Comparison of NIR and Raman spectrometries as quantitative methods to monitor polyethylene content in recycled polypropylene

Abstract

Recycled polypropylene (rPP) often contains a small amount of polyethylene (PE), which frequently compromises the performance of the material and needs to be monitored. In the current work, near-infrared (NIR) and Raman spectrometries were investigated to establish convenient and accurate methods to analyze PE content in rPP. Various spectrum pretreatment methods, including multivariate scattering correction (MSC), standard normal variate transformation (SNV), baseline correction, normalization, smoothing, and first derivative, were tested to reduce noise and improve spectrum quality for partial least squares (PLS) regression. Forward and backward interval partial least squares (FiPLS/BiPLS) methods were employed to optimize spectral range selection. The best NIR model had an R2 of 0.9971 and a root-mean-square error of prediction (RMSEP) of 0.2761 PE wt% in independent validation, and the best Raman model achieved an R2 of 0.9945 and an RMSEP of 0.3818 wt%. The models were further validated with rPP samples. The best Raman model obtained an R2 of 0.9954 and an RMSEP of 0.2404 wt% on non-colored rPP, but it failed to be applied to colored rPP. The best NIR model had a slightly weaker performance on non-colored rPP (R2: 0.9246; RMSEP: 0.9710 wt%), but it was successfully applied to grey commercial rPP samples. This work would help quality assurance of recycled PP materials.

Keywords

• Recycled polypropylene (rPP)
• Polyethylene (PE)
• Quantitative analysis
• Interval partial least squares regression (iPLS)
• Near-infrared (NIR)
• Raman