Research on flow pattern recognition of bidirectional sinusoidal pulsating fluidized bed based on three-camera coupled image analysis

Abstract

Liquid-solid fluidized bed reactors have been widely used in various industrial fields. In order to improve the interphase transfer efficiency, a 2-D bidirectional sinusoidal pulsating intermittent liquid-solid fluidized bed was proposed and constructed in this work. Three high-speed cameras were used to capture the original images of the flow state simultaneously in the reactor. Based on the convolutional neural network toolbox, the image regression model was constructed and used to obtain the local solids holdup and particle distribution standard deviations in the fluidized bed. Then time domain analysis and frequency analysis methods were applied to extract the quadratic parameters. Under different fluidization states, the distribution characteristics of the power spectral or probability density curves show obvious differences. The typical flow patterns can be identified by simultaneously comparing the primary or quadratic parameters at three positions. This method provides a new tool for the study of complex multiphase flow reactors. • The bidirectional sinusoidal liquid pulsed L-S fluidized bed is investigated. • Three-Camera Coupled Image Analysis (TCCIA) method is proposed. • The flow patterns can be identified by comparing the primary/quadratic parameters. • TCCIA method provides a new tool for complex multiphase flow reactors.