Assessing Melt Flow Rate in Post-Consumer Polypropylene via Near-Infrared Hyperspectral Imaging.

Area of Science:

• Materials Science
• Polymer Recycling
• Spectroscopy

Background:

• Mechanical recycling of polypropylene (PP) is hindered by inconsistent feedstock properties, particularly melt flow rate (MFR).
• Variations in MFR across post-consumer PP packaging grades limit the quality and applicability of recycled materials.
• Accurate MFR prediction is crucial for effective sorting and upgrading of PP recyclates.

Purpose of the Study:

• To evaluate the efficacy of near-infrared hyperspectral imaging (NIR-HSI) for predicting MFR in post-consumer PP packaging.
• To compare the performance of various regression and classification models for MFR prediction using spectral data.
• To assess the impact of sample color (white vs. clear) on prediction accuracy.

Main Methods:

• Acquisition of 82 rigid PP samples from a material recovery facility with MFR ranging from 2 to 108 g 10 min^-1.
• Application of thirteen linear and non-linear regression models using median and pixel-wise spectral data.
• Evaluation of binary classification models at various MFR thresholds.
• Analysis of spectral data to identify informative regions for PP characterization.

Main Results:

• Tree-based regression models demonstrated strong performance, achieving R² = 0.85 for white PP and R² = 0.61 for clear PP.
• The combined dataset yielded R² = 0.66 for MFR prediction.
• Binary classification models achieved balanced accuracies between 0.82 and 0.92.
• Median spectral representations consistently outperformed pixel-wise aggregation.

Conclusions:

• NIR-HSI is a viable technology for supporting grade-directed sorting of post-consumer PP packaging, especially for opaque white samples.
• The method shows promise for enhancing the value and consistency of PP recyclates.
• Challenges remain due to heteroscedasticity at high MFR values and inherent data outliers.