Transparent AI-assisted chemical engineering process: Machine learning modeling and multi-objective optimization for integrating process data and molecular-level reaction mechanisms

Abstract

Thoroughly utilizing the first principles of chemical processes, industrial big data, and artificial intelligence algorithms has been a deterministic trend in process modeling technology. However, the complexity of reaction networks triggers austere challenges to the deeper understanding of hybrid modeling techniques. Herein, this study explores a universal framework that integrates artificial intelligence algorithms and process mechanisms. The concept of “transparent AI-assisted chemical processes” is proposed. This study meticulously explains how to build a database integrating process big data and reaction network data. Based on the established machine learning framework incorporating process big data and mechanisms, multi-objective optimization algorithms are combined to achieve process optimization. The results indicate that the comprehensive dimensions of the optimized process's technical performance, economic performance, and environmental impact have energetically upgraded. Compared to the pre-optimization process, the optimized process's conversion rate and high-value product yield have increased by 4.44% and 4.25%, respectively. Moreover, the optimized process's non-renewable energy consumption and greenhouse gas emissions decreased by 6.23% and 12.60%, respectively.