Numerical Simulations of the Fluidized Bed Experiments

Current Status of the Project

The attention of this group has been on the simulation of bed motion and chemical reactions in bubbling fluidized beds, rather than in circulating fluidized beds. In bubbling beds, the particle loadings are much higher (up to close packing). In order to be able to calculate the bed dynamics, heat and mass transfer and chemical reactions in a timely way, the numerical efficiency of the MFIX code was improved dramatically: 1) MFIX was parallelized, to allow the utilization of clustered computers; 2) 4th order spatially differencing was implemented, to allow higher accuracy without increasing computational time; and 3) and 2nd order (Crank-Nicholson) time advancement was added, to guarantee time accuracy and improve stability. These enhancements are purely numerical and are important regardless of the actual simulation.

The NETL group has cooperated with Dow Corning in demonstrating the ability of MFIX to accurately simulate chemical reactions in fluidized beds. A very common method for experimentalists to demonstrate the "contacting" behavior of a bed is by using ozone decomposition as a tracer. A well-documented publication (Fryer and Potter, 1976) was simulated, resulting in very accurate agreement with the experimental ozone conversion, as the mesh size was resolved. Additional simulations of SiH4 and SiCl4 decomposition, as reported in the open literature, were performed. These simulations also provided reasonable results, although there was limited data available for validation. However, the feasibility of the calculations which involved more relevant chemistry schemes was clearly demonstrated.

There were also several improvements to MFIX that are related to improvements in physical models: 1) periodic boundary conditions were implemented, for use by Prof. Sundaresan's group at Princeton, in the development of sub-grid scale models which are essential for the simulation of facilities of commercial size; 2) equations for multiple particle types were included, to allow description of beds with a particle size distribution or mixtures of different particles (as most beds are); 3) frictional flow models were enhanced, again in cooperation with the Princeton group, to more accurately describe motion of the densely packed bed; 4) the scalar transport equation was coupled into the solution scheme to allow for the easy addition of advanced models, such as "granular temperature" transport and the LANL turbulence model.

Methods of analysis of bed hydrodynamics have been developed by the ORNL group and incorporated into the postprocessing code Post-MFIX. These advanced time-series analysis tools can be used to analyze both experimental information (e.g., pressure signals) and detailed simulation results. They have been used to accurately quantify the effect of the variation of model parameters and modifications of numerical methods on the results of computer simulations. Specifically, these tools have demonstrated a type of long-term fluctuation in bed dynamics that had not previously been characterized; this has been, subsequently, verified by laboratory experiments . Additionally, image-analysis algorithms for experiments and methods of creating virtual sensor measurements from detailed simulation results have been developed to accurately and consistently compare experimental and simulation results.

Over the course of these activities a number of collaborators have acquired the MFIX code and are using it in their research projects. Those groups with whom the NETL/ORNL development staff have maintained a close relationship are Prof. Sundaresan's group at Princeton University, Prof. Murthy's group at Purdue University (formerly at Carnegie Mellon University) and a group of three professors (Profs. Rodney Fox, David Hoffman, and Francine Battaglia) and their students at Iowa State University. From the Princeton group, three graduate students have already obtained their Ph. D. degrees for studies using the MFIX code; at Purdue/CMU, three graduate students and one post-doc are working with MFIX; at Iowa State, three graduate students are working with the code. At the University of Tennessee, one student has received his Ph. D. with partial support from this program. In addition, two students of Prof. Aubrey Miller at West Virginia University have utilized MFIX for their M.S. degree work. An industrial staff member of Dow Corning used MFIX for her Ph. D. work under Prof. Gidaspow at the Illinois Institute of Technology. The NETL/ORNL development staff is working closely with industrial workers at Dow Corning, Millennium Inorganic Chemicals, and ExxonMobil (Univation). In order to make this resource widely available, and to encourage collaboration, an MFIX web-site has been established ( www.mfix.org) The code has been acquired from the web site by some 30 professionals at universities, government labs, or in industry.

The activities funded under this consortium have led to several publications and reports at scientific meeting:

1. M. Syamlal and T. O'Brien, " Fluid Dynamic Simulation of O3 Decomposition in a Bubbling Bed," Proceedings of the AIChE Annual Meeting, Los Angeles, November 2000.
2. Daw C.S., J.S. Halow, C.E.A. Finney, K. Nguyen, "Characterizing the hydrodynamics of bubbling fluidized beds with multivariate pressure measurements," Proceedings of the AIChE Annual Meeting, Los Angeles, November 2000.
3. Ed D'Azevedo, S. Pannala , M. Syamlal, A. Gel, M. Prinkey, and T. O'Brien, "Parallelization of MFIX: A Multiphase CFD Code for Modeling Fluidized Beds" presented at the Tenth SIAM Conference on Parallel Processing for Scientific Computing, March 12-14, 2001, Portsmouth, Virginia, 2001.
4. C. Guenther, T. O'Brien, and M. Syamlal, "A Numerical Model of Silane Pyrolysis in a Gas-Solids Fluidized Bed," presented at the International Conference on Multiphase Flow, New Orleans, May 27-June 1, 2001.
5. M. Syamlal, T. J. O'Brien, S. J. Gelderbloom, S. Pawelkowski, and S. Pannala, "Simulation of Trichlorosilane Reactor," presented at the MFDRC review meeting, Midland, MI, October 2000.
6. T. O'Brien and M. Syamlal, "Simulation of the hydrodynamics of a bubbling fluidized bed", Fluidization X . eds. M. Kwauk, J. Li, and W.C-Yang, Proceedings of the 10th Engineering Foundation Conference on Fluidization, Beijing, P.R. China, May 20-25, 2001.
7. C. Guenther and M. Syamlal, "The effect of numerical diffusion on simulation of isolated bubbles in a gas-solid fluidized bed," Powder Technology, 116, 142-154 (2001).
8. Agrawal, K., P.N. Loezos, M. Syamlal, and S. Sundaresan, "The role of meso-scale structures in rapid gas-solids flows," accepted for publication, Journal of Fluid Mechanics, 2001.
9. M. Syamlal and T.J. O'Brien, "Simulation of a catalytic reaction in a bubbling fluidized bed," presented at Chemical Reaction Engineering VII: Computational Fluid Dynamics, August 6-11, 2000, Quebec, Canada.
10. C. Guenther, M.Syamlal, and T.J. O'Brien, "Simulation of the fluidized bed pyrolysis of silane," presented at Chemical Reaction Engineering VII: Computational Fluid Dynamics, August 6-11, 2000, Quebec, Canada.

Plans for Future Research

Utilizing the advanced numerical capabilities within MFIX, developed within the first years of this Consortium, the activities of this group will focus on detailed comparison with experimental activities to validate the capabilities of the code. Specific plans for the next two fiscal years are:

1. Chemically reactive simulations --
   1. Prepare a publication of the 2-D ozone simulation (FY02, Q1-2)
   2. Follow up with the 3-D ozone calculations/paper, using the 4th order numerics and Crank-Nicholson time stepping (FY02, Q3-4)
   3. Prepare a publication of the SiH4 simulation, including a discussion of the effect of chemistry on hydrodynamics (FY02, Q2)
   4. Implement a polyethylene chemistry scheme, in cooperation with Exxon/Mobil (Univation) (FY02, Q3-4)
2. Effect of vessel size on bed hydrodynamics
   1. Direct simulations of lab to pilot scale (FY02, Q3)
   2. Comparison with Werther's experiments (FY02, Q4)
   3. Prediction of effects at industrial scale (FY03, Q1-2)
   4. Use of subgrid scale models (FY03, Q3-4)
   5. Use of granular phase turbulence models (FY03, Q3-4)
3. Implementation of the LANL turbulence model into MFIX
   1. Incorporation into the MFIX solution scheme (FY02, Q1-2)
   2. Application to circulating fluidized beds (FY02, Q3-4)
   3. Application to bubbling fluidized beds (FY03)
4. Exploration of coupled solver strategies for solution of the equation set
   1. using PETSc tools (developed by DOE-OS-BES) (FY02, Q2-3)
   2. plan restructuring of MFIX (FY02, Q4)
   3. initiate restructuring of MFIX (FY03, Q1-4)
5. Simulation of 2-D Bubbling bed experiments at ORNL
   1. Central jet experiments (FY02, Q1-2)
   2. Quantitative comparison of a) with MFIX simulations
      1. 2-D vs. 3-D (FY02, Q1-4)
      2. Model variations: granular stress, boundary conditions, (FY02, Q2-4)
   3. Freely bubbling experiments (FY03, Q1-2)
   4. Quantitative comparison of c) with MFIX simulations
      1. 2-D vs. 3-D (FY03, Q1-2)
      2. Model variations: granular stress, boundary conditions(FY03, Q2-4)
   5. Mixing/segregation experiments with diagnostics (FY03, Q1-2)
   6. Quantitative comparison with MFIX multiple solid phases capabilities (FY03)