Valorization of Polycarbonate Waste by Ultrasound-Assisted Methanolysis: Optimization Using an Artificial Neural Network and Kinetic Investigation

This study presents an integrated approach combining experimental optimization, machine learning (using ANN-GA), and kinetic modeling to enhance ultrasound-assisted methanolysis of polycarbonate (PC) waste. Central composite design (CCD) under response surface methodology (RSM) was employed to optimize methanolysis reaction parameters of solvent ratio, catalyst loading, temperature, reaction time, and lagging time. The optimized process achieved 98.56% PC conversion under 0.138 g of the KOH catalyst, 3.63:1 THF: methanol ratio, 47.7 °C, 26.8 min reaction time, and 7 min lagging time. Product characterization confirmed the formation of high-purity bisphenol A (BPA) and dimethyl carbonate (DMC). To enhance predictive capability, an artificial neural network optimized with genetic algorithm (ANN-GA) was developed. The model with 12 hidden neurons improved the predictions of RSM-CCD with 99.95% PC conversion and identified the solvent ratio and catalyst dosage as the most influential variables through sensitivity analysis. A kinetic model based on four reversible reaction steps was solved using the 4th-order Runge–Kutta method with GA-optimized constants, showing strong agreement with experimental data (R2 > 0.94). The kinetic constants provided mechanistic insight into the role of ultrasound, which enhanced depolymerization by improving mass transfer and reducing activation barriers. This work demonstrates an efficient, sustainable pathway for polycarbonate waste valorization and green synthesis of BPA and DMC.