Ring opening of epoxidized castor oil with applied hybrid kinetic modelling model of particle swarm & simulated annealing

Abstract

This study comprehensively investigated the production of eco-friendly polyols through the in-situ epoxidation of castor oil, employing a hybrid kinetic modeling approach that combined Particle Swarm Optimization (PSO) and Simulated Annealing (SA). The epoxidation process was optimized using the Taguchi method, which identified stirring speed as the most significant process parameter, supported by a p-value of 0.000 and an F-value of 95.92. The reaction was carried out under optimized conditions, where 50 g of castor oil was reacted with hydrogen peroxide and acetic acid at a molar ratio of 1:1:1, a temperature of 65 °C, and a stirring speed of 200 rpm. The relative conversion of oxirane (RCO) was determined using the AOCS Official Method Cd- 957. As the reaction progressed, the near-zero RCO values confirmed complete hydroxylation. The epoxidized castor oil was then mixed with various hydroxylation reagents at epoxide-to-reagent molar ratios of 1:0.5, 1:1, and 1:1.5 to evaluate the hydroxylation rate. The results showed that all reagents achieved the fastest hydroxylation at the highest molar ratio of 1:1.5. The synthesized polyols were categorized based on their hydroxyl values, revealing that polyols produced using peracetic acid (79.3 mg KOH/g), water (85.0 mg KOH/g), and hydrogen peroxide (89.1 mg KOH/g) were suitable for flexible polyurethane applications. In contrast, polyols derived from methanol (127.9 mg KOH/g), acetic acid (139.4 mg KOH/g), and water (108.1 mg KOH/g) exhibited hydroxyl values between 100 and 250 mg KOH/g, making them more suitable for semi-rigid polyurethane applications. Kinetic parameters were determined through MATLAB R2023 A simulations, yielding reaction rate constants for key steps in the epoxidation and hydrolysis processes: k₁ = 0.03 M⁻¹ min⁻¹, k₂ = 0.00 M⁻¹ min⁻¹, k₃ = 30.00 M⁻¹ min⁻¹, and k₄ = 0.050 M⁻¹ min⁻¹. The hybrid PSO + SA simulation model demonstrated a strong correlation with experimental data, achieving an R² value of 0.9961, significantly outperforming the individual PSO model (0.9836) and SA model (0.9779).