Research On Prediction Of Solids Circulation Rate In An Internally Circulating Fluidized Bed With Draft Tube

The key technique for stable operating of internally circulating fluidized bed with draft tube is to control the solids circulation rate appropriately. Solids circulation rate as an key parameter in designing of internally circulating fluidized bed directly affects the residence time and voidage of the solids in the bed. A cold experimental model of above-mentioned corresponding kind was designed and established all by oneself. The effects of several factors such as gas velocities to the draft tube or annulus section, static bed height and the average size of particles in the beds on solids circulating rate were studied. The experimental result indicated that the solids circulation rate increased with increasing gas velocities to the draft tube or annulus section, and the increasing trend slowed down when the gas velocity increased to a certain level. The solids circulation rate showed an increasing trend with increasing static bed height while it decreased with increasing particle size. A correlation equation of the solids circulation rate was given. Comparison of the calculated results and the experimental date proved that the relative forecasting error of calculated results was within±18%.The predicted values of solids circulation rate using a model which was estabished according to the gas-solid two-phase flow theory agree well with the experimental data and the relative forecasting error of was within±20%. A BP Neural Network model was built to predict the solids circulation rate of the internally circulating fluidized bed with draft tube based on MATLAB neural network toolbox. Four input variables, i.e. gas velocities to the draft tube or annulus section, static bed height, particle size and one output variable were selected. 243 experimental data were taken as training and checking samples, and the effects of hidden layer number and nodes of hidden layer on predicted results were studied. It was found that the misdiagnosis rate would minimized with a relative error within±8% and an overall Mean Diversion Extent being no more than 3.09%. The property of network achieving at an optimum condition when the number of hidden layers was one and that of hidden layer nodes were 15.