Figure 6. Box plots of local and overall flux. For local flux, the fast fluidized bed had 1320 datasets15, while the turbulent fluidized bed had 255 datasets21. For overall flux, the fast fluidized bed had 1320 datasets 15, while the turbulent fluidized bed had 378 datasets 21.
To understand the relative influence of the variables on overall and local flux, Figure 7 displays the results from the random forest analysis. For overall flux, the variables were negatively and positively correlated in the cases of fast and turbulent beds, respectively. Two further notes on this are highlighted. Firstly, it should be noted that the magnitude of the estimates for overall flux in fast fluidization was approximately half that in turbulent fluidization, which is because the most dominant influence by far in the former has been identified to be the pressure at the bottom of the riser 27. Secondly, the positive relationship between overall flux and particle properties (namely, dave and ρp ) in turbulent fluidization is expected, since the overall flux has been linked to the collisional momentum transfer effects earlier20. Specifically, it was shown for the binary mixture of narrow PSDs of glass (referred to as ‘fine’ due to the lower terminal velocity) and PS (referred to as ‘coarse’ due to the lower terminal velocity) that the overall flux increased then decreased with the percentage mass of the fines (i.e., the glass species), and the non-monotonic behavior was tied to a trade-off in which collisional momentum transfer from fine to coarse particles dominated at lower fine contents and a decreasing amount of coarse particles dominated at high fine content. In contrast to overall flux, the relative influences for local flux were similar between the fast and turbulent beds for local flux. With respect to local flux, r/R was the most dominant, followed by h/H , implying the importance of position within the riser relative to other variables. The radial dependence is expected for the fast fluidized bed due to the well-acknowledged annulus flow behavior 19,28,29, but surprising for the turbulent bed wherein core-annulus flow structures are not generally acknowledged to exist 4. Notably, the turbulent fluidized bed data presented here has been shown to display more prominent radial variations of the local fluxes for some particle systems, namely, the narrower lognormal PSDs and pure polystyrene systems21, which suggests differences between the core and annulus flows. More studies are needed to ascertain this.