Prediction of Solids Holdup Time Series of a Gas-Solid Circulating Fluidized Bed
气固循环流化床颗粒浓度波动信号的预测

A prediction model based on combination of phase space reconstruction with artificial neural network was proposed. In this model, the Takens phase space reconstruction method is used to reconstruct the attractors, which represent the system hydrodynamics from the single variable time series, and the radial based function artificial neural network is used to fit the attractors in phase spaces. Firstly, the model was verified by the Lorenz chaotic system. Then the model was used to predict the local solids holdup fluctuation in a circulating fluidized bed riser with 16 m height and 0.10 m ID. The experiments were conducted with FCC particles and superficial gas velocity ranging from 3.5~8.2 m/s and the solids circulating fluxes ranging from 50~202 kg/(m2.s). The local solids holdup time series signals were measured at 900 Hz using an optical fiber probe. The results showed that (1) the proposed model based on phase space reconstruction and the artificial neural network is a useful method in predicting local solids holdup fluctuation in gas-solid circulating fluidized beds; (2) the predictability of solids holdup in the CFB is possible only in a short-time, in accordance with that the circulating fluidized bed is a chaotic system.