Kinetic Modeling and Multiobjective Optimization of Ibuprofen Synthesis Using Machine Learning

This study presents a comprehensive application of integrated machine learning tools for modeling and optimizing the ibuprofen synthesis process. Initially, a database of 39,460 input combinations is created using chemical reaction theory and validated with experimental data. The CatBoost meta-model, optimized by the snow ablation optimizer, outperforms conventional algorithms in predicting reaction time, conversion rate, and production cost. Importance analyses through SHAP values identify critical input variables, notably, the concentration of the catalyst precursor (L₂PdCl₂), hydrogen ions (H⁺), and water (H₂O), validating known catalytic principles and providing quantitative parameter guidance through data-driven analysis. Multiobjective optimization using NSGA-II generates a Pareto front of solutions, from which four industrial strategies are derived: balanced performance, maximum output, maximum yield, and minimum cost, each suitable for different production scenarios. The results identify optimal catalyst concentration ranges (0.002-0.01 mol/m³) that achieve high conversion rates while maintaining low costs. Uncertainty analysis conducted through Monte Carlo simulation reveals that reaction time exhibits particularly high sensitivity to parameter fluctuations, with a distinctive nonlinear response peaking at moderate perturbation levels (σ = 0.3). This study provides valuable insights for the rational design of ibuprofen synthesis conditions and demonstrates the effectiveness of integrating physics-based modeling with machine learning for chemical process optimization.