(569bb) Thermal Runaway of Chemical Reactors: An Experimental, Modeling Andmachine-Learning Investigation

Catalytic oxidation reactions are important in the chemical, petroleum, and pharmaceutical industries. The exothermic feature of catalytic oxidation reactions presents the risk of thermal runaway. Identifying the onset of thermal runaways is a critical task in controlling thermal runaways and eventually explosions. There are few experimental investigations of reaction thermal runaway in the literature because conducting an experiment near the onset of thermal runaway is highly risky. On the other hand, mathematical modeling of thermal runaway may be time-consuming although it typically gives reasonable prediction of thermal runaway. Following our prior work on the oxidation of methanol over Pt-Bi bimetallic catalysts, in the present work, we combined experimental, mathematical modeling, and machine learning methods to investigate thermal runaway behavior. An experimental reactor was constructed to measure the axial temperature distribution and observe hot spot formation. The effects of feed temperature, feed oxygen concentration, and catalyst design on the hot-spot formation and in turn reactor runaway was investigated. To increase the efficiency of the modeling phase, experimental data was input into a program that calculated instances of thermal runaway and compiled the calculations into a data set. The generated data set was then coupled with dimensionless parameters with weighted importance, a new parameter denoting the intensity of thermal runaway, and a machine learning model; to accurately identify thermal runaway in chemical reactors. Our method enables rapid evaluation of chemical reactor thermal behavior and for real-time applications in exothermic catalytic processes.