Machine learning to enhance sustainable plastics: a review

Plastic pollution requires advances in the production, use, and recovery of plastics to minimize environmental and human-health impacts. Machine Learning (ML) has been applied to accelerate the replacement of existing plastics with sustainable plastics. However, a comprehensive overview of how ML has been applied to promote sustainable plastics from a life cycle perspective is lacking. This article reviews the current literature on ML and its applications in sustainable polymers, representing a significant departure from previous knowledge. A comprehensive and systematic understanding of ML applications in the sustainable-plastic life cycle is provided by analyzing 47 articles on the subject published between 2019 and 2024. This review aims to increase knowledge of ML methods that are used to enhance sustainable plastics and to highlight the various challenges and opportunities. The findings revealed that ML has been applied at every stage of the polymer life cycle, with a higher incidence in the end-of-life and product-manufacturing stages. The application of ML was lowest in the assessment of the environmental impact of plastics. Neural networks and random forests are the most widely used algorithms because of their ability to deal with complex data patterns. Challenges must be addressed to increase the use of ML, namely the polymer complexity and interdependency of the polymer life cycle, the scarcity and low quality of data, and the validation of results by plastics value-chain specialists to increase trustability.