LAUSR

Abstract A GPU-based, massively parallel coarse-grained CFD-DEM method was adapted to study the scale-up effect of residence time distribution of polydisperse particles in continuously operated multiple-chamber fluidized beds, where the bed is scaled up either by keeping a constant length of each chamber and increasing the number of chambers (Scaling method I) or by maintaining the number unchanged but increasing the length of each chamber (Scaling method II). It was shown that (i) the pressure drop and the residence time distribution of particles in a laboratory scale fluidized bed can be predicted reasonably well; (ii) when the solids feed rate is a constant (Operation I), the mean residence time of each type of particles scaled linearly as the bed length in both of Scaling methods I and II; (iii) when the solids feed rate is linearly increased with increasing bed length (Operation II), Scaling method I results in a continuously reduced ratio of mean residence time between coarse (medium) particles and fine particles, and the ratios level off with increasing bed length in Scaling method II; (iv) the mean residence time of particles can be inferred from the mean solids holdup obtained from CFD simulations and the mass fraction of particles at the inlet, thus offering a much faster way to estimate the mean residence time of particles; and (v) the equivalent number of perfect mixing tanks that corresponds to the mixing characteristics of real fluidized beds is linearly scaled as the bed length in Operation II, irrespective of Scaling methods.