(385y) Understanding the Pyrolysis Process of Polyethylene By a Coupled Method of Machine Learning and Molecular-Level Kinetic Model

The industrialization and the development of the global economy have promoted the production of plastics, resulting in an overwhelming amount of waste plastic that requires effective solutions for resource utilization. Pyrolysis can convert plastic into higher value-added gas/liquid products, which can be utilized as raw materials in the chemical industry.

Several studies have been conducted in experimental aspect of waste plastic pyrolysis, but it is difficult to obtain a comprehensive understanding of the process from scattered experimental data due to its complexity. Thus we adopted the structure-oriented lumping (SOL) method to provide an approach of a molecular-level kinetic model that effectively utilizes limited pyrolysis data. Based on PE pyrolysis system, the model automatically generates complex reaction networks, realizing rapid analysis of data and optimization of model parameters. It also retains the basic information of the reaction mechanism, being suitable for rapid simulation and production guidance of pyrolysis designs.

Machine learning algorithms with powerful data fitting ability can assist in the accurate prediction of product yields and thus guide the scale-up design of pyrolysis processes, but they often face challenges due to insufficient experimental data samples and large search spaces. We extrapolate the molecular weight distribution of the raw materials to obtain sufficient data, which is then re-entered into the molecular-level kinetic model to compute the distribution of gas and liquid phase products. Subsequently, we use the raw material distribution, system parameters, condition parameters, and kinetic parameters as input factors, with gas and liquid yields as output targets, to construct machine learning models and achieve excellent predictive accuracy. Additionally, partial dependence plots quantify the impact among input factors and their effects on yield, providing in-depth insights into the pyrolysis process and enhancing yield optimization.

In conclusion, this study not only advances the theoretical framework for polymer degradation but also provides practical insights into the optimization of pyrolysis processes, offering a pathway to efficient waste plastic recycling and the sustainable production of valuable chemicals.