Study On Component Content Measurement Of Polymer Blends Based On Mathematical Representation Of Raman Spectra

In recent years,in-line quantitative measurement method based on Raman spectroscopy has attracted a lot of attention in the polymer processing field due to its real-time,nondestructive,easy to install and so on.However,for polymer blends,the Raman characteristic peaks of different components may overlap with each other,and the interaction between molecules will lead to spectral signals deviating from linear superposition,which increases the difficulty of quantitative modeling analysis.Traditional soft modeling methods need a large number of samples to train the model,otherwise it is difficult to ensure the accuracy of the model.To solve the above problems,the hard modeling method based on spectral peak decomposition was used to characterize the component contents of PC/ABS blends.In this paper,a fast peak searching algorithm was proposed to solve the problem of overlapping peak searching in analytic process.The number of characteristic peaks,the number of iteration,the method of iteration and the peak-fitting function which affect the accuracy and speed of spectral resolution and studied.In the feature extraction stage,we proposed a screening method based on two-dimensional correlation spectrum to select feature peaks.Finally,based on the above research,the content quantitative models of PC/ABS blends based on spectral peak decomposition were established,and compared with the models obtained by soft modeling method.The main findings are as follows:(1)The fast peak searching algorithm can identify and separate the overlapping vibration peaks in the blends spectra quickly and accurately.The algorithm resolved the actual spectra into the superposition form of multiple independent analytical peak functions and obtained more characteristic information in spectra.(2)The number of characteristic peaks and the number of iterations had a great influence on the analytical time,and the analytical process based on the fast peak searching algorithm had a higher analytical accuracy.Voigt line described the actual polymer spectrum waveform best,and was most suitable for the polymer spectrum resolution process.(3)The peak screening method based on two-dimensional correlation spectrum was able to find out the quantitative peaks that are sensitive to component content.This method can provide high-quality training samples for the modeling.(4)The Raman spectra of PC/ABS blends in both off-line and processing processes had very high analytical accuracy.The nonlinear variation of the spectrum can be well described by optimizing the analytical peak parameters of the component.The component content quantitative model of PC/ABS blends established by this method has high prediction accuracy.Compared with the traditional soft modeling methods,the hard modeling method based on mathematical representation of Raman spectra can obtain a higher precision quantitative model under the condition of fewer training samples,so it has great potential in the field of in-line quantitative characterization of polymer processing.