Research On Rubber And Auxiliary Identification Method Based On Terahertz Spectral Separation And Classification Model

With the development of "green rubber" industry,it is of great practical significance to study a fast,environmentally friendly,non-destructive rubber detection method.Terahertz technology,as an emerging technology with dual characteristics of optics and electronics,has the characteristics of non-destructive,high resolution and accurate identification,which can effectively make up for the lack of traditional rubber detection methods.This paper focuses on the analysis of unknown component mixtures,vulcanizates and accelerators with similar physical properties in rubber testing,which has certainly scientific significance for improving the level of rubber testing.In this paper,nitrile rubber(NBR),neoprene(CR),butyl rubber(IIR)and styrenebutadiene rubber(SBR),vulcanization accelerators MBT,TMTM,and DTDM,and mixtures of some of them are used as research objects.The qualitative analysis of rubber and its multicomponent mixture was realized by terahertz spectroscopy technology.The following is a description of the research content from three aspects:(1)Research on spectral separation technology of multi-component mixtures with unknown component information.For two-component mixtures and three-component mixtures with unknown component information,it's first to determine the number of components in the mixture by using principal component analysis,and then combining the characteristics of smooth and continuous after terahertz pre-processing,using non-negative matrix factorization based on spectral feature information constraints.The matrix decomposition algorithm realizes the separation of the mixture spectrum.It is compared with the traditional non-negative matrix decomposition and the non-negative matrix decomposition method based on pure variables.The results demonstrate that the improved algorithm proposed manifests good results in both spectral separation time and accuracy.Moreover,the correlation coefficients are all above 91%,and the spectral angle is less than0.5.(2)Classification of vulcanized rubber and its accelerator.For vulcanized rubbers with similar physical properties or characteristic absorption peaks,a random classification strategy and asynchronous learning factors are introduced into the particle swarm algorithm at the same time,and a spectral classification model based on improving particle swarm optimization support vector machines is established,which is compared with the support vector machine algorithm based on traditional particle swarm optimization.The results demonstrate that support vector machine model based on the improved particle swarm has improved recognition time and classification accuracy.in addition,the recognition accuracy rate is more than 81%.(3)Software for terahertz spectral data analysis and processing based on MATLAB GUI.In order to improve the detection efficiency of rubber and its additives based on terahertz time-domain spectroscopy technology,an analysis software with friendly interface and parameterized input is designed.The software includes three major functional modules: data preprocessing,spectral separation and identification.It realizes the functions of data loading,preprocessing,parameter extraction,spectral separation,and classification mode construction,which improves the efficiency of rubber detection based on terahertz technology.