OPTIRBLEND · AI-guided design and optimization of recycled polymer blends for high-performance and sustainable composites

Plastic waste is one of the most pressing global challenges, with more than 350 million metric tons generated annually and less than 10% effectively recycled. Polypropylene (PP) and Polyethylene (PE) account for nearly 50% of global plastic production and are often collected as mixed waste streams. Their immiscibility, however, leads to poor mechanical performance, preventing their use in high-value applications. Overcoming this limitation requires new methodologies to optimize manufacturing parameters and microstructures to maximize mechanical performance and reduce their intrinsic variability.This project proposes a data-driven framework for the design and optimization of recycled PP/PE blends. The methodology integrates:1. A manufacturing campaign with systematic variation of processing parameters (blend composition, compounding time, temperatures, screw speed, cooling rate, etc.), combined with morphological characterization.2. Generative machine learning to model and reproduce microstructures conditioned on manufacturing parameters.3. A multi-fidelity model based on finite element analyses to predict yield stress, ultimate strength, and toughness while quantifying epistemic and aleatoric uncertainties.4. Bayesian active learning and multi-objective optimization to identify optimal manufacturing configurations.The framework will be validated through an experimental testing campaign. Expected outcomes include a fundamental understanding of process–microstructure–property relationships in recycled PP/PE blends, efficient predictive tools, and manufacturing guidelines. The project will directly support the circular economy by enabling recycled polymers to be used in high-demand sectors (e.g., transportation, construction), strengthening Europe’s competitiveness while reducing environmental impacts.