Integration of physical information and reaction mechanism data for surrogate prediction model and multi-objective optimization of glycolic acid production

Abstract

With the continuous development of the chemical industry, the concept of advocating green development has become increasingly popular. Glycolic acid (GA), serving as the monomer for biodegradable plastic polyglycolic acid, plays a crucial role in combating plastic pollution and fostering an eco-friendly society. The selective oxidation of ethylene glycol (EG) to produce GA represents a novel green production technology. Controlling reaction parameters to achieve multi-objective optimization of product distribution and direct CO2 emissions is crucial for scaling up the process. With the advent of the big data era, the integration of the chemical industry with artificial intelligence to achieve engineering scale-up is an important trend. This study proposes a neural network model for production prediction and optimization. The model is trained using experimental data, reaction mechanism data, and physical information, enabling rapid prediction of GA production. After validating with 40% of experimental data and 16% of reaction mechanism data, the model's prediction error was within ±5%, and the linear correlation coefficient R2 between the predicted values and actual values was 0.998. Furthermore, this study integrated a multi-objective optimization algorithm based on the model, enabling surrogate optimization of reaction parameters during production. After optimization, the direct CO2 emissions were reduced by over 99% and overall greenhouse gas emissions were reduced by 4.6%. The research paradigm proposed in this research can offer guidance and technical support for the optimized operation of EG selective oxidation to GA.