MACHINE LEARNING–BASED OPTIMIZATION OF NANOCATALYST PROPERTIES FOR ENHANCED BIODIESEL PRODUCTION

Abstract

Biodiesel production through transesterification and esterification reactions is strongly influenced by catalyst chemistry, surface structure, and stability, particularly when low-cost feedstocks such as waste cooking oil and non-edible oils are employed. Nanostructured heterogeneous catalysts have gained significant attention due to their high surface area, tunable acid–base functionality, and potential for catalyst recovery and reuse. However, optimizing nanocatalyst properties remains challenging because catalyst composition, morphology, surface chemistry, and synthesis conditions interact nonlinearly with process variables and feedstock characteristics. In recent years, machine learning (ML) has emerged as a powerful tool for modeling complex reaction systems and optimizing biodiesel production parameters. While many ML studies focus on predicting biodiesel yield using operating conditions alone, fewer efforts explicitly integrate nanocatalyst physicochemical properties into data-driven optimization frameworks. This review critically examines recent advances in ML-assisted biodiesel production with an emphasis on the optimization of nanocatalyst properties. The discussion covers major nanocatalyst classes, catalyst descriptors relevant to ML modeling, commonly applied ML algorithms, validation strategies, and optimization techniques including metaheuristics and Bayesian approaches. Key challenges such as data scarcity, catalyst deactivation, reproducibility, and model generalization are highlighted. Finally, future research directions are proposed toward catalyst property–centric modeling, active learning, and inverse catalyst design for robust and sustainable biodiesel production.