Measuring binary fluidization of non-spherical and spherical particles
using machine learning aided image processing
Abstract
The binary fluidization of Geldart-D type non-spherical wood particles
and spherical LDPE particles was investigated in a laboratory-scale bed.
The experiment was performed for varying static bed height, wood
particles count, as well as superficial gas velocity. The LDPE velocity
field were quantified using Particle Image Velocimetry (PIV). The wood
particles orientation and velocity are measured using Particle Tracking
Velocimetry (PTV). A machine learning pixel-wise classification model
was trained and applied to acquire wood and LDPE particle masks for PIV
and PTV processing, respectively. The results show significant
differences in the fluidization behavior between LDPE only case and
binary fluidization case. The effects of wood particles on the slugging
frequency, mean, and variation of bed height, and characteristics of the
particle velocities/orientations were quantified and compared. This
comprehensive experimental dataset serves as a benchmark for validating
numerical models.