A knowledge-driven approach for automatic generation of reaction networks of methanol-to-olefins process

Abstract

In recent years, automatic generation of reaction networks for chemical processes has attracted considerable attention. However, most of the automatic generators focused on the construction of reaction networks by computer and are essentially dependent upon the predefined reaction rules or categories. The reaction rules for constructing reaction networks are derived manually based on the individual’s understanding of reaction mechanism equation as well as the corresponding experiences, which hinders the automatic generation of reaction network directly from the complex reaction mechanisms or equations. In this article, we proposed a strategy to automatically identifies the reaction rules directly from the reaction equation using a Reaction Rules Topological Matrix Representation (RTMR) method. Taking the induction period of methanol to olefins (MTO) process as an example, we showed how to use RTMR to extract the information of reaction mechanism equation in the literatures, and then establish the reaction networks of the multiphase catalytic process. The results suggest that RTMR not only provides an effective approach for automatic reaction mechanisms equation extraction, but also demonstrates new reactions prediction ability after training with the literature information. It is expected RTMR can be promisingly implemented in knowledge-driven reaction network generation in multiphase catalytic processes.