Numerical modeling and Bayesian calibration of a bubbling fluidized bed with immersed horizontal tubes

Within the field of computational fluid dynamics (CFD), uncertainty quantification (UQ) is becoming increasingly important. Reporting simulation results without uncertainties can be misleading and potentially dangerous. In this paper we consider an isothermal, non-reacting bubbling fluidized bed with immersed horizontal tubes as a test problem for developing a CFD UQ framework. CFD model input parameters (e.g., coefficients of restitution, friction angles, turbulence models, and drag models) have inherent uncertainty associated with their values. A statistical analysis, including sensitivity and uncertainty analyses and Bayesian calibration, of relevant model parameters was performed. Sensitivity analyses suggests that all model parameters were contributing significantly to the output. An emulator (mathematical model) of the system was developed to further explore the state-space of the model parameters. Combined with Markov chain Monte Carlo methods, the emulator results favored a low friction angle for solid-solid phase interaction and the Wen-Yu correlation for the drag model, while remaining parameters were found to be non-influential. These results are currently being implemented in simulations of a bench-scale carbon capture system.