Convolutional neural networks enhance pyrolysis gas chromatography mass spectrometry identification of coated papers, TAPPI Journal August 2024

Application: The results of this study provide valuable insights into the identification of coated papers using Py-GCMS combined with machine learning techniques. This approach improves the accuracy and efficiency of determining coating compositions, which can support quality control processes within the paper industry. The findings are relevant for manufacturers focusing on product quality and sustainability, particularly in the context of recycling and reducing environmental impact.

Author: Jingfan Zhang, Taojing Yuan, Jianyu Wen, and Qingwen Zhang

ABSTRACT: In the evolving paper industry, accurate identification of coated paper components is essential for sustainability and recycling efforts. This study employed pyrolysis-gas chromatography mass spectrometry (Py-GCMS) to examine six types of coated paper. A key finding was the minimal interference of the paper substrate with the pyrolysis products of the coatings, ensuring reliable analysis. A one-dimensional convolutional neural network (1D-CNN) was employed to process the extracted ion chromatograms directly, simplifying the workflow and achieving a predictive accuracy of 95.2% in identifying different coating compositions. Additionally, the study high-lighted the importance of selecting an optimal pyrolysis temperature for effective feature extraction in machine learning models. Specific markers for coated papers, including polyethylene (PE), polypropylene (PP), polyethylene terephthalate (PET), polybutylene succinate (PBS), polylactic acid (PLA), and waterborne polyacrylates (WP), were identified. This research demonstrates a novel approach to coated paper identification by combining Py-GCMS with machine learning, offering a foundation for further studies in product quality and environmental impact.