Particle size segregation in bi and penta-disperse gas–solid fluidized beds: CFD-DEM and recurrence CFD simulations

We studied the time evolution of segregating particle fractions in gas–solid fluidized beds by unresolved CFD-DEM and data-based recurrence CFD (rCFD). First and foremost, we incorporated segregating particle fractions into rCFD by superposing a fractional drift velocity to the velocity of the solid bulk material. From an algorithmic viewpoint, fractional drift was implemented by discrete face swaps. We tested this novel rCFD approach with different superficial gas velocities and different bi-disperse bed inventories. rCFD predictions agreed very well with corresponding full CFD-DEM results with respect to the time-evolution of fractional centre-of-gravity as well as spatial line profiles of fractional mean volume fraction. In order to explore the validity range of the underlying databases, we intentionally ran rCFD at deviating conditions. In a first test, we showed that rCFD simulations could indicate regime-changing de-fluidization even if de-fluidization itself was not pictured in the database. In a second test, we performed penta-disperse rCFD simulations based on an equivalent bi-disperse database. Hereby, we found out that rCFD predictions were only accurate if the penta-disperse bed inventory did not lead to changing overall bed dynamics. Finally, we proved the feasibility of rCFD simulations of a granulation process featuring local particle growth as well as continuous feed input and product output. In all cases, rCFD simulations ran more than four orders of magnitude faster than corresponding CFD-DEM simulations, eventually allowing for real-time simulations of spatially resolved particle segregation in fluidized beds.