4. Conclusion

Using high-speed imaging, the current study provides detailed experimental data on the effects of non-spherical wood particle counts, static bed height, and superficial gas velocity on the binary fluidization behavior of a Geldart-D particle type system. Machine learning-aided image analysis is applied to extract particles mask of a specific type followed by PIV and PTV analysis to quantify the LDPE particle velocity field and the velocity and orientation of wood particles, respectively.
In general, the fluidization for LDPE particles only and LDPE/wood particles behave rather differently. While LDPE particles slugging are achieved at a low gas flow rate of 1.4Umf and 2Umf, the bed transitions to a spouting regime at higher flow rates of 3Umf and 3.4Umf. This transition is clearly observed in the mean and standard deviation of the expanded bed height. To our knowledge, the spouting behavior of Geldart D type particles has not been reported before under uniform flow conditions using porous distributor plates without gas jets. The resultant maximum downward velocity within the fountain regions is significantly lower than those obtained using a central gas jet even with lower gas velocity.
With the addition of wood particles, the spouting phenomenon has not been observed under all tested conditions varying static bed height, wood particles counts, and fluidization velocity. At a low gas flow rate (1.5 Umf), the wood and LDPE particles remain perfectly segregated under different static bed heights and wood particle counts while the LDPE particles form slugging flow. At 2.0Umf– 3.7Umf, wood and LDPE particles are mixed and binary slugging flows are formed. At 2Umf, adding wood particles also causes a 26% increase in the rising speed of the slug and a 112% increase of the maximum fall speed after the slugs disintegration. Furthermore, the spatially averaged vertical velocity shows quasi-periodic fluctuations for all cases except for the spouting bed for the LDPE only case at 3Umf. At 2Umf, the mean amplitude of the signal almost doubled, while the mean frequency sees a 38% increase with the addition of wood particles. The increase of fluidization velocity for the binary fluidization case to 3Umf causes a 110% increase of the mean amplitude and a 35% decrease in mean frequency. Furthermore, the mean bed height seems to collapse at different static bed heights after nondimenlization. The evolution of the spatially averaged vertical velocities behaves similarly under different static bed heights. After using the characteristics falling velocity and time for nondimenlization, the mean periods seem to collapse reasonably well between different bed height and fluidization velocities. However, the scaling only works for mean amplitude under the same static bed height.
Finally, the wood particles preferentially orientate themselves towards 0° (vertical) and 90° (horizontal) for all tested cases at dilute and dense region of the bed respectively. No significant orientation distribution differences are identified under tested conditions. The horizontal wood particle velocities show a Gaussian distribution for all cases and a varying degree of bimodal distribution for the vertical velocities. One unresolved issue is the systematic quantification of distribution. As the result shows, the wood particles demonstrate varying degrees of segregation with varying superficial gas velocity and wood particle fraction. The current system has limited wood particle counts which inherently causes the segregation analysis noisy. A larger system with a significant number of particles is preferred for meaningful segregation analysis and is deferred to future studies.