(74d) CFD-DEM Simulations of Bubbling Fluidization: Global Sensitivity Analysis for the Identification and Validation of Critical Model Parameters | AIChE

(74d) CFD-DEM Simulations of Bubbling Fluidization: Global Sensitivity Analysis for the Identification and Validation of Critical Model Parameters


Bakshi, A. - Presenter, National Energy Technology Laboratory
Shahnam, M., National Energy Technology Laboratory
Li, T., National Energy Technology Laboratory
Altantzis, C., National Energy Technology Laboratory
Gel, A., Arizona state university
Rogers, W. A., National Energy Technology Laboratory
Ghoniem, A. F., Massachusetts Institute of Technology
Bubbling fluidized beds are used extensively in the energy and chemical industries because of their excellent heat and mass transfer characteristics. The performance of these reactors is strongly dependent on particle-scale phenomena such as chemical conversion, clustering, cohesion and their interaction with reactor-scale transport. In this context, CFD-DEM simulations are extremely valuable because particles are individually tracked and their interactions are completely resolved. However, despite significant progress over the past decade, there is considerable uncertainty regarding the choice of particle-interaction parameters such as the stiffness, restitution and interphase drag force, most of which cannot be measured experimentally (under typical operating conditions of reactors). Therefore, they are often treated as tunable parameters in simulations and their impact (highly coupled and non-linear) on hydrodynamic predictions is largely unknown. In this study, an exhaustive global sensitivity analysis is performed and statistical tools are employed to (a) identify critical parameters in CFD-DEM simulations and (b) validate these with experimental data.

For the bubbling fluidization of Geldart B particles, a large parametric space is initially considered including model, operating and numerical parameters in CFD-DEM simulations. The set of insensitive parameters is first dicounted using MOAT (Morris One-At-Time) screening analysis [1] and thorough sensitivity analysis is then performed to quantify the dominant and coupled interactions of the remaining sensitive parameters. Numerical experiments are designed using Latin-hypercube based sampling, optimized to ensure sampling uniformity in the parameteric space. Critical parameters are identified using response surface methodology based on bubble statistics and solids circulation metrics as quantities of interest [2] and these parameters are validated using experimental measurements in a thin rectangular pulsating fluidized bed. All numerical CFD-DEM simulations are conducted using open-source software MFiX and bubble statistics are computed using MS3DATA [3]. This is one of the first large-scale parametric studies of CFD-DEM simulations and the developed framework will guide the selection and validation of critical sub-models in other applications of particle-gas flows.


[1] F. Campolongo, J. Cariboni and A. Saltelli, An effective screening design for the sensitivity analysis of large models. Environmental Modelling and Software 22:1509-1518, 2007

[2] A. Bakshi, C. Altantzis, R.B. Bates and A.F. Ghoniem, Eulerian–Eulerian simulation of dense solid–gas cylindrical fluidized beds: Impact of wall boundary condition and drag model on fluidization. Powder Technology 277: 47–62, 2015

[3] A. Bakshi, C. Altantzis, R.B. Bates and A.F. Ghoniem, Multiphase-flow Statistics using 3D Detection and Tracking Algorithm (MS3DATA): Methodology and application to large-scale fluidized beds. Chemical Engineering Journal, 293: 355-364, 2016