Use of Artificial Neural Networks for Recycled Pellets Identification: Polypropylene-Based Composites.

Area of Science:

• Materials Science
• Analytical Chemistry
• Polymer Science

Background:

• Polymer recycling faces challenges due to difficulties in classifying composite materials.
• Accurate identification of mineral filler content is crucial for effective recycling and quality control.

Purpose of the Study:

• To develop a quantitative method for predicting mineral filler content in polypropylene composites.
• To utilize Fourier-transform infrared spectroscopy (ATR-FTIR) and artificial neural networks (ANNs) for this purpose.

Main Methods:

• ATR-FTIR spectroscopy was employed to collect spectral data from polypropylene composites with talc, calcium carbonate, and glass fiber fillers.
• ANN models were trained using spectral features (600-1700 cm^-1) to correlate with filler concentrations (wt.%).
• The developed models were validated using X-ray fluorescence (XRF) and energy-dispersive X-ray spectroscopy (EDX).

Main Results:

• ANN models achieved high accuracy with prediction errors below 7.5% and R² values above 0.98 for all tested mineral fillers.
• The method successfully quantified filler content in a commercial recycled polypropylene pellet.
• Validation confirmed the reliability of the ATR-FTIR and ANN approach.

Conclusions:

• The combined ATR-FTIR and ANN approach offers a simple, rapid, and non-destructive tool for identifying mineral filler type and content in recycled polymers.
• This method can significantly reduce misclassification issues in polymer commercialization and enhance industrial quality control.
• The technique is suitable for non-expert users, promoting broader application in the recycling industry.