Kinetic modeling and multi-objective optimization of ibuprofen synthesis using the SAO-CatBoost framework

Abstract This study proposes the SAO-CatBoost framework, utilizing the Snow Ablation Optimizer (SAO) to enhance a CatBoost-based meta-model for modeling and optimizing the ibuprofen synthesis process. Initially, a database of 39,460 input combinations is created using chemical reaction theory. The enhanced meta-model, built with CatBoost, outperforms traditional optimization algorithms (Random Search, Bayesian Optimization, and Particle Swarm Optimization) in pre- dicting reaction time (RT) and conversion rate (CR). Importance analyses through SHAP values identify critical input variables impacting RT and CR, notably the concentration of the catalyst precursor (L₂PdCl₂), hydrogen ions (H⁺ ), and water (H₂O). Partial dependence and interaction effects are explored to understand the interactions between input variables and their effects on reaction time and conversion rate. To optimize reaction efficiency and product selectivity, ini- tial concentrations and reaction rate constants are treated as decision variables, and NSGA-II is employed for multi-objective optimization, aiming to simultaneously minimize RT and maximize CR. Within this framework, uncertainty analysis of the meta-model is conducted under various levels using Monte Carlo simulation and Prediction Interval analysis, demonstrating a significant increase in prediction uncertainty with greater input perturbations. This study demonstrates the SAO-CatBoost framework’s effectiveness in optimizing ibuprofen synthesis, highlighting its reli- ability in achieving optimal reaction conditions in chemical engineering.