Simulation Of Non-spherical Particle Two-phase Flow In Fluidized Beds Based On Neural Network

Gas-solid fluidization system has been widely applied in various industries,including energy,chemical engineering,food,aerospace,and so on.In a gas-solid fluidization system,particles are normally irregularly-shaped and non-spherical.To simplify the computation in traditional methods,particles are frequently assumed to be perfectly spherical.However,in many real granular systems,such a kind of oversimplified assumption cannot characterize the flow behavior of the particles in the fluidization systems,which seriously influences the design,optimization,modelling,and actual operation of the fluidized bed reactor.Therefore,it is very necessary to investigate the non-spherical particles hydrodynamics of gas-solid twophase flow in typical fluidized beds.Numerical simulation is a common and effective way to predict the hydrodynamics of complex gas-solid two-phase fluidization systems.In this project,Euler-Euler two-fluid model will be employed to investigate the hydrodynamics of non-spherical particle of gas-solid two-phase flow in typical fluidized beds.Due to the complexity of particle shape,it is difficult to effectively predict the drag coefficient betweent gas phase and solid phase.The gas-solid drag coefficient of spherical particles cannot reflect the real characteristics of drag force.This is one of the difficulties of numerical simulations on hydrodynamics of non-spherical particle gas-solid two-phase flow.Considering the rapid development of artificial intelligence method,the artificial neural network method will be adopted to develop a prediction method and a correlation for non-spherical particle drag coefficient.Then the proposed drag coefficient correlation will be embedded in Euler-Euler two-fluid model to predict the hydrodynamics of non-spherical particles in several typical gassolid fluidized bed systems.The gas-solid drag coefficient of non-sphercial particle is predicted and analzed by the artificial neural network method.First,back propagation neural network(BPNN)and radial basis function neural network(RBFNN)are compared in the prediction of drag coefficient in the experiments of Song et al.Results show the RBFNN method has smaller errors and higher efficiency in predicting the gas-solid drag coefficient of non-spherical particles.Moreover,the RBFNN is applied to predict and analyze the drag coefficients for particles with different sphericities.It is found that the artificial neural network method can be employed in the prediction of gas-solid drag coefficient of non-spherical particles.The research results can provide an effective way to investigate the drag coefficient of complex-shaped particles.The non-spherical particle two-fluid model is further developed in this work.Particle Image Velocimetry(PIV)is adopted to experimentally investigate the flow behavior of non-spherical particle of gas-solid two-phase flow in the spouted bed.The velocity distribution and drag force distribution of non-spherical particles in the regions of spout,fountain,and annulus is obtained.Experimental results reveal that spout height hardly changes while the fountain height increases with the increase of the spout gas velocity.Moreover,the established non-spherical particle two-phase flow model is employed to investigate the hydrodynamics of gas-solid two-phase flow in the spouted bed.Simulation results are compared with the experimental data and the comparison results are satisfactory.It is further validated that the proposed drag coefficient correlation is applicable in the spouted bed.Applying Euler-Euler two-fluid model with the drag coefficient corelation based on the artificial neural network,the hydrodynamics of the bubbling fluidized bed,including the profiles of particle velocity,granular temperature,isotropy and flatness factor distributions are numerically studied.Good agreement is obtained between simulation results and experimental data,which validates the applicability of the modified drag model.Results reveal that particles move upward in the central region,and downward in the wall region,which shows a core-annulus flow.Particle shape hardly affects the horizontal particle velocity and has a certain effect on the vertical particle velocity.The effect is not a monotonous functional relationship.Particle shape mainly influences the fluctuating energy in the high-frequency region and hardly affects it in the low-frequency region.The drag coefficient corelation based on artificial neural network is embedded in two-fluid model to simulate the hydrodynamics of non-spherical particles in the riser.The effect of sphericity,particle density and particle size on the hydrodynamics of gas-solid two-phase flow is analyzed.Simualtion results agree well with experimental data,demonstrating the applicability of the drag model.Results show that the difference of particle sphericity significantly affects the solid concentration distribution and mesoscale cluster formation in the riser.Particle sphericity mainly influences the fluctuating energy in the high-frequency region and hardly affects it in the low-frequency region.