Mathematical Simulation Of Particles Separation Process And Design Of Double Layer Spreading Plate Of Turbo Air Classifier

The powder obtained by crushing and grinding can't meet the demand because of its wide range of particle size distribution, so classification is necessary. Due to its high classification efficiency and large processing capacity, the turbo air classification has become a mainstream dynamic classifier since it is emerged. With the more and more application of micro powder in all walks of life, improvement of the turbo air classification performance becomes an urgent problem to be solved. In this paper, in order to improve the dispersion of the powder particles in the classifier and classification performance of the classifier,the analyses of the particle movement and the simulation using Fluent are carried out to optimize the spreading plate. Moreover, a stochastic model of particle classification based on the analyses of the particles' motion in the annular region is established and the Monte Carlo simulation is performed to further reveal gas-solid classification mechanism. And the prediction model of BP neural network based on the experimental data is established to predict the classification performance accurately. The main research contents are as follows.According to the structural characteristics of the turbo air classifier,the classifier model is divided into four parts generate the grids respectively. The grid independence verification is implemented and the grid number is confirmed. Different turbulence models, discrete schemes and pressure velocity coupling modes are analyzed. The RNG k-? model,the first order upwind scheme and the SIMPLEC pressure velocity coupling method are used to simulate the inner air flow field of the turbo air classifier. The forces on the particles in the gas phase are analyzed.Under the given operating conditions of the turbo air classifier, the quality rate and the volume rate are calculated to determine gas-solid two-way coupling of the DPM discrete phase model.The velocity formula based on the analyses of the particles' motion on the single layer spreading plate is obtained when the particles leave the spreading plate. The particle distribution based on the theoretical calculation is carried out after the particles leave the spreading plate. A double layer spreading plate with uniform fan-shaped slot and radial barrier on the upper spreading plate is designed. The numerical simulation, simulated experiments and material classification experiments are carried out to verify the improvement effect. Numerical simulation results indicate that the flow field distribution is basically the same between the single layer spreading plate and double layer spreading plate.But the discrete phase simulation results show that particle travelling time via the upper and lower spreading plate is different, which can improve the material dispersion and decrease the probability of collision and aggregation. And the cut size is decreased and the classification accuracy is improved using the double spreading plate. The simulated experiment results show that the distribution of the material via the double spreading plate is wider and the maximum relative reduction of the average material amount per unit area is 29.2% compared to the single spreading plate,which means the double spreading plate can improve the material dispersion. The classification experiments results show that compared to the single layer spreading plate the cut size can be decreased up to 25.24% and the classification accuracy can be increased up to 52.47% for the double spreading plate under the different conditions.In order to describe the classification process of particles in turbo air classifier, not only the necessity but also the randomness of particles'motion should be considered. According to the characteristics of a turbo air classifier, the random mathematical model of particle classification using triangular Markov chain is established and Monte Carlo method is used to calculate the stochastic model. The comparing classification experiments show that the partial classification efficiency curve predicted by the Monte Carlo simulation is closer to the ideal grading efficiency curve, and the variation tendency of cut size of Monte Carlo simulation is accordance with that of the classification experiments, which is useful for probing into the classification mechanism of the turbo air classifier.However, the random mathematical model of particle classification and Monte Carlo simulation should be improved further with the consideration of interaction of particles and turbulence of the flow field.A new method is provided for predicting the classification accuracy and cut size of a turbo air classifier using BP neural network and a three layer BP neural network prediction model for classification accuracy and cut size has been established. The training samples are used to train the network and 3-3-2 neural network prediction model is obtained. Testing samples are used to verify this model and the results show that the average relative deviation between the predicted and experimental classification accuracy is 17.59%, the average relative deviation between the predicted and experimental cut size is 7.13% while the average relative deviation between the calculated and experimental cut size is 50.44%, which means the prediction accuracy of BP neural network prediction model is higher no matter the classification accuracy or cut size. In addition, Monte Carlo simulation is based on the analysis of the particle motion in the classifier to predict the partial classification efficiency, while BP neural network model is established based on the experimental data. Thus, the prediction results of Monte Carlo simulation are closer to the ideal classification, and the prediction results of BP neural network model are closer to the experiment.