Smart recycling of LDPE via catalytic pyrolysis: A machine-learning-driven kinetic study

This study constitutes the second phase of a broader research effort aimed at converting waste plastics into valuable hydrocarbons. Here, low-density polyethylene (LDPE) pyrolysis is examined under thermal conditions and in the presence of two catalysts, ZSM-5 and Cr/ZCM-5, with particular emphasis on kinetic behavior and catalyst-driven mechanistic effects. Model-free Friedman, KAS, and OFW analyses revealed conversion-dependent activation energies characteristic of multi-step degradation. In case of ZSM-5 catalyst, the activation energy decreased from ∼120 to 90 kJ mol⁻¹ across α = 0.2–0.8, indicating dominant chain scission behavior. In contrast, Cr/ZSM-5 exhibited a pronounced mid-conversion maximum (∼160–167 kJ mol⁻¹ at α ≈ 0.5) in the Friedman profile, reflecting a mechanistic transition toward energy-intensive secondary reactions such as dehydrogenation, cyclization, and aromatization. Semi-batch pyrolysis at 400 °C showed complete LDPE conversion, with liquid yields of 65% (thermal), 60% (ZSM-5), and 58% (Cr/ZSM-5). The Cr/ZSM-5 catalyst produced the most aromatic-rich oil (39.2%) and the highest Motor Octane Number (MON = 85.8), compared to 12.0% aromatics and MON 69.9 under non-catalytic conditions. All liquid products exhibited high heating values in the range of 41.8–42.6 MJ kg⁻¹. Among the ML models evaluated, the ANN-MLP demonstrated superior predictive performance (R² = 0.9993, RMSE = 0.8129), outperforming CRT, MARS, and BRT approaches. The results demonstrate that chromium incorporation induces bifunctional Lewis-Brønsted acidity, alters kinetic pathways, and enhances aromatization, while ANN-based modeling effectively captures the complex conversion-dependent behavior. This integrated kinetic-ML framework provides mechanistic insight and supports data-driven catalyst design for advanced plastic-to-fuel technologies.