Machine Learning Guides Biomass-Plastic Waste to Aromatics over a Hierarchical Biomass-Derived Zeolite

Abstract

Abstract Biomass-plastic catalytic co-pyrolysis offers a promising route to aromatics, yet high-value monocyclic aromatic (MAH) selectivity remains challenged by the complex interdependencies of reaction parameters. Herein, we present a machine learning-guided strategy to maximize MAH production from the co-pyrolysis of rice straw and polypropylene. Central to this strategy is a hierarchical HZSM-5 zeolite synthesized using rice husk ash as the sole silica source. A hybrid methodology combining response surface methodology and machine learning models was employed to optimize four critical variables: temperature, feedstock mass ratio, catalyst Si/Al ratio, and catalyst loading. Under the optimal conditions, an exceptional MAH content of 71.7% was achieved. The biomass-derived catalyst exhibited superior activity and excellent stability. A cradle-to-gate assessment further revealed category-specific environmental differences between the rice-husk-derived and commercial zeolite production routes under the adopted inventory assumptions. This work establishes a data-driven platform for the efficient upcycling of waste into chemical feedstocks.