TwinLoop: Leveraging artificial intelligence and digital twins to secure recycled plastics

TwinLoop is a collaborative research and innovation project funded by the French National Research Agency (TwinLoop: Leveraging artificial intelligence and digital twins to secure recycled plasticsANR). Its ambition is to contribute to the development of robust, fast, and operational quality assurance methods for recycled plastics intended for food-contact packaging.

In a context of rapidly expanding recycling, industrial facilities must deal with highly variable material streams, both at the input and output of the process. This variability can affect the performance of processing steps and, ultimately, the suitability of the materials produced for their intended uses. TwinLoop aims to address these challenges by exploring artificial intelligence–based solutions capable of supporting the day-to-day control of recycling processes.

The project has a clear objective: to develop rapid methods for assessing sanitary quality, ideally integrable directly into recycling lines. When applied to incoming streams, these approaches would make it possible to identify, at an early stage, materials showing quality deviations incompatible with certain uses, in order to better direct their treatment and avoid downstream process disruptions. When applied to outgoing streams, they would provide a tool for continuous monitoring of proper process operation and the consistency of recycled material quality.

To achieve this objective, TwinLoop relies on the analysis of global chemical fingerprints obtained from a large number of recycled materials and evaluated against sanitary safety criteria. These fingerprints are used to train artificial intelligence models capable of recognizing different alert levels, detecting unusual signals in material streams, and identifying situations requiring particular attention. The goal is not to identify each substance individually, but to quickly detect atypical profiles that are compatible or not with given uses.

Within TwinLoop, digital twins constitute a methodological framework for integrating information from sensors, analyses, and AI models. They aim to link analytical data, process operation, and usage scenarios in order to structure decision-support tools adapted to industrial constraints.

Anchored in real-world conditions, TwinLoop is based on the study of materials representative of current practices in the sector, covering different stages of the recycling chain. This approach is essential for developing tools that are truly transferable and useful to industrial stakeholders, in support of a circular plastics economy that is both efficient and safe.

Stakeholders in the sector interested in the scientific and industrial issues addressed within TwinLoop may contact the project team to discuss the perspectives opened up by this work.

Contacts:

Sandra Domenek – sandra.domenek@agroparistech.fr

Phuong-Mai Nguyen – phuong-mai.nguyen@lne.fr

Pauline Rieu – pauline.rieu@agroparistech.fr