Machine-Learning Approach to Modeling Oxidation of Toluene in a Bubble Column Reactor

A feed forward machine-learning (ML) model is applied to study bubble induced turbulence and bubble mass transfer in a bubble column reactor. Using direct numerical simulation data for forced turbulence, bubble deformations and flow velocities are predicted. To predict mass transfer, ML sub-grid scale (SGS) modeling technique is introduced for the concentration of reactants and products undergoing parallel competitive reactions in the oxidation of toluene. The ML model replaces the iterative approach associated with the use of analytical profiles for previous SGS models for correcting concentration profiles in boundary layers. The present model, thus, offers a significant performance bonus as well as the flexibility to extend to more complex scenarios due to its data-driven nature.