Rapidly and accurately determining the resin and volatile content of CF/PPBESK thermoplastic prepreg by NIR spectroscopy

Abstract

In this study, near infrared (NIR) spectroscopy was used to predict the resin content and volatile content of CF/PPBESK prepreg. Mathematical models between resin and volatile content and spectra were established by PLS, respectively. To improve the accuracy of the models, 8 algorithms were used to preprocess the NIR spectra of the prepreg. The best preprocessing methods were SNV for resin content and SNV + 1stDER for volatile content. Combined with the CARS wavelength screening algorithm separately, the PLS mathematical models of resin and volatile content were established, respectively. The maximum absolute errors in predicting the resin content and volatile content of 14 randomly collected samples were 0.998 % and 0.203 %, respectively. The entire detection process was completed within 60 s, which was 180 times faster than the traditional analysis method. Therefore, the NIR spectroscopy can achieve accurate, rapid, nondestructive and synchronized measurement of resin content and volatile content.