of 20%),and the KS algorithm was used to divide the sample set and develop quantitative analysis models for partial least squares regression(PLSR)and support vector regression(SVR).The predictive effect of SVR was found to be much higher than that of PLSR.The results show that the correlation coefficients of both the calibration and prediction sets of the IVISSASVR model are significantly improved,indicating that the IVISSA-SVR method can be used for the quantitative analysis of textile fibers.(3)Polyester and delusterant mixture were used as the study object,the quantitative analysis model of three data fusion techniques combined with PLSR and SVR was developed for the problem of inaccurate results due to low delusterant content in textiles,using absorption and derivative spectra as the base spectra.Six mixed sample spectra of different concentrations(0%-10% with a gradient of 2%)were acquired using the THzTDS system.The low-level data fusion directly is to fuse the two normalized spectral data.The mid-level data fusion is to fuse the feature variables derived from Monte Carlo uninformative variable elimination method.The high-level data fusion is to fuse the prediction results of the model by the multiple linear regression method.By comparing the PLSR and SVR models,it was found that the models can influence the predictive effect of data fusion.The results show that both data fusion models outperform absorption spectroscopy models,indicating that data fusion can improve the predictive performance of the models,which provides an ideal model for the quantitative regression analysis of textiles and additives.