(65e) CFD-DEM Simulations and Uncertainty Quantification (UQ) of Horizontal Jets in Gas-Solid Fluidized Bed | AIChE

(65e) CFD-DEM Simulations and Uncertainty Quantification (UQ) of Horizontal Jets in Gas-Solid Fluidized Bed


Liu, P. - Presenter, University of Colorado at Boulder
Fullmer, W., National Energy Technology Laboratory
Dahl, S. R., University of Colorado at Boulder
Hrenya, C. M., University of Colorado at Boulder
A recent experiment conducted at PSRI of opposing horizontal air jets impinging into a semi-circular fluidized bed of 6 mm plastic beads is studied here computationally. The experimental particle count is ~60,000, making it amenable to direct comparison with discrete-particle simulations. We use MFiX CFD-DEM to simulate this problem. The gas phase is treated as compressible due to the high velocity of the jets. All uncertainties in model inputs are considered other than those relating to the geometry of the jets, although the uncertainty in the area is considered in the jet velocity. An extensive parameter identification and ranking table (PIRT) is constructed. We use a local sensitivity analysis to rank the impact of all model input uncertainties on the system response quantities (SRQs), i.e., the model output variables of interest in this problem. Four SRQs are considered: the time averaged mean- and standard deviation of the bed pressure drop and the left- and right-side jet penetration depths (which were calculated experimentally from a high speed video imaging analysis). Somewhat surprisingly, the PIRT reveals that the superficial distributor velocity is, by far, the most important parameter controlling the jet penetration depts. After eliminating low impact uncertainties, we directly propagate the remaining uncertainties through the model to determine the uncertainties in the output SRQs. The results show that the CFD-DEM-based uncertainty quantification (UQ) bounds the jet penetration depths, though it is worth noting that the predicted bounds are significantly wider than the experimental data. Results for the bed pressure drop does not bound the experimental data, showing a slight bias, namely a 10 – 15% overestimation. These comparisons and their implications on CFD-DEM+UQ analysis will be discussed.