Numerical Study of Ozone Decomposition Reaction Behaviours in Gas-Solids Circulating Fluidized Bed Reactors
Abstract
This study numerically investigated the reaction behaviours of catalytic ozone decomposition reaction in a 10.2-meter-tall gas-solids circulating fluidized bed (CFB) reactor. A pseudo-homogeneous reactive transport model for ozone decomposition, integrated with the two-fluid model, was developed and validated using experimental data. Based on the model, the impacts of turbulence models, specularity coefficients, and simulation methods on reaction behaviours in the CFB riser reactor were explored. These three factors were found to significantly affect the hydrodynamic characteristics and the reaction behaviours in the riser. A comparative study of CFB riser and downer reactors was conducted. Operations in the direction of and against gravity resulted in drastically different flow fields and particle clustering in the two reactors. More uniform flow and reaction fields make the downer have better gas-solids contact efficiency than the riser. Flow structure and residence time distributions of gas and solids in the riser and downer were characterized by tracing the gas and particles. The results showed that flow in the downer reactor resembles plug flow, while significant axial backmixing occurs in the riser. An internal circulation mechanism is proposed to explain the backmixing. A sub-grid reactive transport model was developed using a filtering method and an artificial neural network (ANN) to explore the impact of particle clustering on reaction behaviours. In the development of the filtered model, employing gradient features as inputs enhanced regression accuracy. Additionally, ANN demonstrated superior performance over traditional fitting methods. Consequently, the filtered reactive transport model showed improvements in predicting the reaction behaviours in a CFB riser. In summary, the hydrodynamic characteristics within CFB predominantly influence reaction behaviours. High-resolution simulations combined with machine learning techniques effectively aid in understanding mechanisms in fluidization system and developing new models, which are crucial for designing and optimizing large-scale reactors.