Machine learning modeling for hydrolysis recycling of PET waste

Abstract

The hydrolysis of polyethylene terephthalate (PET) into terephthalic acid (TPA) can efficiently recycle waste PET, but achieving high conversion efficiency through smart reaction design remains challenging. To develop a robust and accurate machine learning (ML) model for the in-depth understanding and intelligent design of PET hydrolysis, we compiled a new dataset comprising 942 data points and comprehensive information of 44 variables involved in heating type, acid/base catalyst (ABC), organic solvent (OS), co-solvent (CS), phase transfer catalyst (PTC), and operational conditions. The developed Neural Network model demonstrated the best performance in predicting PET conversion, with a testing determination coefficient (R2) of 0.93, a root mean squared error (RMSE) of 8.45%, and a mean absolute error (MAE) of 3.49%. Model-based interpretation suggested that operational conditions contributed the most (45%) to the hydrolysis efficiency, with hydrolysis time, temperature, and ABC concentration being the top-three positive factors, followed by ABC (19%), heating type (12%), and PTC (12%). This work presents a novel ML model and provides valuable insights for researchers in the field, enhancing the understanding and design of PET hydrolysis to advance the plastic circular economy.