Due to a large particle size and a small specific surface, a homogeneous mixing of biomass pellets with bed materials during gasification plays a critical role in the devolatilization and carbon conversion. In this work, the mixing evolution behavior of biomass pellets at different gasification stages is investigated for the first time. Two bubbling fluidized beds are established to perform the preparation of biomass samples undergoing different gasification times and visualized mixing experiments, respectively. An image processing technique is introduced for the determination of the real-time distribution of biomass pellets. The vertical and lateral migration paths of biomass pellets at different gasification stages are revealed. The improvement of binary mixing by adjusting the operating conditions as well as the adaptability to different biomass loadings are discussed. A convolutional neural network is developed to validate the influence of fluidization velocity on the resulting flow and classify the fluidization behavior.
