Intelligent Recognition Technology On AE Signals Of Blended Yarn In The Tensile And Fracture

Industrial textiles have a high technical content and added value,that causes higher intrinsic quality,especially the physical and mechanical performance requirements.Pure fiber of non-high-performance fiber is often difficult to meet these requirements,and a common solution is using multi-component fibers to blend.However,the actual practice found: the actual strength of blended yarn is often lower than its theoretical strength,and the strength utilization of the various components are low.Therefore,it is necessary to analyze the tensile fracture behavior of the multicomponent blended yarn to guide the production of blended yarns so as to improve the tensile strength.The traditional methods to study blending yarn's strength includes two aspects: one is to teste its basic mechanical properties such as breaking strength and elongation at break to evaluate the tensile properties of yarn,which is difficult to fully reveal the reasons which cause such low fracture strength;the other is to construct a predicting model by simplifying tructure of the yarns and fabrics.Because of too many parameters and artificial prediction or simplification have great influence on the result,so the prediction accuracy is limited.Therefore,an analysis method for the blended yarn tensile fracture behavior,which has practical values for industry,has always been looking forward to be proposed.In this paper,an AE signal acquisition platform was set up and the AE signals of four kinds of fibers(Podrun,Aramid 1313,viscose,Vinylon)and their blended yarns were collected.The wavelet denoising,HHT,PCA and(LS-SVM)were used to study the tensile fracture behavior of the blended yarn.The main research contents and conclusions are as follows:(1)An AE signal acquisition platform was set up,and a comparation between the single-fiber AE signals and bundle-fiber AE signals was done with a pre-experiment.In the existing YG065 C electronic fabric strength machine,a sensor was installed,which was also connected to the computer,to built fiber tensile fracture AE signal acquisition platform.The results showed that the signals of the bundle fiber were relatively much more complex and stronger.The signals of single fiber were much cleaner and weaker than that of the bundle fiber.(2)By processing and analysing the AE signals,the feature extraction algorithm called wavelet denoising-intercepting signal-HHT for the AE signals was made sure and the single fiber was confirmed as the experimental object.The feature extraction algorithm based on wavelet de-noising-HHT was explored and the results showed that the spectrums of single fiber were more distinct and suitable than that of bundle fiber.Compared with the Hilbert spectrum of AE signals of four kinds of fiber,it was found that the existence of end effect of EEMD had a great interference to the feature extraction.So an improvement for EEMD was done to eliminate the influence and the feature extraction algorithm(improved EEMD-suitable IMF components-Hilbert transform-seek Hilbert spectrum-seek the marginal spectrum-extract the characteristic frequencies)was finalized.(3)A series of tensile fracture AE signals of single fiber were collected and processed,and the characteristic frequencies library was preminayily built.The tensile fracture AE signals of four kinds of single fibers were collected(100 samples of each fiber were effectively collected),and the signals were processed by wavelet denisinginterception of fracture zone signal signal-HHT prceess,and the characteristic frequencies of four kinds of single fibers(each of the fiber was 100 groups)were obtained.So the characteristic frequencies library of four kinds of single fibers' tensile fracture AE signals was preminayily built.By overall observation to fracture characteristic frequencies,it was found that the characteristic frequencies of four kinds of single fibers were similar and the discrimination degree was low.(4)Based on the characteristic frequency of single fiber modified by PCA and SVM,an intelligent LS-SVM classifier was constructed,and the correct recognition rate was higher.Firstly,according to the characteristic frequencies library of tensile fracture AE signals of four kinds single fibers,the LS-SVM classifier was selected,radial basis kernel function(RBF)was selected and four sub-classifiers were constructed by two classification method;The first 50 groups of the 100 groups' characteristic frequencies were chose as train samples to train classifiers,and the last 50 groups were chose as test samples to test the performance of the classifier.It was found that the average correct recognition rate for four kinds of fibers' characteristic frequencies was just 31%.Therefore,the PCA was used to processs the characteristic frequencies further.And the LS-SVM classifier was constructed by using the characteristic frequencies after PCA processing.It was found that after the LS-SVM classifier identifying,the average correct recognition rate reached 89.5%,the correct recognition rate greatly improved.The results showed that PCA-LS-SVM could realize a reliable identification for the characteristic frequencies of tensile fracture of different fibers and the constructed LS-SVM classifier had a good performance.(5)Obtained the tensile fracture AE signals of the four components' blended yarn and processed them by the feature extraction algorithm to get the characteristic frequencies of the yarn.Using the constracted LS-SVM classifier to test the characteristic frequencies,a more clear and credible tensile fracture behavior of the blended yarn was obtained.The tensile fracture AE signals of the specific blended yarn was effectively collected(15 groups were collected).The each group tensile fracture signal of blended yarn was divided into several segments,and each segment was processed by the feature extraction algorithm to get the different segment's characteristic frequencies.And the constructed LS-SVM classifier was used to test these segments' characteristic frequencies.The recognized results showed: during the tensile fracture of the four-component blended yarn,the aramid fiber 1313 was mainly fractured,and the fracture occurred continuously during the whole process of tensile fracture;Gel's fracture occurred in the final stage of the tensile fracture process;there was almost no fracture in the process for Po Lun;And the fracture points of the Vinylon appeared rarely,only 5.4% of the 240 points,and the fracture time was distributed at the beginning,middle time and end.It could be deduced that in the tensile fracture process of the four components blended yarn,the aramid 1313 fiber had the strongest contribution for the yarn's fracture,followed by Vinylon fiber,viscose fiber had little contribution,and thePodrun fiber had no contribution.The four fibers' tensile fracture didn't synchronize with each other,which was the main cause of the the low strength of four components blended yarn.In order to verify the SVM recognition result of blended yarn,a test and analysis for the mechanical properties of the blended yarn was done.And the fracture surface of the blended yarn was observed under optical microscope.The results were in good agreement with the SVM recognition results.