Detailed simulations of fluid catalytic cracking in a rotating fluidized
bed in a static geometry are presented. A Eulerian–Eulerian
hydrodynamic model using the Kinetic Theory of Granular Flow is
combined with a 10-lump model for the reaction kinetics. To
account for coke formation, the appropriate deactivation functions
are introduced. To operate at low average catalyst coke content,
the catalyst residence time is such that it allows the catalyst to
make only a limited number of rotations in the reactor. Therefore,
the catalyst bed cannot be assumed well-mixed and the local distribution
function of the catalyst coke content is to be accounted for.
Because the latter has no pre-described functional form and to
reduce the calculation time, the catalyst coke content PDF is discretized,
introducing a number of catalyst classes with a given coke
content for which continuity equations are solved. Unsteady simulations
are carried out on a periodical domain containing multiple gas
inlet slots and radial solids inlets and accounting for the gas distribution
chamber.
The simulations demonstrate the importance of accounting for
multiple gas inlet slots and for the distribution of the gas via a chamber.
A dynamic behavior of the particle bed is observed, with both
meso- and macro-scale non-uniformities.
The flow pattern is dominated by slugging, a wave-like phenomenon.
Whereas at the micro-scale a non-solid body type particle bed
rotational motion is observed, the macro-scale slug shows on average
a solid body type rotational motion. Individual particles are seen to
rotate slower or faster than the slug, depending on their position in
the reactor, that is, relative to the slug. The role of the Coriolis effect
and of the tangential acceleration in the particle bed freeboard region
is demonstrated.
Despite the non-optimized fluidization chamber geometry and
resulting non-uniformities in the particle bed, significant process
intensification is achieved. Comparison with a conventional riser
reactor shows process intensification factors between 7 and 16,
depending on the gas oil conversion. The results finally illustrate
the capability of the proposed computationally trackable approach
to account for the catalyst coke content distribution function, and
confirm that the latter has no pre-described functional form.