Resolution-Robust Mid-IR ATR-FTIR Classification of Mixed Plastic Waste with PVC Screening and Polyethylene Subclassification
- 1Korea Institute of Energy Technology
Abstract
Reliable and chemistry-specific presorting of mixed plastic waste is essential for safe and efficient valorization. In this study, a resolution-robust mid-infrared attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) workflow coupled with simple machine-learning classifiers is presented for poly(vinyl chloride) (PVC) screening and subclassification of three polyethylene (PE) grades: low-density polyethylene (LDPE), linear low-density polyethylene (LLDPE), and high-density polyethylene (HDPE). A total of 770 ATR-FTIR spectra in the fingerprint region (2500–500 cm⁻¹) were collected from seven plastics: LDPE, LLDPE, HDPE, PP, PS, PET, and PVC, using commercial pellets together with post-consumer PET and discarded PVC gloves. Principal component analysis (PCA) was used to compress the high-dimensional, correlated spectral inputs into a low-dimensional score space, reducing multicollinearity and overfitting risk while preserving class-identifiable fingerprint information. The resulting PCA structure showed clear low-dimensional separability among the plastic classes and PE subclasses, supporting the use of simple classifiers rather than heavily parameterized black-box architectures. Support-vector machine (SVM) and artificial neural network (ANN) classifiers were trained on the PCA scores, with internal-test and strictly independent external-validation performances evaluated separately. For the baseline dataset (2 cm⁻¹), both classifiers achieved 100% external-validation accuracy. Even after degradation to effective spectral resolutions of 8 and 16 cm⁻¹, SVM and ANN maintained 99.68% internal accuracy and 100% external-validation accuracy across PVC screening, PE-grade subclassification, and full seven-class discrimination. These results demonstrate a resolution-robust, PCA-interpretable, and externally validated ATR-FTIR/ML workflow for practical plastic presorting.
Related
- From Plastic Waste to Power Source: Polyethylene-Derived Anodes for Lithium-Ion Batteries
- Delamination of Multilayer Plastic Films to Recover Polyethylene with Favorable Mechanical Properties
- Ultra-High-Performance Liquid Chromatography-High Resolution Mass Spectrometry and Gas Chromatography-Mass Spectrometry Screening of Classical Psychoactive Substances, Novel Psychoactive Substances, and Metabolites in Postmortem Urine Samples
- High resolution mass spectrometry screening method for routine surveillance of pharmacologically active substance residues in meat and milk
- Large language model-based screening of substances and their composition from safety data sheets for high-resolution chemical exposure assessment
BibTeX
@misc{jeong2026resolutionrobust,
title = {{Resolution-Robust Mid-IR ATR-FTIR Classification of Mixed Plastic Waste with PVC Screening and Polyethylene Subclassification}},
author = {Youngho Jeong},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15000518/v3},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15000518/v3},
note = {CC BY},
}
RIS
TY - UNPB
TI - Resolution-Robust Mid-IR ATR-FTIR Classification of Mixed Plastic Waste with PVC Screening and Polyethylene Subclassification
AU - Youngho Jeong
PY - 2026
DA - 2026/09/11
PB - ChemRxiv
DO - 10.26434/chemrxiv.15000518/v3
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15000518/v3
AB - Reliable and chemistry-specific presorting of mixed plastic waste is essential for safe and efficient valorization. In this study, a resolution-robust mid-infrared attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) workflow coupled with simple machine-learning classifiers is presented for poly(vinyl chloride) (PVC) screening and subclassification of three polyethylene (PE) grades: low-density polyethylene (LDPE), linear low-density polyethylene (LLDPE), and high-density polyethylene (HDPE). A total of 770 ATR-FTIR spectra in the fingerprint region (2500–500 cm⁻¹) were collected from seven plastics: LDPE, LLDPE, HDPE, PP, PS, PET, and PVC, using commercial pellets together with post-consumer PET and discarded PVC gloves. Principal component analysis (PCA) was used to compress the high-dimensional, correlated spectral inputs into a low-dimensional score space, reducing multicollinearity and overfitting risk while preserving class-identifiable fingerprint information. The resulting PCA structure showed clear low-dimensional separability among the plastic classes and PE subclasses, supporting the use of simple classifiers rather than heavily parameterized black-box architectures. Support-vector machine (SVM) and artificial neural network (ANN) classifiers were trained on the PCA scores, with internal-test and strictly independent external-validation performances evaluated separately. For the baseline dataset (2 cm⁻¹), both classifiers achieved 100% external-validation accuracy. Even after degradation to effective spectral resolutions of 8 and 16 cm⁻¹, SVM and ANN maintained 99.68% internal accuracy and 100% external-validation accuracy across PVC screening, PE-grade subclassification, and full seven-class discrimination. These results demonstrate a resolution-robust, PCA-interpretable, and externally validated ATR-FTIR/ML workflow for practical plastic presorting.
ER -