Machine Learning-Driven Kinetic Elucidation for Sustainable Solvent-Free Continuous ε-Caprolactone Production via Propionaldehyde-Mediated Nanocarbon Catalysis

The synthesis of ε-caprolactone (ε-CL) from cyclohexanone (Cy = O) based on the aldehyde/O₂ system represents a promising yet challenging approach, hindered by low byproduct value and intense exothermicity. In this study, propionaldehyde (PRA) was used as a co-oxidant, with nanocarbons as catalysts, enabling solvent-free and isothermal ε-CL synthesis in a continuous-flow reactor. By adjustment of the dispersion in the Cy = O solution and the reaction parameters, carbon nanotubes (CNTs) offered a 10.19% yield of ε-CL with an aldehyde efficiency (η) of 0.16. N-doped CNTs showed enhanced efficiency, achieving an 18.74% ε-CL yield and improved η to 0.25. Machine learning analysis revealed significant influences of catalyst type, catalyst concentration, and the ratio of aldehyde-to-ketone on reaction efficiency. A kinetic model was established by focusing on two primary reactions: Cy = O oxidation (Reaction I) and PRA auto-oxidation (Reaction II), from which the reliable kinetic parameters were obtained via genetic algorithm-based optimization. The results demonstrated that nanocarbons accelerated both reactions and increased the ratio of k_I/k_II, thereby improving the value of η. Validation of the established model by comparing the predicted data and the empirical data was confirmed as well.